78 Commits

Include renames

Author	SHA1	Message	Date
jaap-janandClaude Opus 5	16e0051e89	Draw the tags that have been storable and invisible since the domain landed ... `Tag` has been a full item kind for three commits — a table, a migration, a codec, a merge, a cipher — and `HostSecret.TagIds` has merged per tag so two people tagging one host both keep theirs. Nothing drew a chip. The tags a client could store were ones nothing here could see. Chips on host rows, both heads, from names resolved through the tag list rather than ids: a tag that does not resolve is left out rather than drawn, because it means the tag was deleted elsewhere or belongs to a vault this session cannot read, and a host with one chip fewer is the honest answer where a host wearing a GUID is not. The id stays on the host, so the chip comes back if the tag does. The picker is chips that toggle, matching the chips on the row behind it. A list of names to tick would make the user match an entry to a chip they can see two inches away. The box under it creates a tag and puts it on straight away, because that is when a tag is usually wanted — while tagging a host and finding it does not exist yet. Unlike every other field in that editor it writes to the keychain immediately, since a host can only name an id that exists; cancelling therefore leaves the tag behind, which is honest rather than hidden. A name that already exists is used rather than repeated: two tags called "staging" are storable and must stay storable, because two people creating one offline is how it happens, but typing it into a box beside a chip of the same name is a slip. Renaming and deleting needed a home, or the picker fills with names nobody uses and never empties. That home is a TAGS category on the keychain screen, where every other item kind is managed — and renaming is the whole reason a tag is an item rather than a string repeated inside twenty payloads: it is one write, and no host is touched. The delete confirmation counts the hosts wearing it, which is the difference between a tidy-up and losing a filter somebody relies on. The desktop host editor now scrolls, and that is not a tidy-up. A picker's height is a chip per tag in the keychain, wrapped, so somebody with fifteen tags has an editor half again as tall as somebody with three; no fixed height holds that, and trimming other fields to buy room only moves the failure to whoever has sixteen. The layout suite caught it the moment its seeder grew tags — which is why the seeder now creates ten rather than three, enough to drive the pane onto its cap so the capped shape is what gets measured rather than one no real keychain produces. The cost is named where it is paid: the harness skips anything inside a ScrollViewer, so from here it certifies that pane fits the column rather than that every field in it does. Two smaller things fell out. Five buttons overflowed the keychain header by a few pixels, so GENERATE lost the word KEY — its tooltip carries what the word did. And TotalItemCount had been counting keys and credentials while ALL showed four kinds; it counts all five now, because a number under a chip that disagrees with the rows it opens is worse than no number. An adversarial review of this change found two defects it had introduced, both green against the full suite. NewTag filed into the "new items go to" picker while the tag list only ever holds the active vault's — so with a team vault selected a tag would be created, queued for push, reported as added, and then invisible, with no row, no count, no picker entry and nothing able to rename or delete it, because there is no active-vault switcher to go and find it with. The comment on the host editor's own create path states that exact rule; this was the one place that broke it, and NewObjectStore, whose list is likewise active-vault-only, already ignored the picker. And the tag editor was the only one of five that did not disarm a pending deletion when it opened, so arming a key's deletion and then pressing + TAG left a live DELETE for an item the user was no longer looking at, directly above the boxes they were typing into. Both are fixed, both have a test, and the first was checked against the broken version before being kept. The same review caught a doc comment that had been inserted between SnippetRowViewModel's summary and its declaration, silently taking it over. Verified by the whole suite on a clean build: 1413 tests over nineteen projects, none failing. Both heads build. The rectangles the layout suite cannot reach are phase 9 of docs/manual-checks.md. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>	2026-08-03 12:42:43 +02:00
jaap-janandClaude Opus 5	4925dcf179	Give the phone the + it was promised, and the editors behind it ... Steps 5 and 6 of docs/adding-hosts-on-the-phone.md, which finishes it. The phone can now put things in a keychain rather than only read one. The floating + arrives three designs after it was first asked for. It was refused twice on the honest grounds that it would open an editor this head had not got — and that stopped being true when steps 1 to 4 landed. It raises a sheet rather than a form, because "add" on this screen has been two operations since groups existed: a host, or a group to file hosts under. Both editors are cards in the list's own row, swapped for the list rather than stacked over it, following SnippetsScreen and for the reason written there: a form on top of the list hides what it is about. There is still no dialog and no editor screen anywhere on this head. The port and username boxes may be left empty, and the greyed text inside each is what the host will actually use — following the group picker as it moves, so the form says what leaving a box blank will do rather than making it a guess. The back gesture gains a guard above the switch rather than a case inside it. An editor is not a screen and has no entry there, and it is strictly nearer: the sheet sits over the editor's own screen, so back lowers whatever is topmost before it considers moving between screens. A group editor needs a way in, and the phone draws no groups panel — so the heading in the host list gains a pencil. A + that adds groups with no way to correct one is the same strange thing to ship as one that adds hosts with no way to correct one, and worse: a group's defaults are what every host beneath it falls back to. Scouting step 6 turned up a bug worth more than the prose it was looking for. ReloadGroupsAsync read the active vault alone, which was a cosmetic limitation while a group carried only a name — a teammate's host showed under UNGROUPED and nothing else was lost. Since a group began lending a port, a username and a binding, the same omission silently dropped all three: that host would dial 22 as nobody while the machine is on 2222 as deploy, with nothing on screen saying why. The resolution map now spans every readable vault. The editable list stays narrow, because a row shown across vaults needs a vault id for rename and delete and two vaults may hold groups with one name — but the map needs none of that, since it is only ever asked what an id says. An adversarial review of this change found something I had written a comment denying. The + was a child of the outer Panel rather than of the list's row, so it anchored to the bottom of the screen and sat squarely over the new EDIT button and the lower edge of CONNECT — and because a Panel hit-tests its last-declared child first, tapping the right end of EDIT would have raised the add sheet. The comment beside it claimed it was in the list's row precisely so that could not happen, and manual-checks 8.1 names that exact failure. It is in Grid.Row 2 now. The same pass caught a connect bar that was disabled rather than hidden while its own comment said hidden, a scrim that would have flashed its default pressed chrome across the screen, a fab comment claiming an accent fill no other button had when Button.primary has it, and a back-gesture remark naming two view-model properties that do not exist anywhere in the repo. That last one came from this plan, which invented IsVaultsSurface and IsConnectionsSurface, and pointed step 6 at a ConnectionsScreen.axaml that has never existed. Both are corrected in the plan rather than quietly worked around, along with a fourth "groups are flat" site in README.md that step 1 missed. The phone's rectangles are not measurable and structurally never will be: the layout suite is net10.0, the Android head is net10.0-android, and Avalonia's application is a one-shot process global — so a second head cannot share the process even if the reference were possible. Everything the sheet and both editors bind to is shared and is tested headlessly here; the pixels go to docs/manual-checks.md phase 8, nine checks, which is where this project already sends what it cannot assert. Tags remain the one thing asked for that did not ship. Tag is a full item kind and TagIds merges, encodes and resolves, but no screen draws a chip or offers to add one, so the tags a client can store are ones nothing here can see. Both editors carry the set through a save untouched so a client that can set them does not lose them. Recorded as half-shipped in docs/design-import-gaps.md rather than quietly left out. Verified by the whole suite: 1402 tests over nineteen projects, none failing. Both heads build. Nothing seen on a display — phase 8 is what that costs. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>	2026-08-03 11:30:00 +02:00
jaap-janandClaude Opus 5	b7335743d9	Make a host take what its group lends it, everywhere it is read ... Step 4 of docs/adding-hosts-on-the-phone.md. The domain could resolve a host against its groups; nothing asked it to. This is the wiring, and it is mostly one change repeated: read the resolved host, not the stored one. TryBuildAuthentication and TryBuildConnectionRequest now take the resolved value beside the stored one, which is where group context was being lost. It is the only authentication resolution in the product — both heads and both transports come through it — so a host inheriting its binding would otherwise have been offered a password prompt on every screen at once. The credential-username fallback becomes three levels, and Complete still refuses an empty username, but now only after the chain has been walked; refusing before it would refuse exactly the hosts inheritance exists to serve. HostRowViewModel carries its ResolvedHost, resolved once when the list is built. Address, Authentication and Dialled read it, so a row cannot disagree with itself about what it dials — and MainWindowViewModel.Rank searches Address, so a host inheriting 2222 that displayed 22 would have been unfindable by the port it actually answers on. HostsBoundTo counts over the resolved binding, which is the difference between a warning and a silence: a key bound once on a group and inherited by twenty hosts named nobody, would have been deleted, and would then have refused all twenty at connect time. HostFields.From is answered by a refusal rather than by threading a group list through the sync engine. A relay host may not inherit its port. The reason is stronger than the convenience: a plaintext column is a derived duplicate the client supplies when it pushes *this* host, so an inherited port would make it depend on another item — editing a group would change what the relay dials for every host beneath it, except that nothing re-pushes those hosts, so the server would keep dialling the old port until each was next touched for some unrelated reason. A stale wire on the relay path connects the user to the wrong service. The editor distinguishes unset from explicit in both directions. An empty port box means "take the group's" and shows what that will be as a placeholder, following the group picker as it moves — a pre-filled 2222 would have been indistinguishable from one the user typed, and saving would have pinned it. The authentication picker gains a fourth entry, offered only to a host in a group, because for an ungrouped host it would behave exactly like the first. Which found a real defect while the tests were being written. Filing an ungrouped host into a group silently pinned it to a typed password: the picker had no "Inherit" entry when it opened, so it sat on "Password (ask each time)", and saving wrote that as a decision — the host would have been pinned to a prompt nobody asked for and the group's key would never have reached it. Two guards now: the picker is rebuilt when the group changes, and BuildHost writes AsksForPassword only for a host that had the alternative on offer. The group editor is here too, and the plan never assigned it a step. Without it no group can carry a default, so every line above would have been unreachable. It grows a parent picker that leaves out the group itself and everything beneath it — a courtesy rather than the guarantee, since a cycle assembled from two offline re-parents was never offered that list — and three defaults beside the name, each of which may be left empty because "lend nothing" is an answer. Tags are stored and not editable. TagIds merges, encodes and resolves; no screen can set one yet, and the editor carries the set through a save untouched so a client that can set them does not lose them to somebody editing a port. Eight new tests, and they dial. That is the point of them: a resolved value that never reaches SshConnectionRequest is a label, and every one of these failures would be silent — a host connecting to the wrong port, or being asked for a password it does not need, with nothing on screen admitting it. Verified by the whole suite: 1390 tests over nineteen projects, none failing. Both heads build. Nothing on the phone has changed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>	2026-08-03 10:41:28 +02:00
jaap-janandClaude Opus 5	8c04ba60b0	Build the three things the phone's + needs, before the + exists ... Steps 1 to 3 of docs/adding-hosts-on-the-phone.md: the domain half. Nothing on either head has changed, which is deliberate — the plan orders these first because everything the editors will bind to has to exist and be merge-safe before a screen can offer it. HostGroupSecret gains a parent and four defaults, and the codec gains the version rule it never had. It stamped CurrentSchemaVersion unconditionally, which was harmless with one field and one version and stops being harmless here: upgrading one machine and renaming any group would have made that group uneditable on every machine still on the old build. It now emits the lowest version that loses nothing, so a flat group with no defaults still encodes at version 1, byte for byte, pinned against a literal. Tags become a real item over the reserved slot. Secret, codec, merge, cipher, repository, both registries, the EF entity and a generated AddTagItem migration. TagCipher names AadResourceType.Tag as a constant rather than casting the wire type, because Tag is 5 on the wire and 8 in the crypto enum and 5 there is Credential — a cast would seal every tag under the resource type for a password, encrypt and decrypt perfectly on the machine that wrote it, and only fail when another implementation refused the item, by which time the AAD is frozen into stored ciphertext. HostTag stays reserved and unused: the one thing the join buys over a set on the host is bought instead by merging TagIds per id. HostSecret grows TagIds and Port goes nullable, which is the change with the widest blast radius and the only one that loses an item rather than locking one. A host with no port of its own omits the property, an older build reads int Port as 0, and TryValidate refuses it — unreadable rather than read-only. That cost is confined to hosts which actually inherit, because the version is a maximum over the fields present; the alternative, writing 22 into every host, is the lie inheritance exists to stop telling. One decision the plan did not specify. "Three states where there were two" is four — key, credential, typed password, or the group's answer — and two nullable ids carry three. Naming neither id now means inherit, so AsksForPassword says "a typed password even under a group that lends a key" out loud. Only true is ever written and a decoded false folds back to null, so a host that never touched it encodes as it always did. Nothing already stored changed meaning: no group could lend a binding before this build, so every existing host resolves exactly as it did. HostInheritance is the resolver, and its visited set is load-bearing rather than defensive. Two clients can each re-parent A under B and B under A while offline; the merge sees one item against one item and the server sees ciphertext, so nothing upstream can refuse the pair. With inheritance the chain is walked at connect time, so an unguarded cycle is not an undrawable sidebar — it is a shell that never opens. Stopping at the first repeat degrades it to a group that reads as a root, and clearing the parent is the repair. A tag set turns out to be the one field on a host that can never ask the user anything. TagSet.ToIdMap keys by the value, so no key can hold two values, so the both-sides-moved-differently branch of the keyed merge is unreachable — asserted over the whole eight-row matrix. The conflict loop is kept anyway, because that proof is one edit from ceasing to hold and what it would cause is a discarded tag nothing records. Three guard tests failed by design and were fixed rather than relaxed: the ordered pull filter, the AAD pinning table, and the server's refusal of a plaintext parent — that last one survives with its reason rewritten, because the refusal now means "the parent is not the server's to hold" rather than "there is no such thing as a parent". The prose that said groups are flat is rewritten in all four places it appeared, not deleted. The five view-model sites that read Port directly now go through the resolver, which is a down payment on step 4 rather than the whole of it. HostFields.From still emits the stored port, and that is the one remaining place where an unresolved read would be a wrong wire rather than a wrong label. Verified by the whole suite: 1382 tests over nineteen projects, none failing. Both heads build. Nothing seen on a display, because nothing on a display has changed yet. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>	2026-08-03 10:21:02 +02:00
jaap-janandClaude Opus 5	359087f1ce	Give the tabs their hover back, and the plus the shape its comment claims ... Two things v2 broke in the last commit, both found by reading resolved brushes rather than markup, and both the same mistake: Avalonia has no specificity, so the later declaration wins, and a rule that restated the base instead of excepting from it went in below the rules it was supposed to be underneath. Making a tab a pill gave it a Background of its own, which the flat tab it replaced never had. That one detail moved where the hover has to live. `Button.flat:pointerover` is declared far above and had been supplying it; the moment `Button.tab`'s template rule set a Background, it won, and every tab in the strip stopped answering the pointer. Silently — a tab that no longer lights is not a crash and not a layout change, and nothing in a suite that measures heights and reachability can see it. The `+` lost more than that. It sits below as an exception — no outline, because it is not one of the things being chosen between — and a `.tab` rule declared after it was overriding the exception itself. It drew as a filled, outlined pill identical to a tab, contradicting the comment directly above it. So the base pill and its hover come first now and the exceptions follow, which is the order the rest of this file already uses and the order the Border.rowmark note further down was written about. The `+` clears the fill as well as the border, because an exception to a rule that sets both has to say both. The titlebar's search box had the same shape of error in geometry rather than colour. The design draws it at exactly 380 and centred, and stating that as a Width on the inner Border is what made it wrong: the button around it is free to shrink when the account name or the vault chip beside it is long, and a Border that will not shrink with it arranges outside its own parent — over the name on one side and over the window buttons on the other. MaxWidth on a stretching button gives the same 380 whenever there is room and gives way when there is not. The regression test is the point of this commit rather than an afterthought. It is the only test in that suite that reads a brush, and the gap it fills is exactly the one these two went through: everything else measures rectangles. It hovers a tab through the real input path and asserts the fill changes, then asserts the `+` is neither filled like a tab nor outlined like one. Checked against the broken ordering before being kept — it fails there and passes here, which is the only thing that makes a regression test worth committing. Verified by the whole suite: 1310 tests over nineteen projects, none failing, the layout suite now 70 cases. Both heads build. Still nothing seen on a display. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AZE3u99BNt6LzgTC5jhbz2	2026-08-02 19:54:37 +02:00
