Public Access
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.
This commit is contained in:
@@ -0,0 +1,30 @@
|
||||
<Project Sdk="Microsoft.NET.Sdk">
|
||||
|
||||
<!--
|
||||
S3-compatible buckets as a remote in the file browser.
|
||||
|
||||
Its own project rather than more of DodoSSH.Client.Transfer, because the two answer different
|
||||
questions — that one is about moving bytes and what to do when moving them stops halfway, this
|
||||
one is about one protocol's idea of what a file is — and because the AWS SDK belongs to exactly
|
||||
one project rather than to the whole client.
|
||||
|
||||
It references DodoSSH.Client.Ssh for two types: IRemoteFileStore and SftpEntry. That reads
|
||||
oddly and is deliberate; the reasoning is on IRemoteFileStore itself, and the short version is
|
||||
that moving them would rename a record the entire file browser is written against.
|
||||
-->
|
||||
|
||||
<ItemGroup>
|
||||
<ProjectReference Include="../DodoSSH.Client.Domain/DodoSSH.Client.Domain.csproj" />
|
||||
<ProjectReference Include="../DodoSSH.Client.Ssh/DodoSSH.Client.Ssh.csproj" />
|
||||
</ItemGroup>
|
||||
|
||||
<ItemGroup>
|
||||
<PackageReference Include="AWSSDK.S3" />
|
||||
<PackageReference Include="AWSSDK.Core" />
|
||||
</ItemGroup>
|
||||
|
||||
<ItemGroup>
|
||||
<InternalsVisibleTo Include="DodoSSH.Client.ObjectStore.Tests" />
|
||||
</ItemGroup>
|
||||
|
||||
</Project>
|
||||
@@ -0,0 +1,57 @@
|
||||
namespace DodoSSH.Client.ObjectStore;
|
||||
|
||||
/// <summary>
|
||||
/// Translating between the paths a file browser uses and the keys a bucket has.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// <para>
|
||||
/// <b>A bucket has no directories.</b> It has keys, which are strings, and a convention that <c>/</c> in a
|
||||
/// key means what it means in a path. Everything in this class is that convention written down in one place,
|
||||
/// because the alternative is the same three lines of trimming repeated at every call site with one of them
|
||||
/// subtly different.
|
||||
/// </para>
|
||||
/// <para>
|
||||
/// The browser's side is an absolute POSIX path — <c>/reports/2026/q3.csv</c> — because that is what the
|
||||
/// screen, the breadcrumb trail and the transfer queue already speak. The bucket's side is a key with no
|
||||
/// leading slash: <c>reports/2026/q3.csv</c>. The root is <c>/</c> on one side and the empty string on the
|
||||
/// other, which is the case every one of these methods is really about.
|
||||
/// </para>
|
||||
/// </remarks>
|
||||
internal static class ObjectKeys
|
||||
{
|
||||
/// <summary>The path a file browser opens on.</summary>
|
||||
internal const string Root = "/";
|
||||
|
||||
/// <summary>The object key for a browser path.</summary>
|
||||
internal static string ToKey(string path) => path.TrimStart('/');
|
||||
|
||||
/// <summary>The browser path for an object key.</summary>
|
||||
internal static string ToPath(string key) => Root + key.TrimStart('/');
|
||||
|
||||
/// <summary>
|
||||
/// The prefix that lists one directory's immediate contents.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// Trailing slash, always, and empty for the root. Without it a listing of <c>/reports</c> would also
|
||||
/// return <c>/reports-archive</c>, because a prefix match knows nothing about path segments.
|
||||
/// </remarks>
|
||||
internal static string ToPrefix(string path)
|
||||
{
|
||||
var key = ToKey(path);
|
||||
|
||||
return key.Length == 0 || key.EndsWith('/') ? key : key + "/";
|
||||
}
|
||||
|
||||
/// <summary>The last segment of a key, which is what a row shows.</summary>
|
||||
/// <remarks>
|
||||
/// Trailing slashes are removed first, so the common prefix <c>reports/2026/</c> yields <c>2026</c>
|
||||
/// rather than an empty string.
