Five boxes on this head take a secret and every one of them was drawing dots
and saying nothing. `PasswordChar` is a screen property: Windows has no opinion
about what is being typed into a text box, so the desktop head needs nothing
more. Android's software keyboard has an opinion, and left at its default it
read a vault passphrase as prose — completions offered in the suggestion strip
above the box, and the passphrase itself learned into the IME's dictionary.
Dots on screen with a word bar over them is the worst of both: hidden from the
person typing it and offered to the room. `TextInputOptions.ContentType` is the
property the Android backend maps onto `InputType`, and it is what turns both
off. Both attributes now live in one `TextBox.secret` class rather than being
repeated per box, because they are two halves of one fact and the next box
added would have got one of them.
The keyboard also went on covering whichever box had raised it. That is in
`PhoneShell` rather than on each screen: everything the phone draws is inside
its one root panel, so a bottom margin shortens all eleven screens at once, and
a screen added later cannot forget to handle something it never had to know
about.
Two mechanisms, and it matters that neither is a backstop for the other. Before
Android 15 the activity now declares `AdjustResize` and the platform shortens
the window itself; left unspecified Android chooses, and what it chooses for a
window whose entire content is one native view — which is what an Avalonia
surface is — is to pan, sliding the window by however much it thinks the
focused native view needs and leaving the box exactly where it was. That was
the bug. From Android 15 the attribute is ignored, edge-to-edge being enforced
and the window no longer resized for the keyboard at all, and the reported
inset is what there is. Each is dead where the other applies — where the window
resizes, the inset arrives already consumed and measures zero — which is why
the margin comes from the inset alone. Both added together would strand the
interface an entire keyboard above the keyboard.
Scrolling the box back into view keys off the size change rather than off
either mechanism. `ScrollViewer` already brings a newly focused child into
view; what it cannot know is that the visible region shrank after the focus,
and both ways of losing that region end in the same resize.
None of it is reachable by a test. The software keyboard is an inset the
platform reports and a headless top level reports none, so phase 10 of
`docs/manual-checks.md` is the whole of the verification — including the note
to run it on one device each side of Android 15, since a build exercised on
only one of the two will look correct and be half broken.
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.
jaap-jan