jaap-janandClaude Opus 5	3627021420	Give the desktop the second design too, and the window the size it now needs ... The desktop v2 design is the other half of the one the phone took last commit, and this is its chrome: a 190-pixel labelled sidebar where the 54-pixel icon rail was, a titlebar with the search box centred in it, and session tabs drawn as pills. The palette was already here — it is shared, and moved when the phone's did — so what this changes is shape rather than colour. **The window's minimum grew, and by exactly what v2 added.** The sidebar is 136 wider and the chrome 14 taller, so 880x560 became 1016x574. That is not a round number somebody liked: it leaves every screen the same 826x464 it was designed against, which is the arithmetic the layout suite is built on. Four of the tables stop fitting at 690 wide, so widening the sidebar and leaving the window alone would have broken them somewhere no test was looking. LayoutHarness carries the new constants and the suite still passes at the minimum, which is the whole reason it exists. The rail's five-character abbreviations are gone with the width that caused them — PINS and SNIPS are Pins and Snippets again — and each row gains a glyph and a count. A count is drawn only where one is real, so SFTP, Logs and Preferences show nothing rather than a zero: a transfer queue's depth is not how many files a screen holds, and a log has no total until it is read. The count beside Pins is the vault's own, not the Pins screen's VisiblePins, which is the filtered list and would have made the sidebar count whatever somebody had just typed into a filter box on another screen. Teams has no count for a related reason: they are read from the server when that screen is opened rather than on unlock, so a number there would read 0 until somebody had already been to look. One colour moved with it, finishing what the repalette started: the live-session summaries on the unlock and sign-out cards were Info, so the two heads disagreed about a fact the phone paints green. They match again. **Buckets became a destination rather than a mode**, which is what the design draws and what the phone already does. The HOST / BUCKET pair inside the files screen is gone; ShellScreen.Buckets draws the same TransfersScreen with the other picker, and the sidebar entry is what sets it. That also settles an old disagreement rather than merely moving it: TotalItemCount is keys plus passwords and excludes buckets, so the number beside the keychain used to disagree with the list under it, and now counts what that screen shows. There is one session behind both file destinations, so asking for the other kind while something is open is refused rather than obeyed — and refusing means staying put. An earlier turn of this had it move anyway and only decline to switch the picker, which put the S3 entry in the sidebar over a screen still listing an SFTP host: two pieces of chrome disagreeing about where you are, which is worse than the navigation simply not happening. The message that says so goes to Transfers.Status, which turned out to be drawn in the same grid cell as the connected chip — survivable while it was mostly read before connecting, and not once a refusal reports itself there. It has its own column now. The design has nine entries' worth of screens and draws five. Pins, Teams, Import and Preferences are built, working screens, so they keep their entries — the sidebar is labelled now and has the room, and dropping an entry would have stranded a screen rather than simplified anything. The Team vault card the design pins to the foot is not drawn: it is a second route to a screen already in the list, carrying a seat count nothing here produces. **The status bar survives the design that deletes it**, cut down to one thing. Two of the three facts it carried moved into the titlebar with v2 — the sync word is beside its dot and the shortcut hint is inside the box that uses it — so those are gone from it rather than printed twice. The third is Vault.Status, the only channel this application has for saying a save failed or a merge picked a winner. The design is a mock-up of an afternoon that goes well and has nowhere to put a sentence like that; dropping the bar would have meant dropping the sentence or repeating it on nine screens. What v2 draws and this does not is in docs/design-import-gaps.md, and it is the same list as the phone's for the same reasons: the forwarding screen and both its chips, the host detail's fingerprint, tags and last-session cards, the keychain's rotate button, the logs' FOLLOW pill and severity filters, and the session footer's latency. The terminal is not inset behind a rounded frame either — it is a native child window that composites above everything Avalonia paints, so the frame would clip nothing, which is the same answer the phone gave. **The light theme is not built.** Its accent is #6D5AE6, a different hue rather than a tint of the dark one, so it needs every colour doubled, a variant to switch on, the renderer's own page switching with it, and contrast checked twice. That is a piece of work rather than a setting, and it is separable from the layout — which is why this commit is the layout. The screens themselves are restyled through the shared vocabulary rather than rebuilt: corner radii, chips, cards and the accent's ink, all in App.axaml, so every screen moves at once. Their layouts are left alone deliberately. The design draws read-only detail panes and these screens carry the editors and forms it has no equivalent of, so replacing a layout with the mock-up's would have lost the half that is actually used. Verified by the whole suite: 1309 tests over nineteen projects, none failing, including the 68 layout cases that stand up real Avalonia and measure every screen at the new minimum. Both heads build. Not run on a machine with a display — see docs/manual-checks.md for what wants looking at. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AZE3u99BNt6LzgTC5jhbz2	2026-08-02 19:29:21 +02:00
jaap-janandClaude Opus 5	c00e5dbc5c	Let the terminal's text be made bigger, and remember how big ... Taking pinch-zoom off the phone left nothing in its place, and there was nothing on the desktop either. This is the replacement, and it is deliberately not the thing that was removed: zoom scales what has already been drawn, so the remote goes on wrapping to a width that is no longer on screen. Changing the font size refits the grid and reports the new column count, so the far end is told it has fewer columns. That round trip is the feature. The size is one number, owned by the shell. It has to be, for two reasons that pull the same way: it must survive a relaunch, and it must be reachable from a phone that has no Ctrl key to press. So the page asks and the host decides — a signed step over a new client opcode, answered with a size over a new server opcode. The phone's buttons and the desktop's chords arrive at the same place, and a size set by either is the size both remember. Stored in settings.json beside the cache rather than in it, and that is not laziness about a migration. The cache is encrypted and unreadable until a vault is unlocked, and the first terminal of a locked launch needs the size already. Nothing secret may go in that file; ClientSettings says so out loud, because the next person to add a preference is the one who needs to read it. Where it is reachable from differs per head, and only here. The phone gets A− and A+ on the connection line — not in the accessory row, which scrolls, and a control that fixes unreadable text must never be the thing that is off-screen. The desktop gets the three chords every terminal emulator has, answered by the page while a terminal has focus and by the window when it does not, plus a row in preferences that shows the current value and names the chords rather than replacing them. Someone whose terminal is too small to read is not in a position to go looking. Clamped 8 to 32. Below eight a monospace grid stops being legible and becomes a texture, and every column of it is still a column the remote is being told exists; above thirty-two a phone in portrait has too few columns to hold a prompt. The buttons disable at the ends rather than accepting presses that do nothing, which on a terminal reads as the application having stopped responding. The preferences screen's header comment claimed none of the design's terminal settings could be saved, and listed the three things that were missing to make one work. All three now exist, so it says which one is real and why the other five still are not. Verified with the protocol suite — including that the step byte round-trips signed, since read unsigned a step down arrives as 255 and clamps to the largest font, making "smaller" do the most dramatic available version of "larger" — a data-plane test that the chord is heard with no session registered, and five shell tests: the default matches the renderer's, both clamps hold, reset works, and a size chosen in one shell is there in a second one over the same profile directory. Layout suite and both heads build. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>	2026-08-02 13:15:11 +02:00
jaap-janandClaude Opus 5	ddf0dd6a2b	Stop the transfer tests depending on which thread ran them first ... Every test in TransferQueueingTests failed on the Linux runner with "The calling thread cannot access this object because a different thread owns it", and none of them had anything to do with the commits in that run. The class drained its rows through Dispatcher.UIThread.RunJobs(). That dispatcher is process-wide and belongs to whichever thread touched it first, and xunit runs each test class as its own parallel collection — so the moment a runner scheduled another class onto that thread ahead of this one, all nine died inside DispatcherOperation.Execute having asserted nothing about transfers at all. It passes locally and fails on a machine that schedules differently, which is the whole of why this took a CI run to find. TransfersViewModel now takes the poster it marshals through, defaulting to Dispatcher.UIThread.Post — the seam VaultViewModel's clipboard already is, for the same reason: a view model that reaches a process-wide UI object directly makes every test of it depend on a thread it does not choose. No head passes the parameter, so nothing about the running application changes. The test supplies a queue of its own and drains it, which is the same shape the dispatcher gave it. A poster that ran the action inline was tried first and is wrong: the transfer queue raises Changed from its pump thread as well as from the call that enqueued, so inline execution has a background thread adding rows to an ObservableCollection while the test reads it — it passed once and then failed a different test on the next run. Draining keeps every mutation on the thread doing the asserting, which is the one thing the dispatcher was providing that was worth keeping. One test added for the property that broke: queueing is reachable from any thread and must not care which. The class as a whole guards the seam — remove it and nothing drains, so every assertion about a row fails. Verified by reproducing the failure first: a throwaway probe that touched the dispatcher on one thread and posted and drained on another produced exactly the CI message. Then six consecutive Release runs of the app suite, all green, plus the layout suite, which builds a TransfersViewModel of its own. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>	2026-08-01 21:43:21 +02:00
jaap-janandClaude Opus 5	73c7e2a1e3	Point a release build at the hosted server, and a debug build at the clone ... The shipped default was http://localhost:5233, which is the address the API serves under dotnet run and a machine an installed application is not running. Somebody who installs a release and accepts the field unread is signing in to nothing. Two defaults now, because the two audiences never overlap. A release build offers https://ssh.dodotech.cloud, so a first launch needs no address typed at all. A debug build keeps the loopback address, and that half matters as much: shipping the hosted address into a clone would point every development launch at production, and sign-in is the call that provisions an account there. The remark carries the reason the schemes differ, since the pair now looks like an oversight rather than the deliberate thing it is — the API's first launch profile is plaintext on 5233, and an HTTPS client meeting a plaintext port reports a TLS failure that reads like a certificate problem. The test spells out both branches rather than asserting the constant, which would pass however it were edited. What it is really guarding is that a release never ships a developer's loopback address and a debug build never points a clone at production, and it can only guard those by naming them. Verified by running the shell suite in Debug and in Release. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>	2026-08-01 21:12:00 +02:00
jaap-janandClaude Opus 5	338c1a8647	Move the plaintext exemption out of the test body it made too long ... MA0051: TheWholeSlice reached 68 lines against a limit of 60, because the last commit put a nine-line paragraph and a five-line call in the middle of it. dotnet format --verify-no-changes runs the analysers, so the build stopped there and never reached the two fixes that paragraph was explaining. The explanation was worth keeping and the place was wrong. It is a fact about one call, not about the slice, and this file already keeps its steps in named methods under a "Steps" heading. SignInToTheStackAsync now holds both, which leaves the test body reading as the sequence it is meant to be — sign in, enroll, unlock, write, read elsewhere — rather than a sequence with an essay in it. Nothing about the behaviour changed: same call, same configureOidc, same exemption claimed by the same single caller that starts the provider it is talking to. I should also say how this reached CI, since the answer is not that it was hard to catch. I ran the format gate locally before the last push and read the exit code of a pipeline it was piped into rather than the tool's own, so a failing command reported as passing. Run again against the tool's exit status it is 0, and the end-to-end suite still passes in the Alpine container that reproduces the runner. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>	2026-08-01 14:55:32 +02:00
jaap-janandClaude Opus 5	cf1a321d1e	Pin the font the layout suite measures, and let the slice say it means plaintext ... Two failures left on the runner, with nothing in common except that both only appear on a machine unlike the one anybody develops on. The runner is Alpine, musl, inside a container, with no fonts installed at all — and that combination is now reproducible locally, which is how these were fixed rather than guessed at. Both are verified by running the suite in it. The layout suite had two causes stacked, and the first hid the second completely. Missing libfontconfig stops libSkiaSharp loading, which the last commit fixed and which then revealed the real one: Avalonia takes its default font family from the platform, and on an image with no fonts there is no answer, so FontManager throws "Default font family name can't be null or empty" inside AppBuilder.SetupUnsafe — before a single test body runs, for all sixty-eight of them, naming none of their subjects. WithInterFont does not prevent it: it registers a collection without nominating a default. HeadlessApp's own comment already claimed it measured "the same Inter font the application registers", which was an intention the code never carried out. Both heads now name it, through FontManagerOptions.DefaultFamilyName. That is worth more than getting CI green: a suite whose entire job is measuring text was taking its metrics from whatever the machine happened to have — Segoe UI here, DejaVu there — and reporting the two as one number. It also means the application uses the font it has been shipping and declining to use since it first referenced the package; almost nothing moves visually, because App.axaml already sets MonoFont on essentially everything that draws. The end-to-end slice was the product being right and the test leaning on an accident. ServerConnection permits an http authority only when it is loopback. Testcontainers reports the host a container can actually be reached at, so running the suite directly gives localhost and passes, while running it inside a container gives the bridge gateway 172.17.0.1 and is refused — correctly, since a client that accepted plaintext metadata from a routable address would be a weakness for everyone who is not a test. Loosening that rule was the wrong repair. The slice now passes configureOidc and says out loud that it accepts plaintext from the Keycloak it started itself. Verified by reproducing the runner rather than approximating it: dotnet/sdk:10.0-alpine, musl-x64, fc-list returning zero, the docker socket mounted so Testcontainers resolves the gateway exactly as it does in CI. The whole solution passes there — 19 suites, 0 failures, 4 skipped — and the end-to-end failure was confirmed causal by reverting only that one file and watching it fail again in the same container. The layout suite also still passes on a Fedora desktop with 595 fonts, so the two agree now. Not verified on Windows, and it should be said plainly rather than left to be discovered: pinning the family changed the measured metrics there too, so a tight layout assertion could have moved. platform-flags.md records that, and corrects the entry the last commit added — "libfontconfig, and nothing else" was true of the container it was tested in and false of the runner. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>	2026-08-01 14:48:30 +02:00