|
||||
/// </remarks>
|
||||
internal static string NameOf(string key)
|
||||
{
|
||||
var trimmed = key.TrimEnd('/');
|
||||
var slash = trimmed.LastIndexOf('/');
|
||||
|
||||
return slash < 0 ? trimmed : trimmed[(slash + 1)..];
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,69 @@
|
||||
using Amazon;
|
||||
using Amazon.Runtime;
|
||||
using Amazon.S3;
|
||||
using DodoSSH.Client.Domain;
|
||||
using DodoSSH.Client.Ssh;
|
||||
|
||||
namespace DodoSSH.Client.ObjectStore;
|
||||
|
||||
/// <summary>Opens a bucket as a place with files in it.</summary>
|
||||
/// <remarks>
|
||||
/// An interface so the file screen can be tested without a bucket, exactly as <c>ISftpSessionFactory</c> is
|
||||
/// what lets it be tested without a host.
|
||||
/// </remarks>
|
||||
public interface IObjectStoreFactory
|
||||
{
|
||||
/// <summary>Builds a client for one bucket.</summary>
|
||||
/// <param name="store">The bucket and its credentials, decrypted.</param>
|
||||
/// <remarks>
|
||||
/// Synchronous and cheap: nothing is contacted here. S3 is request-per-operation, so there is no
|
||||
/// connect step to fail — the first thing that can fail is the first listing, which is where the
|
||||
/// credentials and the endpoint are actually tested.
|
||||
/// </remarks>
|
||||
IRemoteFileStore Open(ObjectStoreSecret store);
|
||||
}
|
||||
|
||||
/// <summary>Opens buckets with the AWS SDK.</summary>
|
||||
public sealed class S3ObjectStoreFactory : IObjectStoreFactory
|
||||
{
|
||||
/// <inheritdoc />
|
||||
public IRemoteFileStore Open(ObjectStoreSecret store)
|
||||
{
|
||||
ArgumentNullException.ThrowIfNull(store);
|
||||
|
||||
if (!store.TryValidate(out var reason))
|
||||
{
|
||||
throw new ArgumentException(reason, nameof(store));
|
||||
}
|
||||
|
||||
var config = new AmazonS3Config
|
||||
{
|
||||
// On for nearly every self-hosted service and off for AWS. It is a stored setting rather than
|
||||
// something inferred from the endpoint, because inferring it wrongly produces a DNS failure that
|
||||
// says nothing about buckets — see ObjectStoreSecret.UsePathStyle.
|
||||
ForcePathStyle = store.UsePathStyle,
|
||||
};
|
||||
|
||||
if (store.Endpoint is { } endpoint)
|
||||
{
|
||||
config.ServiceURL = endpoint;
|
||||
|
||||
// Still set when there is one, because SigV4 signs the region into every request and several
|
||||
// S3-compatible services check it. The ones that do not, ignore it.
|
||||
if (store.Region is { } named)
|
||||
{
|
||||
config.AuthenticationRegion = named;
|
||||
}
|
||||
}
|
||||
else
|
||||
{
|
||||
// No endpoint means Amazon, and then the region is what resolves the host. Validation has
|
||||
// already refused the case where neither is set.