jaap-janandClaude Opus 5	43d76d0f2d	Let the suite run on Linux, and fix the three things that stopped it ... The pipeline finally reached the tests and found four failures. None was the pipeline's, and only one of the four was a test being fussy about a platform rather than telling the truth about one. The local pane's roots bar was the real bug. LocalDirectory.Roots built it from DriveInfo.GetDrives on every platform, and its own summary — "the drives on Windows, and the root elsewhere" — had been describing an intention rather than the code for as long as nobody ran it off Windows. On Unix that call answers with every mount the kernel holds: /proc, /sys/fs/bpf, one per installed snap, /run/user/1000/doc, some forty on an ordinary laptop. The transfers screen draws a button per root, so the bar ran to about five thousand pixels inside an eight-hundred pixel window. Anybody running the Linux build has been looking at that. Filtering GetDrives is not the fix and the comment now says why at length, because it is the obvious thing to try: DriveType answers Fixed for / and /home and equally for every squashfs snap, for efivarfs and for tracefs, while /boot/efi comes back Removable, and DriveFormat would need a hand-kept list of every virtual filesystem Linux might grow. So Unix now names what somebody would want instead of subtracting what they would not — the root, their home, and whatever is mounted under /run/media/<user>, /media, /mnt or /Volumes. Anything else is still reachable by navigating from /, which is what the pane is for. Windows is untouched. ClientPathsTests looked for "odoSSH" in the profile directory. ClientPaths spells it DodoSSH on Windows and dodossh on Unix deliberately, one per platform convention, and that substring was clever enough to survive either spelling of the leading D while still only ever matching one of them. Now OrdinalIgnoreCase. WhyTheWindowItselfIsNeverShown asserted a COMException with HResult RPC_E_CHANGED_MODE, which is WebView2 refusing an MTA thread — a Win32 component raising a COM error. On Linux the adapter is a different implementation with no apartment to disagree about, so showing the window works and Should.Throw catches nothing. Skipped there rather than loosened to accept both outcomes: the assertion is the documentation in that test, and one that passed everywhere would have stopped recording the constraint it exists to record. The fourth was CI's alone, and the diagnosis is the useful part. All 69 layout tests failed on the runner while 6 failed here, which looked like missing fonts and was not: Avalonia's headless renderer is Skia, libSkiaSharp.so links against libfontconfig, and without it the suite dies in HeadlessUnitTestSession with a TypeInitializationException on SKImageInfo naming none of its actual subjects. The job installs the one library now. Verified in a container where fc-list returns zero and the suite passes regardless, because the application carries Inter itself — fonts were never the problem, only the thing that would have looked for them. The whole solution now passes on Linux: 19 suites, 1295 tests, 0 failures, 4 skipped, the end-to-end Testcontainers suite included. README and platform-flags.md said testing was Windows-only, which CI now contradicts on every push, so both say what is true instead and the two findings are written down where the next person will look for them. macOS is still untested and now says so on its own rather than hiding inside "not Windows". Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>	2026-08-01 14:28:06 +02:00
jaap-jan	4300d917a8	Stop making people wait for a handshake, and give the host list a pointer ... Connecting held the vault's busy gate, which meant a window that did nothing visible for as long as a machine took to answer — and against one that is merely asleep, that is the whole timeout. The gate is gone from that one command. A tab now appears in the strip in the same turn as the click, carrying "connecting…" rather than a pane, and the terminal's rectangle draws a card naming the host and the address being dialled. Every other screen stays usable, and two connections can be in flight at once. That splits the vault's one connection event into three, carrying an attempt id, because "which tab is this about" can no longer be answered by "the most recent one". The id also buys the two kinds of not-connecting their different endings: a refusal stays in the strip as a tab holding its reason, since by then the user is quite likely three screens away and a status line they are not looking at is not where a failure should end; a host key question takes the tab away and puts the window back on HOSTS, because the prompt is drawn there and a tab claiming failure would be competing with the thing about to resume it. ConnectAsync takes no CancellationToken any more, and that is load-bearing rather than tidying. A [RelayCommand] over a method that takes one generates a command that cancels the previous execution's token on every invocation — so asking for a second machine silently abandoned the first, measured as the first tab disappearing with "Cancelled." the instant the second was asked for. Giving up on a connection is closing its tab, and a session that lands after that is adopted rather than dropped: a shell running with nothing naming it cannot be closed at all. A tab is marked active on IsShowing rather than IsSelected. The selection survives navigating away — that is what makes the strip a way back to a terminal instead of a way to lose one — so a tab lit while preferences filled the window was a second "you are here" mark pointing at something nobody could see. The nav rail's own entries have always made this distinction. The host list grows the two gestures it looked like it already had. A right click selects the row under the pointer before opening a menu of Connect, Edit and Delete — the menu is on the list rather than in the item template, so its entries are the vault's own commands and not a row's, and it is cancelled outright over a group heading. Dragging a host onto a heading files it there, onto a host files it beside that one, and onto UNGROUPED takes it out of a group; the write is one field of one host through the same repository a save uses, refused while the editor is open because a drop is a gesture on the list and not on a half-typed form. Clicking a result in the palette connects, which is what a list of hosts under a search box looks like it does. It went through the shell's own command, so the pointer and Enter take one path. And the files screen's two pickers followed the vault's lists once, at unlock: a host or a bucket created afterwards could not be picked until the keychain had been locked and opened again, with nothing on screen explaining why the machine plainly in the host list was missing. They follow the collections now, re-finding the selection by id across the rebuild a sync pass causes every minute. 165 shell tests and 69 layout tests green, including the connecting tab, both failure endings, two connections at once, a connection in flight across a lock, and the right click acting on the row under the pointer rather than on the selection. The drag itself is in docs/manual-checks.md with the rest of phase 7 — headless Avalonia has no platform drag, and a test that claimed to have dropped something would pass while confirming nothing.	2026-07-31 22:59:33 +02:00
jaap-jan	7a3a521c59	Give the phone the rest of its screens, and a way in ... All seven screens of the design, plus the two it does not draw because it starts at an enrolled phone: naming a server, and choosing a passphrase. The five states docs/android-port.md worried about losing at 360dp are all here and none of them softened. The changed-key refusal is a full-screen panel rather than a bottom sheet, because a sheet is swipe-to-dismiss by convention and that screen must have no way forward. The recovery code raises FLAG_SECURE for its own state and lowers it afterwards, so the sentence about screenshots is true rather than decorative. The delete confirmations keep their counts and replace the row in place. Signing in works, and the seam it needed is worth more than the implementation: IAuthorizationCallback now sits between OidcClient and the loopback listener, so the two heads differ in where the response arrives and in nothing else. PKCE, the state check, discovery, the token exchange and the key binding stay one implementation — a second OIDC client would be a second place for a security bug to live. The phone registers a private-use scheme with the system rather than binding a loopback port, which on a shared device any other app can do first. The accessory key row needed TerminalWorkspace.SendInputAsync: ordinary typing goes from the renderer straight down the socket, and there was no way in for the keys a software keyboard does not have. Ctrl latches, because one thumb cannot chord, and the latch is drawn — a modifier that is on and does not look on is how somebody sends ^L to a database prompt believing they typed an l. 597 client tests green, including two new ones for the input path and one for the terminal surface command. Nothing has run on a device.	2026-07-31 21:43:11 +02:00
jaap-jan	2caedd93ff	Merge branch 'main' into the Android head ... Main grew the screens the host-management plan called for — hosts, pins, snippets, logs, import, teams — plus the ObjectStore and Import projects behind two of them, and moved WindowsDeviceKeyStore into the desktop head's Platform folder. Five of those view models landed in a directory this branch had already moved, so they join the rest in DodoSSH.Client.Shell: git spotted the rename and put them there, and the namespaces followed. Shell picks up ObjectStore and Import as a result, which the Android head then gets transitively and will use neither of at first — scoped storage means there is no ~/.ssh/config to import, and file transfer is out of its first scope. Desktop suites green at 155 and 64.	2026-07-31 21:03:22 +02:00
jaap-jan	fe9d7fc289	Give DodoSSH a phone, and a shared shell for both heads to drive ... The Android head from docs/android-port.md, taken as far as its step 6. Step 3, the spike, is answered and its throwaway screen is gone: libsodium.so and libe_sqlite3.so are both in the arm64 APK, so NSec resolves its native half on Android despite shipping no Android build, and the local cache opens. Two findings the audit could not have had: Avalonia.Controls.WebView only ships net10.0-android36.0, which settles the open "which Android versions" question at targetSdk 36; and Android has blocked cleartext HTTP since API 28, so the terminal renderer needs a network security config scoped to 127.0.0.1 or the WebView loads nothing. DodoSSH.Client.Shell is new and is why the phone can exist: the view models, the terminal renderer files and the palette moved there so both heads drive one state machine and draw from one set of tokens. The desktop head is otherwise untouched and its 144 tests still pass. The platform pieces behind interfaces that already existed: the profile directory from filesDir, a device key wrapped by a StrongBox-backed key that a fingerprint releases, and a foreground service so a shell outliving a vault lock stays true on a platform that stops backgrounded processes. Sign-in is deliberately absent rather than approximated. It needs an app link, because reusing the desktop loopback listener is the attack RFC 8252 section 8.3 names.	2026-07-31 20:58:48 +02:00
jaap-jan	5cbda59a34	Merge branch 'main' into claude/host-management-ui-plan-7f20ab ... Seven files needed a hand. Most were two branches adding something in the same place, but three were one branch changing what the other had moved or renamed, and those are the ones worth reading. The shell keeps both new fields and both constructor lines: the connection recorder this branch built and the teams view model main did. Where main put a teams load inside OnScreenChanged, it now sits beside the logs refresh rather than inside RaiseSurfaceState — this branch extracted that notification block and it is called from two properties, so a screen-specific side effect in there would fire on every terminal switch as well. Main gave four row types a vault id and a vault name, and this branch had moved one of them — KnownHostRowViewModel — into its own file when the pinned keys became a screen. Git resolved that as "deleted here, modified there" and took the delete, which compiles as long as nobody looks: the moved copy still had the two-argument constructor and the call site had grown to four. Carried over by hand, along with the ordering the pins list now does on them. The status line's quiet rule was the subtle one. Main extracted it into IsWorthReporting; this branch had changed the same condition to read item counts rather than raw ones, because every user action queues a log entry a moment later and this machine reads its own entries back on the next pull. Take main's structure and the merge builds, passes, and silently restores a bug this branch existed partly to fix — every save's message overwritten a second after it appears. The method now reads PulledItems and PushedItems, with the reason in its remarks. Two conflicts were prose that had gone stale rather than code. The keychain screen's comment said team vaults are refused by the server's access service, which was true when it was written and is not now; main's replacement stands, in this branch's vocabulary. The design-gaps row for groups was claimed by both — real host groups here, per-vault headings there — and they are different things, so both rows stay and the difference is stated: a group is a shelf the user chose, a vault is who can read the item. One defect the tests found and the compiler could not. Generating a key opens the same editor as pasting one, but not through NewKey — so it never set the target vault main added, and a generated key was filed into whatever vault was edited last, or none. Both key-generation tests failed on it. Fixed where the editor opens, with the reason recorded there. One gap is left deliberately and is written down rather than half-built. Hosts, keys, credentials and pins are read across every vault this session holds a key for; groups are read from the active vault alone, so a host a teammate filed shows under UNGROUPED. Nothing is lost or misfiled — it is what the sidebar already shows for a group that has been deleted — but closing it needs a vault id on every group row for rename and delete, and a way to tell two vaults' identically-named groups apart under a layout with one heading per group. Both are worth doing and neither is a merge's business. It is in the remarks on ReloadGroupsAsync and in docs/design-import-gaps.md. dotnet build, dotnet test and dotnet format --verify-no-changes are all clean: 1282 tests, including the end-to-end suite against real containers.	2026-07-31 20:44:39 +02:00
jaap-jan	d07b336868	Free the terminal from the Hosts screen, and fill the room it left ... The WebView sat inside the Hosts grid, so navigating to Files or the keychain hid every open terminal and the strip that named them. A connection you had opened was invisible from four of the five screens. The window now has two surfaces rather than one: a nav rail that says which page you are on, and a terminal strip that is always there and switches the whole content area to a shell. Screen keeps meaning "which page" and never becomes a sixth kind of page, which is why this is two properties instead of one enum with a terminal member in it. Every screen lives inside one wrapper panel that collapses when a terminal is showing. That is not tidiness — the WebView hosts a Win32 child window that composites above everything Avalonia draws, so a screen left visible over its rectangle is a screen sliced in half, and this window has shipped that defect once already. One decision point, IsTerminalShowing, and a nested panel rather than five compound bindings nobody would remember to extend. The focus choreography is the part no test in this repo can see. Every reveal path now focuses in the same turn the WebView appeared, so all three of them post at DispatcherPriority.Loaded and let the native control re-push its bounds first. Going the other way had a real bug: the screen-changed branch called a bare Focus() where it had to release the keyboard from the native child, so switching from a terminal to Files silently ate the first keystrokes. Rare before this commit and the primary gesture after it. The tab strip grew a cross inside each tab, a plus that opens the quick-connect palette, and middle-click close. Nested buttons are correct here: Avalonia handles a left press on the cross and deliberately does not handle other buttons, which is exactly what lets middle-click bubble up from the cross as well as the tab. The test is PointerUpdateKind rather than IsMiddleButtonPressed, because the latter reports button state and is also true for a left press made while the middle button happens to be held. The handler is on the tab and not the strip, so the background closes nothing by construction. Plus opens the palette rather than a flyout, since a menu dropping into the WebView's rectangle may or may not composite above a child HWND and this repo does not make rendering claims it has not photographed. Everything a user reads now says keychain. The wire, the database and the cryptographic spec still say vault, deliberately: renaming those is a migration and a protocol change for a word. That split is written down rather than left to be rediscovered as an inconsistency. Four things that were squeezed into the keychain's category rail, or into nothing at all, now have screens. Pinned host keys get one, with fingerprints never truncated and a filter that matches them, because comparing what you have against what the operator published is the whole workflow; the approved date is read out of the item's UUIDv7 rather than added as a column, and says so, since it means first approval and not last use. Keys can be generated in the client, which needed the openssh-key-v1 container written by hand — there is no BCL or NSec helper, and the PKCS#8 route is unverified in the SSH library this uses. The armour carries no passphrase: encrypting it needs bcrypt_pbkdf, which is Blowfish with a swizzle, in a project whose crypto is otherwise entirely libsodium, for a protection the key's own remarks argue is redundant inside a vault. Generation fills the existing editor and stops, so SAVE stays the one thing that writes. ~/.ssh/config can be imported behind a preview that is ticked per row and writes nothing until the button; IdentityFile records the path and imports the key material only on an explicit opt-in, because reading somebody's private key into a vault is precisely the act this product exists to make deliberate. Match blocks and ProxyJump are reported rather than obeyed — one cannot be evaluated statically and the other has nothing behind it to route with, and a preview that implied otherwise would be worse than one that admits it. Files can be dragged in all four directions that are honestly available. Remote to Explorer does not ship and is not pretended to: the shell wants the bytes during the drop, which needs a virtual file and a native COM data object, outside what Avalonia offers. Note for the next person that Avalonia 12 replaced the drag model outright — DataObject and DataFormats are no-op stubs and IDataObject is not in the reference assembly, so every tutorial written for 11 does not compile here. Hosts can be grouped, flat and never nested. A parent id merged as a scalar lets two offline clients each re-parent A under B and B under A, producing a cycle inside an encrypted payload that no server can police and every reader would have to detect for ever. Membership lives in that payload rather than in the one plaintext concession ADR 0001 allows, whose test is that the relay cannot function without it — nothing on the server reads a group, so what plaintext would hand over is a clustering of the estate for nothing. The plaintext column reserved for it is dropped, provably always null, and the server now refuses a client that sends one; it was never populated, was copied on apply, and was not cleared on delete, so a group id would have outlived the host it described. Snippets insert through xterm rather than through the pump, because xterm is the only thing that knows whether the remote has bracketed paste on, and that is what makes a shell treat embedded newlines as text instead of as execute. The host process moves opaque bytes and never parses output, so it would have to guess, and guessing wrong runs every line. Running is off by default and the copy says the text goes into whatever is there — the terminal has no notion of being at a prompt, and may be in vi or at a password prompt with echo off, so the Enter the user presses themselves is the entire safety property. Connections and keychain changes are recorded as synced encrypted items, which is what makes them auditable by a team later and costs the server knowledge of connection rate and timing from row counts alone. ADR 0001 already concedes it cannot hide that class of metadata; the trade is now written into it rather than left implicit. A connection entry is written once, at close, which is what makes a synced log tractable: nothing to merge, one outbox row, no chance of colliding with itself. Live sessions come from memory, not from the log. The write is void by contract and posts to a bounded channel, because putting an encrypt-and-write on the teardown path of every session is how closing the application comes to take four seconds. A ticket opened before a lock still closes afterwards, since a shell outlives the vault. The activity log hooks the one generic repository every kind writes through, so it cannot miss a caller — which is also why the log kinds themselves declare they are not audited, or the first entry would write an entry about writing an entry. It records the names of the fields that changed and never their values; a log with an old password in it would be a plaintext credential store with no vault around it. Retention is 90 days or 5,000 entries, whichever bites first, pruned on the sync loop rather than on a second timer. That log traffic then broke the status line, which is worth recording because the fix is a shape and not a patch: background sync counted its own log rows as pushed items, so the quiet rule stopped being quiet and every action's message was overwritten a second later by a sync report. The report now separates log rows from user items and the rule reads the latter. S3 buckets appear as a remote in the file browser, behind the same interface an SFTP session implements, so the queue and both panes did not have to learn what they are talking to. Uploads go through a pipe, because the queue wants to write and the SDK wants to read; memory is then bounded by the part size instead of buffering a file to disk twice. Finally, the Windows device key store moved out of the session project, which was the one thing keeping it from being portable — everything else in it is platform-neutral, and a Windows CNG dependency in the middle of the vault code meant a second head could not reference it without dragging Windows along. The seam that made the move free was already there. docs/android-port.md is the audit behind that: what ports, what does not, in order of cost, the four decisions taken, and an inventory of every screen and state the interface has to carry, written so a design can be made from it directly. dotnet build, dotnet test and dotnet format --verify-no-changes are all clean: 1240 tests at zero warnings, including the end-to-end suite against real containers. The manual checks that headless Avalonia cannot make — the drag from Explorer, a generated key against a real host, twelve tabs at the minimum window width — are listed in docs/manual-checks.md and are still outstanding.	2026-07-31 20:30:05 +02:00
jaap-jan	23eca3a21b	Merge branch 'main' into claude/m3-implementation-57f9d7 ... Three files conflicted, and two of the resolutions are more than a choice of side. QuickConnectTests had both branches fixing the same build break — main's M2 merge left the shell's constructor with an ISftpSessionFactory nobody passed. Main's version wins because it carries a comment saying why the palette never needs a session. VaultSession's conflict is adjacent edits: main added the remembered sign-in members and this branch changed SyncAsync's summary from "the active vault" to "one vault". Both kept. VaultViewModel is the one that matters. Main taught the background pass to report a sync that had to start over, on the grounds that a machine which silently re-read a whole vault has had something happen to it; this branch turned a pass into one report per readable vault. Taking either side alone would have lost the other, so ResyncedFromStart is now one of the conditions IsWorthReporting checks, per vault. Merging also broke something neither branch could have caught alone, and the build would not have said a word. SyncOnceAsync cleared LastSyncFailed unconditionally, which was right while a pass was one vault and a failure was an exception that never reached that line. A failure is now a report — one unreachable team vault must not stop the others syncing — so the flag was being cleared over a vault that had just failed, lighting the titlebar SYNCED. It is computed from the report instead, in the one place both callers go through, so the manual command gets it as well as the loop. The background pass still swallows the message and keeps the fact, which is what AnAutomaticPassThatFails_LeavesTheStatusAlone is there to hold it to. Two comments the auto-merge left describing a world with one vault in it: the SCOPES rail's, which said team vaults are refused by the access service, and the host sidebar's "One heading, for one vault".	2026-07-31 12:26:59 +02:00
jaap-jan	95816de0c5	Share a vault with a team, without the server holding a key ... M3's teams, sharing and ACLs. Teams with roles, a public-key directory, the append-only key log served for clients to check it against, team-owned vaults, and vault key grants wrapped by a client and stored opaquely by the server. VaultAccessService resolves team membership to PermissionFlags, so a viewer may pull and may not push; the desktop client reads and syncs every vault it holds a key for, and a real TEAMS screen replaces the one that said it did not exist. No migration: team, team_membership, vault.team_id and vault_key_grant have all been there since the first one, which is what carrying two unused tables bought. Membership is authorisation. A grant is access. The obvious model is one concept — "access", with a role attached, handed out by the server — and this architecture cannot implement it: a vault key is sealed to each member's X25519 key, and only a client holding the plaintext can seal it for somebody else. So "give Bob access" decomposes into a database write and a wrap, which happen on different machines. Adding a member makes the server serve them the vault; it cannot make it readable. VaultSummary.WrappedVaultKey is null in the meantime and the vault appears in their list saying it is waiting for a key, because hiding it until a grant existed would have been tidier and would have implied the server was the thing granting access. The screen says the same thing after every add, in the status line. ADR 0009 records the whole decision. Sharing verifies or refuses. A directory lookup is a claim by the server about a third party's public key, and wrapping to an unverified claim hands the vault to whoever made it — no amount of transport security helps, because the server is inside the threat model. KeyLogAudit reads the whole log, recomputes every entry's hash from its own contents, checks the chain from genesis, and refuses unless the offered key appears in it unchanged. There is no override flag: one that exists gets used on the day the log is briefly unreachable, and the resulting grant is indistinguishable from a correct one afterwards. What it still cannot promise is that the key is the right person's, so the fingerprint comes back for an out-of-band comparison and the success message says so every time. A test corrupts the fake server's log by one byte and watches the client refuse rather than warn. The roles are only the ones that are enforceable. There is no ConnectOnly, despite the design asking for one and TeamRole having room: SSH terminates on the client, so a session needs the credential's plaintext on that machine, and "may connect but may not read the key" cannot be enforced here. Shipping it as an option in a dropdown would have been a lie. Connect rides along with Read and is documented as an interface hint. Removal is named for what it does — it revokes grants and flags the vault for rekey, and claims nothing about what is already on somebody's laptop. Three things are deliberately absent, and each is a refusal rather than an omission. The rekey itself, because re-wrapping every item's data key under a new vault key needs a client holding the current one; the server records that a rotation is owed and the interface reports it, which is more honest than a button that only appears to do it. Ownership transfer, because allowing an owner to be removed without one leaves a team nobody can administer. And cross-vault host key trust: a pin in a team vault is listed but not consulted at connect time, because any member with Write could otherwise pre-approve a fingerprint another member's client then trusts silently for a host in their own vault. Scoping trust properly needs a scope on the SSH connect path, which IKnownHostStore has not got; until then the narrow direction is the safe one and the cost is in the README rather than hidden. Reading now spans vaults and writing still does not. Every list on the vault and hosts screens covers each vault the keyring opened, rows carry the vault they came from, and an edit goes back to that vault rather than to the active one — writing it to the active vault would fork the item and only show up when a colleague wondered why their change never arrived. A new item goes wherever a picker says, defaulting to the personal vault and never moving on its own, because an item filed into a team's vault is visible to that team and moving it back means deleting and retyping. The sidebar heading stops naming one vault once there are two, and each row names its own. The server checks what it can and nothing it cannot. It will not record a grant for a key its recipient no longer holds, for a superseded generation, or for somebody who is not in the team — each of those would otherwise surface days later at the far end as a tag failure indistinguishable from corruption. It does not verify the wrap or the signature, and the grant service says so: that would be a convenience and never the boundary, and would put an asymmetric implementation on a machine that is supposed to hold no keys. Two bugs the tests found. TeamsViewModel's busy gate blocked its own reload, so a team created a moment earlier was missing from the list it had just been added to. And syncing every vault turned a failure from an exception into a report, which made a background pass announce an unreachable vault once a minute — the exact behaviour AnAutomaticPassThatFails_LeavesTheStatusAlone exists to prevent. The fact is recorded and the message swallowed, as it was before; pressing Sync still names the vault and the reason. Also fixes a build break this branch started with: QuickConnectTests was never updated when M2 added ISftpSessionFactory to the shell's constructor, so nothing built at all.	2026-07-31 12:18:28 +02:00
jaap-jan	03e902a2d2	Colour the host's file rows by what their mode says ... The remote pane's NAME column was blue for a directory and plain for everything else, and the PERMS column was faint whatever it said. Two colours now come off the mode, split across those two columns on purpose: NAME says what a row is, so a file with an execute bit is green there, and PERMS says what is notable about how it is set, so a file anyone may write to is amber over the characters that actually say so. Because the two never compete for one TextBlock, a world-writable executable shows both facts instead of one winning an argument. No new blue is spent, which is what App.axaml asks for: it reserves blue for a directory, a distinct scope, and calls it deliberately rare. Both are files only, and each exclusion is a wrong answer avoided rather than a case not got to. Every symbolic link is lrwxrwxrwx by convention and its mode governs nothing — what may be written is the target, whose mode an lstat listing never fetched — so amber there would fire on every link on the host. A world-writable directory is /tmp, made safe by a sticky bit PosixMode does not render, and warning about it would be warning about the half of the mode that is on screen while the half that answers the warning is not. And the execute bit on a directory means "may be searched", which is true of very nearly every directory a host has, so green there would paint the whole pane and mark nothing. The two questions read back the string PosixMode wrote rather than carrying its nine booleans through SftpEntry as well. That is the point rather than a shortcut: two representations of one fact is how a row ends up coloured for a bit the column beside it does not show. A mode of the wrong length answers false rather than throwing, since these decide a colour and a listing is not worth failing over one. The amber is Warn rather than WarnText, which is the muted amber a warning card writes its sentences in. At 9.5px against TextFaint that one is a shade rather than a signal, and a marker nobody notices is the same as no marker. The local pane is untouched, on the grounds it already gives for having no PERMS column at all: a POSIX mode is not a fact about a file on Windows, and colouring one there would invent exactly what the column declines to print. Twenty cases in RemotePathTests, which needs no container — the execute bit in any of the three triples rather than only the owner's, the others-write bit alone, a mode of the wrong length, and the file-only rule for both questions from all three kinds. dotnet format is clean and the app and layout suites pass at 109 and 35.	2026-07-31 12:17:56 +02:00
jaap-jan	1292084af9	Merge branch 'claude/delete-confirmations-becf4a'	2026-07-31 11:52:27 +02:00
jaap-jan	91438fb382	Ask before deleting, and connect a host by double-clicking it ... DELETE on a host, an SSH key, a stored password or a file on the host now puts a question where the button was, and only answering it deletes anything. It is a state rather than a dialog, which is the arrangement signing out already had and for the same reason: this is the moment that has to be able to say what is about to go before it goes. What the question says is counted rather than generic, because a confirmation that only asks whether you are sure is a click to train people out of. A key names the hosts that authenticate with it and says they will refuse to connect afterwards rather than falling back to a typed password, which is what the connect path actually does. A host discloses a terminal open on it, because deleting the host does not close the session. Every vault deletion says how far it travels and whether this machine can push the tombstone yet or is queuing it. Deleting on the host carries the strongest warning of the four on purpose: everything else here is a tombstone against a copy the server still holds, and a file on somebody's machine is bytes with nothing behind them — so that one names the full path, since a bare name identifies nothing. The armed request carries the item's entity id, so nothing that moves the selection between the question and the answer can redirect it, and answering about something that has since gone says so instead of doing nothing quietly. Disarming compares ids rather than rows, which is the subtle half: a reload replaces every row object, so the naive rule would have let the pass that runs every minute take the card away from somebody halfway through reading it. Forgetting a pinned host key is deliberately still unguarded. It costs one fingerprint check on the next connection and it is the safe direction to be wrong in — the dangerous button there is the one that adds trust, and that one is already a prompt at connect time. Discarding a stopped transfer is likewise unguarded: it removes a resumable part file and leaves the source alone. Double-clicking a host in the sidebar connects to it, wired as a gesture in the control exactly as the transfers screen opens a directory. CONNECT stays, since it is the button with the password box beside it. Ten existing delete call sites now go through arm-and-confirm helpers, and eight new flow tests cover asking first, cancelling, the counted warning, disarming on a selection change and on an editor opening, surviving a sync, and the stale-item guard. Three layout tests measure the new shapes — the sidebar card is the one card in the application a user cannot scroll — and one of them also asserts the card renders its text, because a card whose compiled bindings did not resolve would lay out perfectly as empty rows. The double-click test performs the real gesture and proves it reached the connect command through a refusal that never touches a network. dotnet build, dotnet test and dotnet format --verify-no-changes are all clean: 853 tests, including the end-to-end suite against real containers.	2026-07-31 11:52:13 +02:00
jaap-jan	9608d73747	Come back from a sync position the server will not accept ... "The server returned 400: The sync cursor is not valid for this vault. Resync from the beginning." told the user exactly what to do and gave them no way to do it. The cursor is the only thing a pull sends, so the refusal was permanent: the next pass read the same stored cursor and was told the same thing, once a minute, for ever. And because the pull runs first, the exception ended the pass before it reached the outbox — so the vault stopped receiving other machines' changes and stopped sending its own. A machine that met this went quietly read-only until somebody deleted its cache. The engine now does what the message asks. A pull refused with the invalid-cursor problem code — the code, never the prose, which is free to change — drops this vault's position, writes that down, and reads the log again from the beginning. The restarted request carries no cursor, which is the one position a server cannot reject, so the retry cannot loop; a refusal of that is rethrown rather than retried, and a restart is allowed once per pull. The position is saved before the replay starts, so a process that dies halfway through begins the next one from the beginning too rather than meeting the same refusal again. The mirror is deliberately kept. Replaying rewrites every row the server still has and applying a change is a blind overwrite, so the re-pull repairs the mirror on its way past; clearing it first would claim more than the evidence supports — the position was refused, not the contents — and would leave a machine that lost its connection mid-replay with less than it started with. That leaves one gap, named in the remarks rather than left to be discovered: once tombstone collection exists, a replay stops carrying deletions older than the retention window. None of the causes are the user's doing — a rotated cursor signing key, a vault served from a restored database, a cache copied between machines — so nothing asks them to decide anything. The report carries ResyncedFromStart and the status line says the position was not recognised and the vault was read again. It is kept out of NeedsAttention, because nothing is outstanding, but the background pass breaks its usual silence for it: a sync that pulled the whole vault on a day nobody changed anything otherwise reads as a fault. The fake server grew a switch that refuses cursors the way a rotated signing key does, including ones it minted itself. Three cases: the vault is re-read and the change on the far side of the refused position arrives; the edits waiting in the outbox are still pushed in that same pass, which is the half that made this worth recovering from rather than merely reporting; and a server that refuses the beginning itself is surfaced instead of replayed against. dotnet build is clean at zero warnings, dotnet format is clean, and the sync and app suites pass — 109 and 101.	2026-07-31 11:32:14 +02:00
jaap-jan	240aadb746	Merge branch 'main' into claude/vault-unlock-logout-autosync-a84c35 ... Four files needed a hand, and all four were two branches adding something in the same place rather than either changing what the other did. The shell's constructor now takes both new parameters: main's SFTP session factory, which it must have because it builds the transfers view model, and this branch's optional resume handler, which stays last so every existing test that constructs a shell without one still gets a shell that can only be online because somebody signed in during this run. App.axaml.cs, ShellFlowTests and QuickConnectTests pass the pair; the layout suite keeps both of its new fields. Signing out now detaches the transfers screen exactly as locking does, and the confirmation says that an open transfer session survives it. That is the same policy both sides already argue for their own case: signing out destroys this machine's copy of the vault, not work that authenticated before it. QuickConnectTests did not compile on main — the SFTP commit added a constructor parameter and the quick-connect suite, merged from a parallel branch just before it, was still calling the old one. Fixed here rather than worked around, since the merged tree has to build. dotnet build, dotnet test and dotnet format --verify-no-changes are all clean: 980 tests, including the end-to-end suite against real containers.	2026-07-31 11:16:49 +02:00
jaap-jan	d1700f5a34	Merge branch 'claude/m2-file-transfer-1b9951'	2026-07-31 11:08:05 +02:00