|
||||
config.RegionEndpoint = RegionEndpoint.GetBySystemName(store.Region!);
|
||||
}
|
||||
|
||||
var credentials = new BasicAWSCredentials(store.AccessKeyId, store.SecretAccessKey);
|
||||
|
||||
return new S3FileStore(new AmazonS3Client(credentials, config), store.Bucket);
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,449 @@
|
||||
using Amazon.S3;
|
||||
using Amazon.S3.Model;
|
||||
using Amazon.S3.Transfer;
|
||||
using DodoSSH.Client.Ssh;
|
||||
|
||||
namespace DodoSSH.Client.ObjectStore;
|
||||
|
||||
/// <summary>
|
||||
/// One S3-compatible bucket, as a place with files in it.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// <para>
|
||||
/// <b>A bucket is not a filesystem, and the three places that matter are documented on the members rather
|
||||
/// than smoothed over.</b> There are no directories, only keys with slashes in them; an object cannot be
|
||||
/// appended to, so an interrupted upload cannot resume; and there is no rename, only copy-then-delete. Each
|
||||
/// is refused with a reason or implemented with its cost stated, because a file browser that quietly did
|
||||
/// something adjacent would be worse than one that said no.
|
||||
/// </para>
|
||||
/// <para>
|
||||
/// <b>Listings are one page.</b> <c>ListObjectsV2</c> returns up to a thousand keys and this asks for one
|
||||
/// page, so a prefix with more than that in it is shown truncated — which the screen says out loud. Paging
|
||||
/// the whole way through a bucket with a million objects under one prefix is a request storm behind a
|
||||
/// scrollbar nobody asked for; the filter box is the answer, and a prefix that large is not a directory
|
||||
/// anybody browses.
|
||||
/// </para>
|
||||
/// </remarks>
|
||||
internal sealed class S3FileStore : IRemoteFileStore
|
||||
{
|
||||
/// <summary>
|
||||
/// The most keys one listing asks for.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// The service's own maximum. Asking for less would page more often for no benefit; asking for more is
|
||||
/// not possible.
|
||||
/// </remarks>
|
||||
private const int PageSize = 1000;
|
||||
|
||||
private readonly IAmazonS3 client;
|
||||
private readonly string bucket;
|
||||
private int disposed;
|
||||
|
||||
internal S3FileStore(IAmazonS3 client, string bucket)
|
||||
{
|
||||
this.client = client;
|
||||
this.bucket = bucket;
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Always true, because there is no connection to be up.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// S3 is request-per-operation over HTTPS; there is no session to drop and nothing to poll. Answering
|
||||
/// false when the network is down would be a claim this type cannot make without a request of its own,
|
||||
/// and every operation already reports its own failure.
|
||||
/// </remarks>
|
||||
public bool IsConnected => Volatile.Read(ref disposed) == 0;
|
||||
|
||||
/// <inheritdoc />
|
||||
public string HomeDirectory => ObjectKeys.Root;
|
||||
|
||||
/// <summary>
|
||||
/// Lists one prefix: its immediate sub-prefixes as directories, its immediate keys as files.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// The delimiter is what makes this a directory listing rather than a recursive walk — without it, a
|
||||
/// listing of the root returns every object in the bucket. Common prefixes come back as directories;
|
||||
/// the marker object some tools write for a "folder" is dropped, because it is the directory itself and
|
||||
/// showing it would put an empty-named row inside every one.
|
||||
/// </remarks>
|
||||
public async Task<IReadOnlyList<SftpEntry>> ListAsync(string path, CancellationToken cancellationToken)
|
||||
{
|
||||
ArgumentNullException.ThrowIfNull(path);
|
||||
|
||||
var prefix = ObjectKeys.ToPrefix(path);
|
||||
|
||||
ListObjectsV2Response response;
|
||||
try
|
||||
{
|
||||
response = await client.ListObjectsV2Async(
|
||||
new ListObjectsV2Request
|
||||
{
|
||||
BucketName = bucket,
|
||||
Prefix = prefix,
|
||||
Delimiter = "/",
|
||||
MaxKeys = PageSize,
|
||||
},
|
||||
cancellationToken).ConfigureAwait(false);
|
||||
}
|
||||
catch (AmazonS3Exception exception)
|
||||
{
|
||||
throw new SftpPathException(path, Describe(exception), exception);
|
||||
}
|
||||
|
||||
return Project(response, prefix);
|
||||
}
|
||||
|
||||
/// <summary>Turns one listing into rows a file browser can show.</summary>
|
||||
/// <remarks>
|
||||
/// Directories first and then by name, which is the order every caller of this interface expects and
|
||||
/// what saves the screen sorting it again.
|
||||
/// </remarks>
|
||||
private static IReadOnlyList<SftpEntry> Project(ListObjectsV2Response response, string prefix)
|
||||
{
|
||||
var entries = new List<SftpEntry>();
|
||||
|
||||
foreach (var common in response.CommonPrefixes ?? [])
|
||||
{
|
||||
entries.Add(new SftpEntry(
|
||||
ObjectKeys.NameOf(common),
|
||||
ObjectKeys.ToPath(common),
|
||||
SftpEntryKind.Directory,
|
||||
Length: 0,
|
||||
LastWriteTimeUtc: default,
|
||||
|
||||
// Blank rather than invented. A bucket has no POSIX mode, and printing drwxr-xr-x beside a
|
||||
// prefix would be a fact this store made up.
|
||||
Permissions: string.Empty));
|
||||
}
|
||||
|
||||
foreach (var item in response.S3Objects ?? [])
|
||||
{
|
||||
// The marker object for this prefix itself, which several tools write to make a folder appear
|
||||
// in a web console. It is this directory, not something in it.
|
||||
if (string.Equals(item.Key, prefix, StringComparison.Ordinal))
|
||||
{
|
||||
continue;
|
||||
}
|
||||
|
||||
entries.Add(new SftpEntry(
|
||||
ObjectKeys.NameOf(item.Key),
|
||||
ObjectKeys.ToPath(item.Key),
|
||||
SftpEntryKind.File,
|
||||
item.Size ?? 0,
|
||||
Utc(item.LastModified),
|
||||
Permissions: string.Empty));
|
||||
}
|
||||
|
||||
return
|
||||
[
|
||||
.. entries
|
||||
.OrderByDescending(entry => entry.Kind is SftpEntryKind.Directory)
|
||||
.ThenBy(entry => entry.Name, StringComparer.OrdinalIgnoreCase),
|
||||
];
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// The SDK's timestamp as an unambiguous instant.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// Stated rather than converted implicitly. S3 returns <c>Last-Modified</c> in UTC and the SDK hands it
|
||||
/// over as a <see cref="DateTime"/> whose <c>Kind</c> is not reliably set — so an implicit conversion
|
||||
/// would read it as local time on some paths and shift every timestamp in the listing by the machine's
|
||||
/// offset. The file browser shows this column beside an SFTP one.
|
||||
/// </remarks>
|
||||
private static DateTimeOffset Utc(DateTime? moment) =>
|
||||
moment is { } value
|
||||
? new DateTimeOffset(DateTime.SpecifyKind(value, DateTimeKind.Utc))
|
||||
: default;
|
||||
|
||||
/// <summary>
|
||||
/// What one path is, or null when nothing is there.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// Two requests in the worst case, because a bucket cannot answer "is this a directory" directly: a
|
||||
/// HEAD tells us whether an object with that exact key exists, and only a listing can tell us whether
|
||||
/// anything lives under it as a prefix. The order matters — a key can be both, and the object is the
|
||||
/// more specific answer.
|
||||
/// </remarks>
|
||||
public async Task<SftpEntry?> StatAsync(string path, CancellationToken cancellationToken)
|
||||
{
|
||||
ArgumentNullException.ThrowIfNull(path);
|
||||
|
||||
var key = ObjectKeys.ToKey(path);
|
||||
|
||||
if (key.Length == 0)
|
||||
{
|
||||
return new SftpEntry(
|
||||
string.Empty, ObjectKeys.Root, SftpEntryKind.Directory, 0, default, string.Empty);
|
||||
}
|
||||
|
||||
try
|
||||
{
|
||||
var head = await client.GetObjectMetadataAsync(
|
||||
new GetObjectMetadataRequest { BucketName = bucket, Key = key },
|
||||
cancellationToken).ConfigureAwait(false);
|
||||
|
||||
return new SftpEntry(
|
||||
ObjectKeys.NameOf(key),
|
||||
ObjectKeys.ToPath(key),
|
||||
SftpEntryKind.File,
|
||||
head.ContentLength,
|
||||
Utc(head.LastModified),
|
||||
Permissions: string.Empty);
|
||||
}
|
||||
catch (AmazonS3Exception exception) when (exception.StatusCode == System.Net.HttpStatusCode.NotFound)
|
||||
{
|
||||
// Not an object. It may still be a prefix with things under it, which is what a browser means
|
||||
// by a directory.