jaap-jan	0b261c4d39	Stay signed in, come back online by itself, and let a machine be given up ... Three things a machine that has been set up could not do. Unlock now takes Enter, which is the gesture everybody makes after typing a password and which did nothing until they found the button. Signing in survives a relaunch. The refresh token is kept in the local cache, sealed under the vault's own cache key, so a later launch resumes the session through the refresh grant with no browser and nobody present — and because it is sealed under that key, only an unlocked vault can resume it. A locked client therefore cannot reach the server at all, which is a consequence worth stating rather than working around; docs/crypto.md §3.2 records it. Every sync pass asks the shell for a connection rather than reading one captured at unlock, so a laptop that unlocked on a train is online within a minute of finding a network, with nothing pressed. Unlocking itself still never waits on a socket. Signing out empties this machine: the profile, the cached items, the outbox and this machine's device key, with the account's row withdrawn when the server can be reached. It asks first and says what it costs — the outbox count when the vault is open, an admission that it cannot be counted when it is not, and the shells that keep running either way. The vault is on the server and is untouched, which is what makes the same button the only honest answer to a forgotten passphrase, so it is on the unlock screen as well as in preferences. It cannot end the session at the identity provider, and says so. Two defects surfaced on the way. The synchronisation pass that runs when the vault opens never ran at all: the loop is started from inside the unlock command, so the busy flag it yields to was raised by that command — the first sync was a minute late on every launch. And signing in from preferences while unlocked threw an unlock screen over an open vault whose keys were still in memory. The unlock card and the new confirmation live in their own controls because MainWindow cannot be laid out headless, so markup left inside it is markup no test can measure; both are now measured at the window's minimum size in the shapes that grow. What is still unverified is the composed window itself.	2026-07-31 11:07:36 +02:00
jaap-jan	04faef6597	Move files to and from a host over SFTP ... M2's file transfer, built bottom-up: an SFTP session on the SSH layer, a transfer queue in a project of its own, and the two-pane browser the design asked for replacing the screen that said it did not exist. Remote listings carry names, sizes, modification times and a real drwxr-xr-x — nothing in this repository could render a POSIX mode before — and the queue moves one file at a time with progress, throughput and resume. The design import assumed this would be an SFTP subsystem channel on ISshConnection, beside the shell on a transport that is already up. SSH.NET does not offer that: SftpClient derives from BaseClient and owns its own transport, and there is no supported way to hand it an SshClient's session. So file transfer opens a second authenticated connection, and it is named for that rather than dressed up as a channel — OpenSftpAsync is on ISftpSessionFactory, not on a connection. The difference is visible to a user: the host records a second login, and a host whose password is typed each time asks for it again on this screen. It goes through the same host key gate, the same pin and the same two refusals a shell does, so a fingerprint approved for a terminal is approved here and one approved here reaches the other machines with the next sync. docs/design-import-gaps.md is corrected, and marked as the one row where what shipped differs from what it predicted. Nothing is written at its final name until it is complete. Every transfer goes to a .dodossh-part file beside its destination and is renamed into place at the end, so an interrupted transfer can never be mistaken for a finished one — which matters most for what this screen is actually for, which is copying a build artefact onto a server and then running it. A destination that already exists is refused outright rather than overwritten: the queue has no way to ask, and silently replacing a file somebody's process is serving is the worse of the two failures. The remote pane has DELETE and MKDIR so that refusal is not a dead end. A test against the container pins the assumption underneath all of this — that SFTP's rename does not clobber. Resume works within a run of the application and not across a restart, and the limit is deliberate rather than unfinished. Nothing records which source wrote a part file, and resuming one on the strength of its name matching is how a corrupt artefact gets delivered with nothing reporting a failure; a part file found at startup is started over. Making it survive a restart needs the preferences store this client still has not got. The offset a resume starts at is the part file's own length rather than the transfer's recorded progress: a cancellation can land between a write completing and the counter moving, and only one of those two is a fact about the bytes that are there. The queue and its connection outlive a lock, as shells do. LockAsync already argues that locking must not destroy work in flight — it is what somebody does when they walk away from the machine, which is exactly when a long transfer is most likely to be running — so TransfersViewModel is created once and the vault is attached on unlock and detached on lock. What locking takes is the host list, and it has to: those rows carry decrypted secrets. DodoSSH.Client.Transfer is a new project rather than more of Client.Ssh. The two answer different questions — one is about reaching a host, the other about moving bytes and what to do when moving them stops halfway — and this is the only client project that deliberately touches the local filesystem. Three defects the tests found, none of which review would have. SftpPath.Name answered an empty string for the root. NavigateRemoteAsync wrapped itself in the busy guard, so navigating from inside another command did nothing at all and the remote pane simply stayed empty after connecting, with no failure anywhere to explain it. And opening an SFTP session per test made two handshakes per test — this client learns a host key by being refused — which pushed the SSH assembly past sshd's MaxStartups and failed a different few unrelated tests each run; the session is shared through the fixture now, with the reason written where the next person will hit it. 1004 tests green across 18 projects, 24 of them new: the SFTP subsystem against the OpenSSH container, the queue against a real temporary directory and a fake host, and three more layout measurements because a screen this window has never laid out is a screen never checked. Not verified: the screen has not been looked at running. The layout harness measures it at the window's minimum in three shapes, which is the class of defect that has shipped here before, but reaching it in the application needs the compose stack, the migrations, the API and a browser sign-in. What is still absent — the status bar's transfer count, dragging between the panes, transferring a directory, and sftp over a bastion — is in docs/design-import-gaps.md.	2026-07-31 11:07:29 +02:00
jaap-jan	f7c5096bc6	Keep the stub servers on loopback ... Running the tests raised a Windows Firewall prompt, and raised it again from every worktree. WireMockServer.Start() with no settings listens on 0.0.0.0 and [::], and the prompt is keyed to the binary that opened the socket — so each test executable asks once per bin path, which a new worktree or a switch between Debug and Release makes new again. The three suites that hold a firewall rule on this machine are exactly the three that use WireMock; every other listener in the repository already binds 127.0.0.1. The stubs now say so explicitly. Port 0 is still WireMock's own free-port search and still comes back on server.Url, which is what each stub builds its base URL from, so the authority the API validates against and the issuer its tokens claim follow the binding rather than being pinned to a host name. Sampling the listening sockets of a full DodoSSH.Api.Tests run afterwards finds one, 127.0.0.1, where there were previously three.	2026-07-31 11:06:46 +02:00
jaap-jan	312d766c30	Let the quick-connect palette answer for itself ... Clicking outside the palette did nothing, because nothing was listening: the wash took no pointer input at all, so the only ways out were a key and the button that opened it. It now closes on a press whose source is the wash itself, which is what separates outside from inside — a press on the card bubbles through the same handler on its way to the window, and closing on those would make the palette impossible to click into. The caret never reached the query box either. The window focused it from the view model's PropertyChanged, and that handler runs before the binding which reveals the control — measured, with the same wiring, in a replica window. So it focused a control that was still collapsed, which Avalonia treats as a no-op and does not replay when the control is revealed, and the keyboard stayed wherever the click that opened the palette had left it. Becoming visible is now what triggers it, posted rather than called: a control that has never been laid out has no visual children, and at the instant IsVisible turns true the box still reports IsAttachedToVisualTree() == false. Escape, Enter and the arrows move to the palette as a tunnelled handler. Answering them only on the window was fragile in the way that matters here: anything on the route that took a key first would silence them, and with the focus never landing in the palette the key was being pressed at whatever the opening click had focused — a focused Button eats Enter. The window keeps Ctrl+K, which has to work when the palette is not showing, and forwards the rest as the net for a press that arrives from outside the palette. Which is also why this moved out of MainWindow rather than being fixed there. Showing MainWindow initialises WebView2 on a thread it refuses, so nothing on that window can be tested — the palette shipped with no test of any kind. As a UserControl it hosts in a bare window and takes real key and pointer input, and there are now six: press on the wash closes, press on the card does not, Escape closes, the arrows move the selection without taking the caret out of the box, Enter takes the highlighted host, and the palette takes the keyboard when it appears.	2026-07-31 10:43:30 +02:00
jaap-jan	94e11f5e38	update packages	2026-07-31 10:12:05 +02:00
jaap-jan	0b49cfb3c6	Merge branch 'claude/api-fastendpoints-migration-020431'	2026-07-31 10:10:22 +02:00
jaap-jan	9a76eced14	Give hosts and terminals their own screen, and the rest of the vault another ... Rebuilds the client's shell from an imported design: a titlebar and nav rail it draws itself, real multi-session tabs over the one WebView, a Ctrl+K host search, and a vault screen that merges keys, passwords and pinned host keys into one table. Hosts left the vault column for their own screen beside the terminal, which is what the design asks for and turned out to be the better split anyway. Two screens the design shows have nothing behind them yet — file transfer and teams — and say so plainly rather than rendering invented data; every other gap between the design and this build is recorded in docs/design-import-gaps.md.	2026-07-31 08:39:37 +02:00
jaap-jan	9bc28f1c0f	Move the API onto FastEndpoints, without moving the wire ... Eight endpoints today, around sixty planned. The minimal-API shape — a static class per area holding static local functions, route and policy and name asserted in one fluent chain with the handler somewhere below it — has not hurt yet, and would. A handler's dependencies are parameters rather than injected, a group's RequireAuthorization sits far from the handler it governs, and there is no type to hang an endpoint's own documentation on. FastEndpoints is one class per endpoint, its route and authorization in Configure(), its handler a method on the same type. Nothing about the wire moves, and the evidence is that the 94 existing HTTP tests pass with zero edits to any of them. Same routes, verbs, route constraints, status codes, operation ids, and the same RFC 9457 bodies with the same code values. Every place the idiomatic FastEndpoints answer would have changed one of those, it was refused: Endpoints are registered from an explicit List<Type>, not found by scanning. ADR 0002 rejected reflection discovery by name, and the reason it gave is sharper here than in general — under WebApplicationFactory the scan reaches the test assembly, so an endpoint written in a test would be registered into the host under test. The cost is a line per endpoint that can be forgotten, which is what the endpoint-inventory test is for. That test is the one ADR 0002 promised and never got. Handlers still return Results<Ok<T>, NotFound, ProblemHttpResult> from ExecuteAsync. The union executes as an ordinary IResult, which is what keeps problem bodies going through the host's serialiser and IProblemDetailsService, and what keeps the compile-time record of which statuses an endpoint can produce. No Send.* call appears anywhere; the moment one does, a response has left the host's serialiser. Validation stays in the feature services. A Validator<T> short-circuits before the handler and answers with FastEndpoints' own envelope, which carries no code — and the code is the only part of an error the client branches on. Twenty-odd tests assert a specific code on a 400. It is banned in BannedSymbols.txt rather than merely avoided, because the framework's documentation leads straight to it and it looks like an improvement. Three defects arrived with the framework and were caught in review. All three were green at the time, which is the part worth remembering. FastEndpoints maps GET /_test_url_cache_ unconditionally, in every environment, with no policy and no way to opt out; it answers with the whole endpoint-name-to-route table. It is short-circuited to 404 — by asking routing which endpoint it selected, after the first attempt compared the request path with Ordinal and was therefore bypassable at /_TEST_URL_CACHE_, certified by a test that only ever tried one spelling. The default request binder writes query-string values over the deserialised body, which would have let ?identityProviderToken=... put an ID token in a URL and from there into every proxy log on the path; every endpoint now binds from the body alone. And a route value read with Route<T>() is invisible to ApiExplorer, so the generated document named {vaultId} in a path template with nothing declaring it — invalid OpenAPI, and unusable by the client generators the document exists for. Two changes to the surface, both deliberate. A body that cannot be deserialised now answers with a problem document carrying malformed-request, rather than an empty 400: FastEndpoints' default announces application/problem+json while sending something else, and names the failing .NET type on the wire, in a codebase that sets IncludeErrorDetails = false to prevent exactly that. And the route table above returns 404 where it would otherwise have answered any authenticated caller. Each of the three fixes has a regression test that was checked by reverting the fix and watching it fail — four failures for the route table and the binder, four for the document. That check is the whole reason to trust them, since all three defects passed a full green suite on the way in. 950 tests green across 16 projects, 14 of them new and no existing test edited. Zero warnings, format clean, locked restore clean. FluentValidation, JobQueues and Messaging are in the graph now and none is used. Not verified: the generated document's response schemas, which differ from before — FastEndpoints contributes its own Produces metadata. Nothing consumes the document yet, and MapOpenApi runs only in Development behind the fallback policy. It needs pinning if ADR 0002's build-time artifacts/openapi/v1.json is ever built.	2026-07-31 08:39:06 +02:00
jaap-jan	d162271a45	Show the host keys this vault has approved ... Trust was created by the connect prompt and withdrawn from one host's editor, so a pin for a host that had since been deleted or re-addressed was unreachable from the interface entirely. It went on refusing connections and nothing in the application would admit it was there. Two of the four recorded debts were really this one: leftover pins, and no list to see them in. A fourth section in the vault column, and the first that adding one has been cheap for — three edits and two layout tests, which is what #8 and #9 were for. No editor and no Add, which makes it the only section with neither. A pin is not something anybody writes: it appears when somebody approves a fingerprint at the moment of connecting, which is the one place a person can actually check it against what the operator published. A form for typing one in would be a form for pasting whatever a man in the middle just offered. So the section exists to show and to withdraw, which is exactly what was missing. The fingerprint is shown in full, wrapped, in a monospace line. The only thing anybody does with one is compare it against a fingerprint an operator published, and half of one cannot be compared — it can only be glanced at, which is the habit pinning exists to replace. Nothing here is secret; a host key fingerprint is published on purpose. A pin no host in this vault dials is badged rather than hidden or deleted. That is the leftover the debt was about, and keeping it is still right: the address may be reached by something without a bookmark, and trust is about the endpoint rather than the bookmark. The badge is a hint and not a verdict, which is why nothing acts on it. Matched case-insensitively, because a host name is, and because a list that called DB.internal unused next to a host saved as db.internal would be inviting somebody to delete trust they rely on. Forgetting goes through the same ForgetAsync as the host editor's button, which withdraws every pin for the address rather than the selected row. Deliberate: somebody who has stopped trusting a machine has not decided to keep trusting one of its keys, and a second pin under another algorithm would go on being offered at the next handshake — which reads as a withdrawal that did not work. The status line says how many went, and the change is pushed immediately, because the other machines are the ones still refusing to connect to a rebuilt server. The list is read through the repository rather than through VaultKnownHostStore, whose snapshot is shaped for the SSH handshake: one pin per endpoint, deduplicated, no entity ids. This list has to show duplicates, because a duplicate is one of the things worth seeing. Two mutations, both caught: calling every pin dialled (3 tests), and defaulting the selection to the first row (1) — the same hazard as the credential list, since Forget acts on the selection. The selector now holds four buttons in 340 pixels, and TheSelectorIsBigEnoughToClick measures how much of that they use rather than leaving a fifth section to discover it as "a button falls outside the window". 936 tests green across 16 projects, 6 of them new. Zero warnings, format clean. Not verified: how the section looks. It joins the list in outstanding item #7.	2026-07-30 17:44:33 +02:00
jaap-jan	f86791e817	Finish revoking a device, instead of half of it ... ForgetDeviceAsync stopped this machine unlocking without a passphrase and left the server's row exactly where it was, so the account went on listing a device nobody could account for. ADR 0007 recorded that as a deliberate gap needing an endpoint. This is the endpoint, and the two things that turned up behind it. DELETE /api/v1/me/devices/{id}. The device row is not the dangerous half: a kind=device wrap is the user's identity bundle sealed to a key somebody may be holding, and that is what has to go. It goes on the foreign key's cascade rather than a second statement, and RevokeDevice_TakesItsWrapWithIt asserts the cascade rather than trusting the configuration to keep saying so. Scoped to the caller's own account, which is the only authorisation check there is. The id is an unguessable v7 GUID, but unguessable is not a permission — without the scope one user could withdraw another's device key by pasting an id they saw once, and the victim's next launch would ask for a passphrase with no explanation. 404 rather than 403 for somebody else's device, so a stranger does not learn the id exists. Never refused for being the last device. ADR 0001 makes an enrolled device a recovery path, so removing the last one does cost the user something — but the machine being revoked is most likely the one they have just lost, and a server that argued about it would be refusing the one request that has to work immediately. The passphrase wrap is untouched either way, which RevokeDevice_LeavesThePassphraseWrapAlone pins. --- Two things found on the way --- Registering twice from one machine left two devices on the account. The server is idempotent on the public key, but the client generates a fresh key pair every call and the keystore holds one — so the second registration orphaned a wrap whose private half had just been overwritten, which is precisely the leftover this change exists to remove. Registering now withdraws the previous device. Found by a test that asserted the property and failed. And the fakes were lying about it. FakeAccountServer's comment claimed the real service's idempotence while handing back a fresh Guid on every call, which is invisible until something revokes by id — at which point a test would be revoking an id the server never issued, and passing. Both fakes now issue one id per public key and drop the wrap with the device, as the cascade does. --- Reachable at all --- ForgetDeviceAsync had exactly one caller and it was a test, so "Stop unlocking here" now sits in the account bar where "Use Windows Hello here" was. Its own flag rather than the negation of that one: a machine with no TPM and a machine that is already registered are both "cannot register", and only the second has anything to take back. No confirmation prompt, deliberately. The cost of pressing it by accident is one passphrase and one re-registration; the cost of a dialog is a moment's hesitation at the point somebody has realised a machine is in the wrong hands. Offline it does the local half and says so rather than refusing. Whether this machine may unlock itself is decided entirely by the local cache and the local keystore — the unlock path never asks the server — so forgetting here is what actually revokes, and "you are offline, so this machine will go on unlocking itself" would be the worst available answer. DeviceRevocation.LocalOnly is what the interface reports and the status line explains what is left to do. The local half runs first for the same reason, and the keystore call is the first thing in the method that can yield: on Windows it raises a consent dialog, and a dialog wants the thread it was called from. That ordering is currently load-bearing and shakier than it looks — see the open device-unlock hang. Four mutations, all caught: dropping the user scope from the server query (1 test), skipping the stale-device revoke on re-registration (2), skipping the server call in ForgetDeviceAsync (2), and the earlier version of the client that never called it at all. 930 tests green across 16 projects, 13 of them new. Zero warnings, format clean.	2026-07-30 17:33:31 +02:00
jaap-jan	d17a60e7c3	Stop asking the server to delete things it has never seen ... Add a host on a laptop with no network, change your mind, delete it: the outbox holds a tombstone for a row the server has never heard of, the push answers Invalid, the change is parked, and the user is left looking at a rejected change for an item they already deleted and a pending count that will never reach zero. It applies to all four item types, because they all go through the one generic repository — the known-host path is only the likeliest way to meet it, since trust is pinned by connecting and withdrawn from the host editor. DeleteAsync now drops the queued create instead, when the server cannot be holding the item. A null expected version means the row is a create — including a create that has since been edited, because coalescing keeps the original expected version — so there is no server row and no mirror row, and dropping the queued change makes the item genuinely gone. The attempt count is what makes that safe rather than merely convenient. Nothing sent cannot have landed. A parked row cannot have landed either, because parking is what the pusher does when the server has refused, so the refusal is the evidence — and a parked create that the user then deletes could not be got rid of at all before this: the tombstone replacing it was parked in its turn. What is left is a create that went out and whose answer was never seen. That one still gets a tombstone, because the server may be holding the item and a local drop would strand it there for ever. A refused tombstone is recoverable; an orphan nobody can see and nobody can delete is not. Eight tests, and the interesting half is the other direction. A repository that quietly dropped tombstones would pass a suite written only around the bug and would lose data on every machine but the one that pressed the button. Which is not hypothetical, because the mutation pass found exactly that hole in the first draft of these tests. Removing the expected-version guard left every test passing: after a sync there is no queued row at all, so deleting a synced item never reaches the shortcut and proves nothing about it. The way to hold an unpushed Upsert over an item the server holds is to edit it offline, and EditingASyncedItemOfflineAndThenDeletingIt_StillQueuesATombstone is the test that was missing. Without the guard it deletes the item here, leaves it on the server, and the next pull brings it back. Three mutations, all caught now: removing the shortcut (5 tests), removing the expected-version guard (1), removing the attempt-count guard (1). The Upsert check itself is conservative rather than load-bearing — a queued Delete with no expected version is not reachable from the interface, and completing one locally would discard a tombstone that might be needed, so it stays and is not independently covered. 106 tests green in Client.Sync, 8 of them new. Zero warnings, format clean.	2026-07-30 17:15:51 +02:00
jaap-jan	da7462e41f	Show one kind of vault item at a time, and let the vault hold passwords ... Outstanding items #8 and #9, in one commit rather than two. They are separable as work and were built in that order, but not as a diff: the section enum has three members, the one-editor guard has three arms, and the picker offers keys and credentials from the same list. Reconstructing an #8-only state would mean hand-writing an intermediate version of VaultViewModel that never existed and that no test has ever run. One honest commit beats two invented ones. --- #8, the type selector --- The column showed two lists and two editors stacked in 340 pixels, and only just: the key list needed a MaxHeight and had to hide itself whenever its editor opened, both to stop the host list above it pushing the buttons off the bottom edge. Credentials would not have fitted at all. It now shows one kind at a time, chosen by a selector at the top, and both workarounds are gone because a section owns the whole column. Three departures from the plan, each with a reason found while building it. The selector is plain Buttons and a parameterised command, not a TabControl, a TabStrip or a ListBox. All three of those hold the selection themselves, so a click moves the highlight before the view model can refuse it — and this column does refuse, while an editor is open. A selector lit on a section the column is not showing is worse than the refusal it would be hiding. Buttons carry no state and cannot disagree with the vault. The one-editor-at-a-time rule survives with its justification replaced. That rule was a workaround for the sizing problem above, and sections dissolved it: the editors are in different sections and only one section is ever laid out. BothEditorsAtOnce_DoNotFit_WhichIsWhyTheRuleExists is now BothEditorsOpen_NowFit_BecauseOnlyOneSectionIsLaidOut — the same test, inverted, because its own comment said that if it ever started passing the rule had become unnecessary. It has. The rule stays for a better reason: an open key editor holds a pasted private key in a bound string, and letting the column move on would leave key material in a form nobody can see, with nothing on screen to say it is there. A sizing hack became a rule about not hiding a secret from the person holding it. KeyEditorIsInTheWay and HostEditorIsInTheWay are one AnEditorIsInTheWay, called by the section switch and by every editor-opening command. And releasing the keyboard from the terminal has never worked. MainWindow takes Win32 focus off the WebView's child window and then calls Focus() on VaultColumn.KeyboardTarget — and a ListBox is not focusable by default in Avalonia, which leaves focus to its items. So the call returned false, the window ended up with nothing focused, and the keystrokes went nowhere: exactly the state that method's own comment says its second half exists to prevent. Found by writing the test to assert focus was taken rather than that the right control was named — the cheap assertion was already passing. Fixed with Focusable="True" on every list. --- #9, credentials --- Credentials have synced since they were added and could not be created. They can now, and the sync layer needed no change at all: fourth item type, same result, which is the item-kind seam working as intended. One picker for all three ways a host authenticates, which is what makes the illegal combination unrepresentable rather than merely invalid. SshKeyChoice became AuthenticationChoice carrying an AuthenticationKind, and BuildHost reads both SshKeyId and CredentialId off that single selection, so a host naming a key and a credential — which HostSecret.TryValidate refuses — cannot be expressed. Two pickers would have expressed it and then rejected it at save time. The kind travels with the id in three places and none is padding: Missing takes it, the placeholder lookup matches on kind as well as id, and Bound(kind) returns null unless the selection is that kind. Drop any one and a dangling credential comes back as a dangling key, which saves as a key binding to an id no key has. A credential's username had to reach the SSH request, not just its password. TryBuildCredential returned only the secret and the connect path read the username off the host, so a stored credential would have gone out under the wrong account — wrong in a way a server only reports as "authentication failed". It is now TryBuildAuthentication returning a (Username, Credential) pair. The no-username refusal moved, and had to. It ran before anything looked at the binding, which made a credential's username unreachable in the one case it is most useful: a host somebody never filled a username in for. It is now the last thing every branch agrees on, so such a host is perfectly usable through a credential that carries one, and a host with neither still refuses and now says where to put one. --- What the measurements cost --- Ten mutations, all caught. Two are worth naming. Removing a section's IsVisible is caught by OnlyOneSectionIsOnScreenAtOnce and by nothing else: two visible sections overlap in the row they share rather than clip, so every fit test still passes while the column shows one list through another. Defaulting the credential selection to the first row is caught by ReloadingKeepsACredentialSelectionButNeverInventsOne, and the property is a safety one rather than tidiness — Delete acts on the selection, so a list that picked a row on every background sync would aim a one-click password deletion at something nobody chose. The key list has the same property, and its comment cited a method that has not existed for some time; both now name the delete command they actually protect. One test of mine could not fail, and the mutation pass is what found it. AHostBoundToACredential_SendsItsPasswordAndItsUsername gave the credential and the host the same username, so it passed whichever one the code read. An override is only tested when the two values differ. Two shipped statements went false and were corrected rather than left: the class remark saying passwords were "not yet" in the vault, and the terminal column's "Keys are in the vault; passwords are not yet." That column's hint is now a tooltip on the password box rather than a sentence in the row, which was measured the hard way — by looking. At the window's 820px minimum the column gets 480, and a 220px box plus Connect plus any sentence does not fit; the row has shipped clipped for as long as it has had a hint in it. That strip is the one part of the window nothing can measure, because MainWindow cannot be laid out headlessly at all. Extracting it into its own control, as the vault column was extracted for exactly this reason, is what would fix that, and is not done here. 911 tests green, 30 of them new. Zero warnings, dotnet format clean. Seen by a person, which is how the two defects above were found. Still open from that pass: unlocking with the device key raises its consent dialog and then never returns, while registering one works — the difference is which thread the CNG call lands on, and diagnosing it properly is its own change.	2026-07-30 17:08:59 +02:00
jaap-jan	573f5d5668	Keep the device key in the TPM, behind a consent Windows enforces ... The last of ADR 0007's three pieces, and it does not implement what that ADR originally decided — because writing it exposed a flaw in the decision. The ADR said "a Windows Hello gesture gating a protected blob". That does not deliver what the rest of the document claims for it: a gate inside the process is not a gate. A store that showed a prompt and then read a DPAPI blob would be bypassed by malware that skipped the prompt, read the file and called CryptUnprotectData itself — which is exactly the attacker the whole decision was made against, and exactly the reason DPAPI alone was rejected. The presence requirement has to be a condition of using the key, enforced below the application, or it is decoration. So the device key is encrypted to an RSA key created in the Microsoft Platform Crypto Provider — the TPM — under CngUIProtectionLevels.ProtectKey. Windows requires consent to use that key, so the prompt is not something this code can be talked out of showing. Malware can ask for the key; it cannot answer the dialog. That is strictly stronger than the ADR described, and most of what option D was being saved for: the wrapping key genuinely never leaves hardware. The X25519 device key still lands in memory to open the wrap, because DSH1 fixes that wrap at a curve the TPM cannot do — the remaining gap, and now a smaller step than it was. CngKey is in-box, so this needed no WinRT projection and no Windows target framework. Which is worth stating plainly because the opposite was planned: the piece was scoped as "where the Windows TFM lands", and it turned out a platform guard on one class was enough. Client.App and its two test projects stay on net10.0. Two things were measured on real hardware rather than assumed, and the second changed the shape of the work. The platform provider works here and holds an RSA key — confirmed by creating and deleting one before writing anything that depended on it. And ProtectKey prompts at key *creation*, not only at use. The comment in the first draft of this file said the opposite, with a confident explanation: sealing uses only the public half, so it should be silent. It is not. CngKey.Create blocks on a dialog, because the policy means "protect this key with a PIN" and Windows asks the user to set that up there and then. Found by writing tests around save and forget and watching the suite hang for ten minutes waiting for somebody to type one. That has two consequences worth knowing before touching this file. SaveAsync is user-facing code — it belongs on a UI thread, behind a button somebody pressed, never on a background pass. And almost nothing in the store can be covered automatically: two tests remain, availability and the empty-blob case, both of which provably reach no dialog. Disabling the UI policy to make the rest testable would remove the one property worth having. The interface offers two things and hides both where they cannot work. "Use Windows Hello" appears on the unlock screen only when this machine has a cached wrap and a keystore still willing to release the key; "Use Windows Hello here" appears in the account bar only when the machine can keep a key and has not already registered one, so it is spent once used. Absent rather than disabled, in both cases: a greyed-out button on a machine that never had a TPM reads as something broken, and the passphrase box beside it is not a fallback — it is the ordinary way in. Both unlock paths now share AdoptAsync rather than each opening the known-host store, building the vault and starting auto-sync. The ordering in there is load-bearing and a second copy would be a second chance to get it wrong. The shell's tests drive a fake keystore. Not for speed: the real one prompts on every save and load, so a suite using it would block forever. What the shell has to get right is which buttons appear and what happens when one is pressed, and a fake answers exactly that. It is shared from Client.Session.Tests by source link rather than reimplemented. 882 tests green, 6 of them new. Zero warnings, dotnet format clean. Not verified, and not verifiable here: the dialogs. Whether the consent prompt appears at the right moments, reads sensibly, and returns to a usable window when declined needs the application run by a person on a machine with a TPM. That is the remaining half of outstanding item #7, and it is now the only thing between this feature and being finished.	2026-07-30 15:17:30 +02:00
jaap-jan	1faea42b94	Unlock with this machine's device key, without a passphrase or a network ... The second of ADR 0007's three pieces: the seam a keystore plugs into, the wrap cached where an offline unlock can reach it, and the unlock path itself. What is still missing is the keystore — UnavailableDeviceKeyStore is what the application composes for now, so behaviour is unchanged until piece three lands. IDeviceKeyStore holds exactly 32 bytes, and only because the cache key moved first. It would have had to hold the local cache key alongside the X25519 scalar — a second live secret at rest, going stale on every passphrase change — had 7016ce3 not re-keyed that to the identity bundle. ADeviceUnlock_ReadsTheSameCache ThePassphraseWrote is the test that ties the two commits together: under the old derivation this session would have opened the identity and then found its own cache unreadable. The wrap is cached at registration rather than fetched at unlock, which is the whole point. The one unlock path that exists to save the user typing must not be the one that only works online; a laptop on a plane is precisely where a gesture should help. Every way this fails returns a status rather than throwing, because none of them are exceptional — a cancelled fingerprint prompt is the most ordinary thing in this file. Three statuses rather than one, because the caller says a different sentence for each: no device registered (the normal state of a machine nobody opted in on), the machine would not release the key (declined gesture, or a Hello key invalidated by a PIN reset — deliberately indistinguishable, since the remedy does not differ), and the key was released and did not open the wrap (a rotated identity, which will never succeed again and needs re-registering). A keystore returning the wrong number of bytes lands in the third rather than crashing the unlock screen. The subtle defect this could have shipped is in UnlockStore.Apply. That method runs on every sign-in from a /me response, which knows nothing about this machine's keystore — so assigning the device columns unconditionally would delete the wrap on the next launch, and the user's fingerprint would stop working for no visible reason and no error anywhere. The columns are therefore written only when the incoming material carries them, with AttachDeviceAsync and DetachDeviceAsync as the only paths that set them deliberately. Mutation tested: removing the guard fails RefreshingTheProfile_DoesNotDiscardTheDeviceWrap and nothing else. RegisterDeviceAsync lives on VaultSession because sealing the bundle is the one step only an open session can do, and the session is the bundle's custodian. Everything else arrives as a parameter, exactly as SyncAsync takes its transport, so the session still knows nothing about how either the wire or the keystore is implemented. Its steps are ordered so a failure cannot leave a lie behind: the key is generated, saved locally, and only then registered with the server. A server row whose private half was never stored is a device that can never unlock and that the account claims can — worse than not offering the feature at all — so the write that could produce it happens after the one that prevents it. ForgetDeviceAsync is deliberately half a job, and says so. It stops this machine unlocking without a passphrase, which is what a user turning the feature off means, but the server's wrap row survives and the account will go on listing a device that cannot unlock. Deleting it needs an endpoint that does not exist yet. Half with the gap recorded beats a method whose name promises the other half. The client cache gained two nullable columns and a migration, generated rather than hand-written this time. 876 tests green, 10 of them new. Zero warnings, dotnet format clean. Remaining: the Windows Hello store and the unlock-screen UI. That is where the Windows target framework lands, and where automated testing stops — a gesture needs hardware and a person, so the last piece is the one that has to be looked at rather than asserted.	2026-07-30 13:37:35 +02:00