|
||||
}
|
||||
|
||||
var listing = await client.ListObjectsV2Async(
|
||||
new ListObjectsV2Request
|
||||
{
|
||||
BucketName = bucket,
|
||||
Prefix = ObjectKeys.ToPrefix(path),
|
||||
MaxKeys = 1,
|
||||
},
|
||||
cancellationToken).ConfigureAwait(false);
|
||||
|
||||
return listing.KeyCount > 0
|
||||
? new SftpEntry(
|
||||
ObjectKeys.NameOf(key),
|
||||
ObjectKeys.ToPath(key),
|
||||
SftpEntryKind.Directory,
|
||||
0,
|
||||
default,
|
||||
string.Empty)
|
||||
: null;
|
||||
}
|
||||
|
||||
/// <inheritdoc />
|
||||
public async Task<Stream> OpenReadAsync(string path, long offset, CancellationToken cancellationToken)
|
||||
{
|
||||
ArgumentNullException.ThrowIfNull(path);
|
||||
ArgumentOutOfRangeException.ThrowIfNegative(offset);
|
||||
|
||||
var request = new GetObjectRequest { BucketName = bucket, Key = ObjectKeys.ToKey(path) };
|
||||
|
||||
if (offset > 0)
|
||||
{
|
||||
// A ranged GET, which is what makes an interrupted download resumable — and the one place where
|
||||
// a bucket is better at this than SFTP, because the range is part of the protocol rather than a
|
||||
// seek on an open handle.
|
||||
request.ByteRange = new ByteRange(offset, long.MaxValue);
|
||||
}
|
||||
|
||||
try
|
||||
{
|
||||
var response = await client.GetObjectAsync(request, cancellationToken).ConfigureAwait(false);
|
||||
|
||||
return response.ResponseStream;
|
||||
}
|
||||
catch (AmazonS3Exception exception)
|
||||
{
|
||||
throw new SftpPathException(path, Describe(exception), exception);
|
||||
}
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Opens an object for writing, from the beginning.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// <para>
|
||||
/// <b>A non-zero offset is refused, and this is the one capability a bucket genuinely does not have.</b>
|
||||
/// Objects are immutable: there is no append, and no way to write into the middle of one. Multipart
|
||||
/// upload can rebuild an interrupted transfer, but only by keeping the upload id and every part's ETag
|
||||
/// across the interruption — state this store would have to persist somewhere, on behalf of a queue that
|
||||
/// already has its own idea of what resuming means. Refusing with a reason is the honest answer;
|
||||
/// silently starting from zero would corrupt a resumed file.
|
||||
/// </para>
|
||||
/// <para>
|
||||
/// The returned stream is the writing half of a pipe. A background upload reads the other half and
|
||||
/// chunks it into parts, so a large file never lands on disk twice and memory stays bounded by the part
|
||||
/// size — which is what the alternative, buffering to a temporary file and putting it afterwards, would
|
||||
/// have cost.