jaap-jan	db4a8ed3d3	Let an already-enrolled account register a device key ... The first of the three pieces ADR 0007 needs, and the one that was a discovery rather than a plan. EnrollmentService.AddDevice runs only during enrollment, so without an endpoint the device-unlock feature would have reached accounts created after it shipped and no others — which is to say none of the ones that exist. The code even said so: "the devices endpoint sets it properly when it lands." POST /api/v1/me/devices takes a name, an X25519 public key and the bundle sealed to it, and writes a device row plus a UserKeyWrapKind.Device wrap. Possession is proved by construction, so there is no challenge. The wrap is the secret bundle sealed to the supplied public key, and only something that has opened that bundle can produce it. A caller who seals the wrong bytes registers a device that cannot unlock, which harms nobody else; the server cannot tell the difference and must not pretend to, because it holds no key that opens either. That is also why the client must be unlocked to call this at all. It is the one endpoint in the /me group that requires enrollment, and it says so itself rather than relying on the group. The group deliberately does not: GET / and POST /enrollment are how a client discovers it needs to enroll and then does so, and gating those on enrollment would make enrollment unreachable. Adding the stricter policy to this route alone means an unenrolled caller is told "enrollment-required" by the authorization handler rather than getting a 400 about the shape of a request that was fine. Idempotent on the public key, and 200 rather than 201 for the reason enrollment gives: a retry of an identical request returns the same body, so there is no single moment of creation to point a Location header at. A second row for one key would mean a device list with a duplicate in it and two wraps to revoke instead of one. Mutation tested — removing the lookup fails RegisterDevice_TwiceWithTheSameKey_ReturnsTheSameDeviceAndAddsNoSecondWrap and nothing else. That test also found a real defect, in the way these usually surface: two timestamps that print identically and are not equal. TimeProvider reports 100-nanosecond ticks and PostgreSQL's timestamp with time zone keeps microseconds, so the first call returned a value that no later read of the row would ever produce, and the idempotent retry answered with a different timestamp for the same device. Nothing breaks, which is what makes it worth fixing: the service now truncates to the precision the column actually holds, so the response is the same value every time it is asked for. The repo already had a precedent for this class of thing in KeyLogChain.TruncateTimestamp; it just had not been applied here. The platform is deliberately not carried on the wire, which leaves Device.Platform unreported and the stale comment corrected rather than fulfilled. It would be a display-only field, and a Contracts enum mirroring the domain's DevicePlatform is exactly the shape of duplication that has produced three self-consistent bugs in this repository. A device list that wants it can add a mapping table and a test pinning the two together, which is what the sync entity types already do. Its own problem code and exception rather than reusing enrollment's, whose rules it largely shares. Registering a device is not enrolling, and a client showing "your enrollment was rejected" because somebody set up a fingerprint reader would be describing the wrong thing. The validation shares the limit constants — MaximumWrapBytes, MaximumDeviceNameLength, PublicKeySize — and not the four-line guards, which would have had to be parameterised over which exception to throw for less than they cost. Both in-memory fakes implement it properly rather than throwing: they record the wrap so a test can assert it arrived, and refuse before enrollment as the real endpoint's policy does. A fake that answered where the server refuses is a fake that can make a real bug pass. 866 tests green, 8 of them new. Zero warnings, dotnet format clean. Still to come: the protector seam with the wrap cached locally so device unlock works offline, then the Windows Hello implementation and the unlock-screen UI — which is where the Windows target framework lands and where automated testing stops.	2026-07-30 13:18:09 +02:00
jaap-jan	7016ce36f1	Key the local cache to the identity, not to the door it was opened through ... Groundwork for a device key, and a spec change rather than a feature. ADR 0007 records the decision it clears the way for: a Windows Hello gesture guarding a protected blob, with the passphrase kept as a permanent fallback. The reason that decision needed this first is that a device key cannot open a session on its own. SessionOpener derived two things from the passphrase master key — the bundle, and the local cache key — and a device wrap is SealTo(device_x25519_pk), which yields the bundle and never computes a master key at all. A device unlock could therefore have opened the identity and still not read the cache it had itself written. So LocalCacheKey now derives from the bundle: dsh1/localcache/v1 → v2, specified in crypto.md §3.2. Every wrap that opens a vault ends up holding the bundle, so every door reaches the same cache. Extract-and-expand, not expand alone. Everything derived from the master key uses HKDF-Expand directly, which is sound because an Argon2id output is uniformly random over its whole length. The bundle's encoding is not — it opens with a fixed 14-byte label and carries a version, a generation and a timestamp before reaching any key material — so it needs the extract step to become a pseudorandom key first. Two consequences fell out, both improvements and neither the point: - A passphrase change no longer discards the local cache. The bundle is unchanged by a re-wrap, so the cache key is too. Under v1 changing a passphrase silently orphaned every cached row and the next launch re-pulled the whole vault. - Recovery-code unlock is fixed before it ships. It derives a different master key from a different secret and a different salt, so under v1 it would have had the same defect as the device path, and nobody would have noticed until it landed. The cache becomes unreadable exactly when the identity is rotated, which is the correct moment to discard it. Existing caches are discarded and re-pulled on upgrade — already the specified behaviour for a stale cache, and the reason the label is versioned rather than reused: a v1 cache must fail to open rather than decrypt to nonsense. One stated guarantee got weaker and now says so. crypto.md §10 claimed locking meant "nothing on disk can be read again without the passphrase." Where a device wrap exists that is no longer true, and it would have been untrue under either candidate design — the alternative was storing a copy of the cache key in the device blob, which is the same door with an extra key lying next to it. The wording now points at ADR 0007, because what guards the device key is a platform decision and not a property of this specification. A golden vector was quietly lying, which is the part worth reading twice. The "local-cache" entry pinned HKDF-SHA512-Expand over a fixed PRK — a construction the cache key no longer uses. Regenerating it would have produced a green suite describing a derivation this code does not perform. It is replaced by a vector over a bundle whose every byte is pinned: the label, version 1, generation 1, a fixed timestamp and two recognisable key scalars, all visible in the fixture so a second implementation can check itself against it. UserSecretBundle.TryDecode is internal for this, because Create draws fresh randomness and so can never produce a reproducible input. Mutation tested, and this one earns its keep: dropping the extract step now fails CommittedVectors_MatchCurrentImplementation. The vector it replaced could not have caught that, because it never touched the bundle at all. One test became false and says so. ARecordSealedUnderAnotherPassphrase is now ARecordSealedByAnotherIdentity: a different passphrase deliberately no longer changes the cache key, and TheLocalCacheKey_SurvivesAPassphraseChange pins that. What must still be unreadable is another user's cache. CacheHarness therefore generates an identity rather than deriving from a passphrase, and has no passphrase parameter left — the cache key is not a question about passphrases any more. SyncHarness's two simulated machines now derive the same cache key, which is what keying on the bundle means: they are the same user holding the same identity. They still have separate cache databases, so nothing is shared between them but the key that would open either. Both harnesses lost a MasterKey field that existed only to make a protector. 858 tests green. Zero warnings, dotnet format clean. Not done: the device key itself. Three pieces remain, and the middle one was a discovery rather than a plan — EnrollmentService.AddDevice runs only during enrollment, so every already-enrolled account, which is all of them, needs an endpoint to add a device wrap while unlocked. The client proves possession by producing the wrap, so that shape falls out of the crypto. After that: the protector seam with the wrap cached locally for offline unlock, then the Hello implementation and the unlock-screen UI, which is where the Windows TFM lands and where automated testing stops.	2026-07-30 12:46:55 +02:00
jaap-jan	c5dec2d68e	Measure the vault column instead of arguing about it ... Nothing in this repository loaded a .axaml, so the one class of defect this window has actually shipped — a control arranged past the edge of its container, where it cannot be clicked — was the one class nothing could catch. The setup screens rendered sliced once, with their buttons unreachable. The vault column is the next candidate: 340 pixels wide, two lists and two editors, and the only thing keeping it from clipping its own Save button at the window's 520-pixel minimum is a state rule that one editor may be open at a time. That rule was added on the strength of an argument. This adds an Avalonia.Headless project that lays real XAML out at a real size and reports what a user could not reach, and the argument is now a number: with both editors open the column overflows, so the rule is load-bearing rather than defensive. BothEditorsAtOnce_DoNotFit_WhichIsWhyTheRuleExists is the test, and it says what to do if it ever starts passing — the column has room, so delete the rule, not the test. Two findings arrived by measuring rather than by reasoning, and the first one changed the design. MainWindow cannot be shown headlessly at all. Showing it attaches the terminal's NativeWebView, whose Win32 adapter initialises WebView2 on attach, and WebView2 refuses a non-STA thread — which is exactly why Program.Main carries [STAThread] and is written down in that comment. A HeadlessUnitTestSession owns its dispatcher thread and offers no apartment choice, so the whole window is out of reach at any size. That is pinned as a test asserting RPC_E_CHANGED_MODE by HResult rather than by message, so a future Avalonia that makes the adapter lazy will fail it and the harness can be widened. So the column had to become its own control to be measurable, which is the extraction the type-selector rework wanted anyway. Keyboard release moved with it: MainWindow used to call Focus() on HostList by name, and now asks VaultColumn.KeyboardTarget. The window decides that the keyboard should leave the terminal and the column decides where it lands — which is the seam the rework needs, because once the column shows one list at a time, "which list owns the keyboard" is a question only the column can answer. The second finding is the way this kind of test lies quietly. The hint class lived in MainWindow.Styles and carries TextWrapping. A Window's styles reach its whole tree, so nothing about the application depended on where it lived — but a control laid out on its own loses them, and every hint paragraph would have measured as a single line. The harness would have passed while measuring heights that were all too small. The three shared classes now live in App.axaml, which changes no rendering and makes the measurement honest. The detector is calibrated in both directions, because a clipping detector that never fires reads as a guarantee: a deliberately clipped Save button is caught by name, and a list longer than its viewport is exempt. Scrolling is how a list is supposed to handle more rows than fit, and without that exemption the host list would fail the moment it had content. It also mis-fired once and the rule is narrower for it — an empty ListBox is zero pixels tall and correct, so "arranged with no size" now applies only to controls the theme gives a height to. Skia rather than the headless drawing stub, deliberately. The stub's font manager invents glyph metrics, and text height is an input to every stacked panel in this column, so measuring against it would produce numbers that are self-consistent and unrelated to the application. A separate test project rather than more tests in DodoSSH.Client.App.Tests. Avalonia's application, dispatcher and platform are process-global singletons initialised once, and that project's identity is the shell's state machine without Avalonia — the whole reason sign-in is a delegate. The fakes needed to reach a real unlocked vault are shared from DodoSSH.Client.Session.Tests by source link: a project reference would make one test project a library of another, and a copy would be a third implementation of the same decision table drifting from the other two. 855 tests green, 10 of them new. Zero warnings, dotnet format clean. Not done, and this is groundwork rather than the item itself: the type selector. The column still holds both lists at once, so a third item type would still recreate the defect the one-editor rule works around. What is different is that the rework can now be checked instead of eyeballed — including the claim it is being made for, that one editor at a time stops being a runtime rule and becomes a fact about what is in the visual tree. What this harness will never catch is the terminal's native child window compositing over Avalonia content. That is a Win32 property of a real window, no headless surface reproduces it, and it is the reason the WebView is collapsed rather than covered.	2026-07-30 11:34:09 +02:00
jaap-jan	211eba0666	Keep host key trust in the vault, and make it withdrawable ... A fingerprint approved once is now approved on every machine and survives a restart, because host key trust is a vault item type rather than a dictionary that dies with the process. InMemoryKnownHostStore was what shipped, so the user was asked to verify a fingerprint on every single connection — which is the gap most likely to train somebody to click through the one warning that actually matters. A warning that appears when nothing is wrong teaches that nothing is ever wrong. The fourth item type, and like the third it cost no sync logic: a row, an EF configuration, a migration, a server kind; a secret, a codec, a merge, a cipher, a repository facade and a session property. One row in the client registry. The reconciler, the mirror, the repository, the outbox and the pull filter were not touched. SyncEntityType.KnownHostKey and AadResourceType.KnownHostKey were already reserved, so neither the contract nor docs/crypto.md changed. One item per (host, port, algorithm), because a server legitimately offers several host keys and which one gets negotiated is not ours to predict. Pinning per endpoint would make an algorithm change indistinguishable from an attack. The label is derived rather than stored, which is the one place this type departs from the other three. A user never names a pin — there is nothing to name it after but the three fields it already has — and a stored label is a second copy of data that can disagree with the first after a merge. Relabel returns the secret unchanged, and says why. The store answers the handshake without touching the disk. SshNetConnectionFactory calls FindAsync from inside SSH.NET's synchronous HostKeyReceived event, over .GetAwaiter().GetResult(), which cannot be avoided; doing SQLite I/O plus an AEAD open per lookup there would put the handshake behind the cache. So decryption happens in OpenAsync and RefreshAsync — on unlock and after each sync pass, exactly where the host and key lists already reload — and FindAsync is a dictionary read under a lock with no await inside it. That snapshot is where the one real bug in this change lived. Install originally merged the live pins over the freshly loaded snapshot, to protect a TrustAsync that had landed while the read was in flight. It would also have resurrected every pin the user had just forgotten, and stopped a withdrawal made on another machine from ever taking effect — the store would have healed the deletion back into existence on every refresh. Replacing wholesale and discarding the read instead is correct because writes are the rare case: every write bumps a generation counter, and a refresh whose stamp is stale throws itself away rather than winning. Nothing found this but reading the method again; it is the kind of mistake that passes every test written before it, because the test that catches it is the one the bug tells you to write. Forgetting is new, and persistence is what made it mandatory rather than convenient. A mismatch is a hard refusal with no way to continue — deliberately, and that stays — so pinning a key permanently is also a way to make a legitimately rebuilt server permanently unreachable. Before this change the pin died at exit and the problem solved itself; now it does not. ForgetAsync drops every algorithm for an endpoint, and it is reachable from the host editor rather than from the warning. Putting it on the mismatch banner would have made it two clicks from "this may be an attack" to "connect anyway", which is the affordance the hard refusal exists to deny. The banner already promised the key could be removed in the host's settings; that promise is now true and points at the button. Trust recorded on another machine becomes visible at the next sync pass, not immediately, and that is a decision rather than an oversight. The failure it produces is a first-contact prompt for a host a colleague approved a minute ago: answerable, and self-correcting on the next pass. The opposite trade — polling the vault on the handshake thread to close a one-minute window — buys nothing and costs the property above. The dangerous direction is not reachable at all: a pin recorded here enters the snapshot as part of recording it, so a refresh can never discard a local trust decision. The server learns nothing, and this is the item type where the temptation was real. A plaintext host column would let a known-hosts screen sort and page without decrypting anything, and it would hand the operator the map of every user's estate — assembled, as these things are, out of facts that are each individually harmless. A host row concedes an address only when relay is switched on and the database refuses to store one otherwise (ADR 0004); there is no equivalent excuse here. The table has no column to put one in, and the EF configuration says so where somebody adding it would be standing. Two things about the migration in this commit are worth knowing, because both came out of getting it wrong. It was hand-written first, including its .Designer.cs, and that version is not what is here. Verifying it turned up something that had been quietly assumed: Migration_AppliedCleanly_WithNoPendingModelChanges does not check the model snapshot. It asserts that migrations applied and that none are pending, which a wrong snapshot satisfies perfectly — the snapshot only matters as the diff base for the *next* migrations add, so an incorrect one passes the whole suite and corrupts the following migration instead. The real check is to generate a throwaway migration and confirm its Up and Down come out empty. They did, and the generated designer was byte-identical to the transcribed one across all 1255 lines, so the hand-written work was in fact correct. Then dotnet ef migrations remove --no-build deleted the wrong migration. With --no-build the tool reads the previously compiled assembly rather than the files on disk, and the probe had just changed which migration was last, so it removed AddKnownHostKeyItem and reverted the snapshot. That turned out to leave exactly the right diff base, so the migration here is EF's own output rather than a transcription — a better outcome than the one that was interrupted, arrived at by accident. Never pass --no-build to migrations remove. Mutation tested, all three sabotages detected: dropping the algorithm from KnownHostIdentity.For, merging instead of replacing in Install, and pointing KnownHostKeyCipher at PortForward — which is what a cast from the wire enum's 10 would silently produce. Each is caught both by an assertion about the mechanism and by a behavioural test that never mentions it; the resource-type sabotage is caught by the table from d10a38d and nothing else, which is what that table is for. The end-to-end slice now approves the real sshd's host key through the vault, pushes it, and reads it back on the second simulated machine — including a check that the server learned no address, and that the second machine answers null for an algorithm never offered. 845 tests green. Zero warnings, dotnet format clean. Three things are deliberately not fixed. A tombstone queued over a create that was never pushed is refused by the server as Invalid and parked; that is pre-existing for all four item types, and the fix belongs in VaultItemRepository.DeleteAsync rather than here. Deleting a host, or changing its address, orphans its pins — both are correct as trust decisions, since a pin describes an endpoint and not a bookmark, but nothing surfaces the leftovers. And there is no interface listing pins at all: trust is created at the connect prompt and withdrawn in the host editor. A known-hosts list is where the orphans would become visible, and it wants the vault column rework first, for the same reason the credential editor does.	2026-07-30 11:00:39 +02:00
jaap-jan	d10a38d8e6	Pin every cipher's AAD resource type from one table, not one test each ... Mutation testing found that pointing CredentialCipher at AadResourceType.Vault passed the entire suite. Every credential test compared the cipher against itself — round trips, cross-type refusals, two-machine sync — and all of those stay true when both halves of one cipher are wrong together, because Seal and TryOpen share the constant. A password sealed under the resource type for a vault encrypts cleanly, decrypts cleanly, syncs cleanly, and violates docs/crypto.md in a way nothing surfaces until another implementation refuses the item. By then the AAD is frozen into stored ciphertext and only clients can re-encrypt it. This is the third time that hole has appeared in this file, and the second time mutation testing rather than review is what found it. So the fix is structural rather than another hand-written test: one table of wire type to resource type, a theory that seals a sample through each cipher and opens it with the resource type the table names — never the one the cipher holds — and a guard asserting the table covers ItemKinds.SyncedTypes. A fourth item type can no longer be added without pinning its resource type: the coverage test fails, and the sample switch throws with an explanation. The two per-cipher tests it replaces said the same thing for hosts and keys, so nothing is lost and the credential row is no longer something someone has to remember. Verified by re-running the mutation matrix. All seven sabotages are now detected: the credential merge dropping its redaction, the key/credential exclusivity check disabled, the schema version ladder flattened so a key-bound host claims the credential version, a credential sending the server an empty fields record instead of none, CredentialKind claiming to be a host, CredentialCipher sealing under the wrong resource type, and the credential noun reading "host". Two of those were unproven before this run — one because the earlier sabotage did not compile, and one because it was genuinely undetected. Sync.Tests 88/88, Domain.Tests 117/117. Zero warnings, dotnet format clean.	2026-07-29 21:17:36 +02:00
jaap-jan	e24012b039	Sync credentials as a vault item type, and bind one to a host ... Closes the largest remaining M1 gap in the data layer: a username and password can live in the vault, sync between machines, and be named by a host as how it authenticates. What is not here is the interface for creating one — see the end of this message. The third item type, and the first one that cost almost nothing to add. Server: a VaultCredential row, an EF configuration, a migration, and a CredentialKind. Client: a secret, a codec, a merge, a cipher, a kind, a repository facade and a session property. No new reconciliation logic, no change to the sync engine, no client cache migration. That was the whole point of the item-kind seam, and this is the evidence it holds. The narrowest type of the three on plaintext, and not for symmetry. A host has a deliberate concession — the relay needs an address it can resolve. A key has a fingerprint, public by nature, which this client still declines to send. A password has no part that is safe to expose: not its length, not a hash, not a hint. So CredentialKind refuses every plaintext field there is, hydrates none, and the table has no column to put one in. HostSecret.CredentialId is the password counterpart of SshKeyId, and the two are mutually exclusive. SSH itself would happily try a key and fall back to a password, but a host naming both leaves "how does this authenticate?" without a single answer — the interface, the connect path and the user would each be free to guess differently. TryValidate refuses it. One consequence was not anticipated: "a full host" stops being a coherent idea, which is what broke AFullHost_RoundTrips and is now written into that test. The schema version became a ladder rather than a maximum: credential-bound is 3, key-bound is 2, neither is still 1. Adding credentials therefore does not drag every key-bound host in every vault onto a version that clients understanding keys perfectly well would refuse to edit. A test pins exactly that, because it is the property the whole content-dependent-version rule exists to provide, and the obvious implementation would quietly lose it. Two tests had become false and said so: - Push_AnUnsupportedEntityType_IsInvalidNotAFailedBatch used Credential as its example of a type this server does not implement. It now asks the server's own registry what is still missing, so it cannot go stale again, and skips with a reason if that set ever empties. - ThePullFilterNamesEveryTypeThisBuildSynchronises pinned the exact list, which is what it is for. Also fixes ten nullable warnings — eight in SyncEndpointTests, two in a test file added earlier today. Neither set was introduced here; both were invisible until an unrelated change forced their project to recompile, which means the zero-warning claims made earlier in this work only ever covered what happened to be rebuilt. 777 tests green. Zero warnings, dotnet format clean. Not done, and deliberately: the credential interface. The vault column is 340 pixels wide and already holds two lists and two editors, kept from clipping its own buttons at the window's minimum height only by the one-editor-at-a-time rule added earlier today. A third list and a third editor would recreate that defect rather than avoid it, so the column needs a shape decision first. Credentials sync; they cannot yet be created in the interface.	2026-07-29 21:09:08 +02:00
jaap-jan	70b3290a77	Bind an SSH key to a host instead of picking one per connection ... A host now names the key it authenticates with, or none, as a field in its encrypted payload — so the choice follows the host to every machine rather than being made again each time somebody connects. The per-connection "Use key" switch it replaces was a stopgap for not having this, and keeping both would have left two mechanisms answering one question. This is the first payload schema version bump, and it does not work the obvious way. A host is written at the *lowest* schema version that can represent it: one that binds a key is written at 2, one that does not is still written at 1, byte for byte as it was before the field existed. The version is what makes an older client refuse to edit an item, so stamping 2 unconditionally would mean upgrading a single machine and renaming a single host made that host uneditable on every machine that had not upgraded yet. Confining the cost to the hosts that actually use the field is the difference between a team noticing a bump and a team being blocked by one. HostSecretCodec states the rule so the next field added follows it, and a test pins the version-1 bytes against a literal rather than against the codec, because the claim is about history: every host already in every vault has to re-encode to what it encoded before, or the first sync after an upgrade would push the whole vault as changed. A binding is an item id, not a copy of the key — a second copy of a private key is one that goes stale — which means the reference can dangle when the key is deleted on another machine. Both places that meets are handled the same way, by refusing rather than falling back: - Connecting to a host whose key is gone is refused outright. A host somebody deliberately set up for key-only access must not quietly start offering a password. - Opening such a host in the editor keeps the binding, selected, labelled as missing. The quieter version of the same failure is someone editing the port and saving, silently converting the host to password authentication with nothing ever having said so. Two things this found by being falsified: - The merge was untested for the new field, and "just take the server's value" passed the entire suite — a local binding change would have been discarded with no conflict recorded. HostSecretMergeTests already had a test written for exactly this class of omission; it simply had not been extended. - Adding a nullable field exposed a defect in HostSecretMerge.Field: it short-circuited when the discarded value was null, so the formatter never ran for the one case where null is a value rather than an absence, and a field whose absence has a name could not report it. Now the formatter always runs, and "no key" appears in the conflict log where an empty string used to. Also fixes eight nullable warnings in SyncEndpointTests left by the server-side SSH key commit, which had omitted the null-forgiving operator the rest of that file uses. They were invisible until an unrelated change forced the project to recompile. The end-to-end slice now binds its host to its key, so a schema-version-2 payload goes through the real API, the real PostgreSQL and back out on a second machine. 745 tests green. Zero warnings, dotnet format clean.	2026-07-29 20:42:51 +02:00
jaap-jan	e3fd3e1728	Sync and authenticate with SSH keys on the client ... Completes the client half of SSH keys: they sync alongside hosts, appear in their own list, and can be selected to authenticate a connection instead of typing a password. The reconciler and the repository were Host-typed throughout, so the choice was to generalise them or to keep a second copy per item type. Generalised, because ItemReconciler's whole premise is that the pull and the push paths must answer the same collision the same way — two copies would drift the first time one of them was fixed. What is genuinely per-type now arrives through IItemKind<TSecret>: the cipher, the merge, the plaintext columns, and the noun to use when telling a person what happened to their item. Generic where the server's IItemKind is not, and for the reason that reverses there — the client needs the concrete type, because it merges field by field. The pull filter is derived from the same registry that builds the reconcilers. That is the specific failure being designed out: an item type that encrypts, merges and lists perfectly and is never once requested from the server, so it works on the machine that made it and exists nowhere else. No client cache migration. The item table's primary key and the outbox's unique index already carry the entity type, and AadResourceTypes already mapped SshKey — so a host and a key may share an id and never see each other's rows, which SshKeySyncTests now arranges deliberately. A key hands the server nothing in plaintext. There is a public_key_fingerprint column and it would be accepted; leaving it null is deliberate. A fingerprint is not secret but it is a stable identifier for a key pair, so filling it would let an operator tell which of their users hold the same key and correlate one across vaults, for a column nothing reads. The design allows itself one plaintext concession — the relay address, which the relay cannot work without — and this is not that. A key is chosen per connection rather than bound to a host, which works the way ssh -i does. Binding one needs a field on HostSecret and therefore a payload schema bump, which makes every host written afterwards read-only on an older build; worth doing deliberately rather than as a side effect of adding keys. Three things this found, all of them by being falsified rather than by review: - Making the reconciler generic silently turned a record comparison into reference equality, because == on a type parameter is not value equality. The effect would have been a conflict recorded on every pass for an unacknowledged create that had in fact landed. Sabotaging the fix left all 73 tests passing — nothing covered that branch — so ConflictMatrixTests now has AnUnacknowledgedCreateThatDidLand_IsDroppedQuietly, which fails without it. - A test asserting that a blank passphrase reaches SSH.NET as null was vacuous: it exercised the editor, not the credential path, and passed with the guard deleted. Resolved by making SshKeySecret.Passphrase normalise an empty string to null, so there is one spelling of one state — which also keeps two clients from producing different payload bytes for an identical key. That exposed a wider gap: SshKeySecret, its codec and its merge had no direct unit tests at all. They have 25 now. - The reason first given for that normalisation was false. It claimed SSH.NET rejects a passphrase supplied for an unprotected key; measured against a real sshd it ignores it and authenticates anyway. Corrected everywhere it was stated and recorded in docs/platform-flags.md. The same test file also closes a real hole: SshPrivateKeyCredential had never been exercised against a server, because the existing key test builds SSH.NET's auth method directly and bypasses the path a vault-held key actually takes. Only one editor may be open at a time. Both sit in the same 340-pixel column as Auto rows and their heights together exceed it at the window's minimum size, so two open editors put the lower one's Save and Cancel past the bottom edge — the same failure this window already shipped once with the setup screens. Expressed as a state rule because that is the only form of it this repository can check: nothing here loads a .axaml. The refusal keeps what was typed, since in the key editor that is a pasted private key the user may have nowhere else. The end-to-end slice now carries a key as well as a host, so both item types go through the real API, the real PostgreSQL and the real crypto in one pass — the three hand-kept mappings between enums that do not line up are the reason that is worth doing rather than trusting the unit suites. 735 tests green, including the container-backed SSH and end-to-end suites. Zero warnings, dotnet format clean.	2026-07-29 20:27:23 +02:00
jaap-jan	586cb303d5	Merge branch 'claude/gallant-brahmagupta-1f8244' ... Writes down that locking the vault leaves shells running, and shows the count on the unlock screen rather than leaving it to be inferred. Conflict resolution: - ShellFlowTests' fixture keeps main's FakeSshConnectionFactory. The branch added an IdleSshConnectionFactory for exactly what main's fake already does — a shell that is open, silent and never closes on its own — so FakeSshConnections.cs is dropped rather than merged, leaving one fake SSH stack in the suite instead of two that would drift apart. - MainWindowViewModel and TerminalWorkspace: both sides added their own members, so both are kept. - TerminalWorkspaceTests was added by both branches, with the renderer gate on one side and session lifetime on the other. Merged into one class over one set of helpers; the gate tests now use FakeConnectionFactory rather than an NSubstitute stub, since the suite already has the fake. gallant's polling Timeout constant is PollTimeout, which no longer reads as the renderer's. - platform-flags.md keeps main's measured focus section and drops the short "nothing hands the terminal keyboard focus" entry the branch still carried, which that section supersedes. One genuine disagreement between the branches, left visible rather than flattened: this branch measured that a collapsed WebView cannot be typed into and attributed it to a hidden WS_CHILD window being ineligible for keyboard focus, while main's focus work measured Win32 focus still held by that hidden window and added a lock path that moves the keyboard off it. Both results stand; the mechanism sentence now defers to the focus entry, which makes the input barrier something the lock path maintains rather than something the platform guarantees. Full suite green, including the container-backed SSH tests.	2026-07-29 15:46:41 +02:00
jaap-jan	74341d41e0	Merge branch 'claude/distracted-ritchie-53fc70' ... Bounds the renderer wait, so a WebView2 that never initialises reports itself instead of hanging Connect with the busy flag stuck. Conflict resolution, all of it in the App test suite, which main had changed under the branch when sleepy-chebyshev landed: - The workspace fixture keeps main's fake SSH factory and its FakeRenderer-aware page, and takes the branch's RendererTimeout on top. One second rather than the branch's 250 ms, because the timeout now also bounds FakeRenderer's own wait for the attach it just made. - FakeRenderer arrived on main after the branch was cut and still called the no-argument WaitForRendererAsync. Both sides merged cleanly and left the build broken; it now passes its own token. - ConnectingWithNoRenderer's remark claimed the suite never starts the workspace and never attaches a renderer. Both are false here, so it now says what is true of the test: it is the one connect test that attaches no renderer.	2026-07-29 15:40:25 +02:00