|
||||
/// </para>
|
||||
/// </remarks>
|
||||
public Task<Stream> OpenWriteAsync(string path, long offset, CancellationToken cancellationToken)
|
||||
{
|
||||
ArgumentNullException.ThrowIfNull(path);
|
||||
|
||||
if (offset != 0)
|
||||
{
|
||||
throw new SftpPathException(
|
||||
path,
|
||||
"An object cannot be written to from the middle, so an interrupted upload to a bucket "
|
||||
+ "starts again rather than resuming.");
|
||||
}
|
||||
|
||||
return Task.FromResult<Stream>(
|
||||
new S3UploadStream(client, bucket, ObjectKeys.ToKey(path), cancellationToken));
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Creates the marker object that makes an empty prefix visible.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// A zero-byte object whose key ends in <c>/</c>, which is the convention every S3 console and most
|
||||
/// tools use. It is not a directory — nothing in the service knows what one is — and it disappears by
|
||||
/// itself once real objects live under the prefix, which is why the listing above drops it.
|
||||
/// </remarks>
|
||||
public async Task CreateDirectoryAsync(string path, CancellationToken cancellationToken)
|
||||
{
|
||||
ArgumentNullException.ThrowIfNull(path);
|
||||
|
||||
var prefix = ObjectKeys.ToPrefix(path);
|
||||
|
||||
if (prefix.Length == 0)
|
||||
{
|
||||
throw new SftpPathException(path, "The root of a bucket already exists.");
|
||||
}
|
||||
|
||||
try
|
||||
{
|
||||
await client.PutObjectAsync(
|
||||
new PutObjectRequest
|
||||
{
|
||||
BucketName = bucket,
|
||||
Key = prefix,
|
||||
ContentBody = string.Empty,
|
||||
},
|
||||
cancellationToken).ConfigureAwait(false);
|
||||
}
|
||||
catch (AmazonS3Exception exception)
|
||||
{
|
||||
throw new SftpPathException(path, Describe(exception), exception);
|
||||
}
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Deletes one object, or an empty prefix's marker.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// Deliberately not recursive, matching SFTP's own rule and for the same reason: a recursive delete
|
||||
/// against a bucket is the one operation on this screen that can destroy something no undo reaches. A
|
||||
/// prefix with anything under it is refused and says so.
|
||||
/// </remarks>
|
||||
public async Task DeleteAsync(string path, CancellationToken cancellationToken)
|
||||
{
|
||||
ArgumentNullException.ThrowIfNull(path);
|
||||
|
||||
var key = ObjectKeys.ToKey(path);
|
||||
|
||||
if (key.Length == 0)
|
||||
{
|
||||
throw new SftpPathException(path, "A bucket cannot delete its own root.");
|
||||
}
|
||||
|
||||
if (await StatAsync(path, cancellationToken).ConfigureAwait(false) is { Kind: SftpEntryKind.Directory })
|
||||
{
|
||||
var listing = await client.ListObjectsV2Async(
|
||||
new ListObjectsV2Request
|
||||
{
|
||||
BucketName = bucket,
|
||||
Prefix = ObjectKeys.ToPrefix(path),
|
||||
MaxKeys = 2,
|
||||
},
|
||||
cancellationToken).ConfigureAwait(false);
|
||||
|
||||
// One key is the marker object for this prefix itself; anything more is contents.
|
||||
if (listing.KeyCount > 1)
|
||||
{
|
||||
throw new SftpPathException(
|
||||
path, "There are still objects under this prefix, so it was not deleted.");
|
||||
}
|
||||
|
||||
key = ObjectKeys.ToPrefix(path);
|
||||
}
|
||||
|
||||
try
|
||||
{
|
||||
await client.DeleteObjectAsync(
|
||||
new DeleteObjectRequest { BucketName = bucket, Key = key },
|
||||
cancellationToken).ConfigureAwait(false);
|
||||
}
|
||||
catch (AmazonS3Exception exception)
|
||||
{
|
||||
throw new SftpPathException(path, Describe(exception), exception);
|
||||
}
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Copies to the new key and deletes the old one, which is what a bucket has instead of rename.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// <para>
|
||||
/// Not atomic, and it cannot be. Between the two requests both keys exist; if the delete fails, both
|
||||
/// still do. The copy is server-side — no bytes come to this machine — so the window is short, but it is
|
||||
/// real and a failure leaves a duplicate rather than a loss, which is the safe direction.
|
||||
/// </para>
|
||||
/// <para>
|
||||
/// Only objects. Renaming a prefix means copying every key under it, which is a bulk operation wearing
|
||||
/// a rename's clothing, and the failure mode is a half-moved directory.
|
||||
/// </para>
|
||||
/// </remarks>
|
||||
public async Task RenameAsync(string fromPath, string toPath, CancellationToken cancellationToken)
|
||||
{
|
||||
ArgumentNullException.ThrowIfNull(fromPath);
|
||||
ArgumentNullException.ThrowIfNull(toPath);
|
||||
|
||||
if (await StatAsync(fromPath, cancellationToken).ConfigureAwait(false)
|
||||
is not { Kind: SftpEntryKind.File })
|
||||
{
|
||||
throw new SftpPathException(
|
||||
fromPath,
|
||||
"Only an object can be renamed in a bucket. A prefix would have to be copied key by key.");
|
||||
}
|
||||
|
||||
try
|
||||
{
|
||||
await client.CopyObjectAsync(
|
||||
new CopyObjectRequest
|
||||
{
|
||||
SourceBucket = bucket,
|
||||
SourceKey = ObjectKeys.ToKey(fromPath),
|
||||
DestinationBucket = bucket,
|
||||
DestinationKey = ObjectKeys.ToKey(toPath),
|
||||
},
|
||||
cancellationToken).ConfigureAwait(false);
|
||||
|
||||
await client.DeleteObjectAsync(
|
||||
new DeleteObjectRequest { BucketName = bucket, Key = ObjectKeys.ToKey(fromPath) },
|
||||
cancellationToken).ConfigureAwait(false);
|
||||
}
|
||||
catch (AmazonS3Exception exception)
|
||||
{
|
||||
throw new SftpPathException(fromPath, Describe(exception), exception);
|
||||
}
|
||||
}
|
||||
|
||||
/// <inheritdoc />
|
||||
public ValueTask DisposeAsync()
|
||||
{
|
||||
if (Interlocked.Exchange(ref disposed, 1) == 0)
|
||||
{
|
||||
client.Dispose();
|
||||
}
|
||||
|
||||
return ValueTask.CompletedTask;
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// What went wrong, in words that name the bucket rather than the protocol.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// The SDK's own messages are accurate and unhelpful at a file browser: "The specified key does not
|
||||
/// exist" is fine, and "Access Denied" against a bucket somebody has just typed the keys for is the
|
||||
/// moment to say which of the two is more likely.
|
||||
/// </remarks>
|
||||
private static string Describe(AmazonS3Exception exception) => exception.StatusCode switch
|
||||
{
|
||||
System.Net.HttpStatusCode.NotFound => "There is nothing at that key.",
|
||||
System.Net.HttpStatusCode.Forbidden =>
|
||||
"The bucket refused that. Check the access key and what it is allowed to do.",
|
||||
System.Net.HttpStatusCode.BadRequest when exception.ErrorCode is "AuthorizationHeaderMalformed" =>
|
||||
"The bucket is in a different region to the one configured.",
|
||||
_ => exception.Message,
|
||||
};
|
||||
}
|
||||
@@ -0,0 +1,204 @@
|
||||
using System.IO.Pipelines;
|
||||
using Amazon.S3;
|
||||
using Amazon.S3.Transfer;
|
||||
|
||||
namespace DodoSSH.Client.ObjectStore;
|
||||
|
||||
/// <summary>
|
||||
/// A stream you write an object into.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// <para>
|
||||
/// <b>The direction is the whole problem.</b> The transfer queue asks for somewhere to write and then copies
|
||||
/// a local file into it; the S3 SDK wants a stream it can read from. Something has to bridge the two, and
|
||||
/// there are only three ways to do it: buffer the whole object to a temporary file and upload afterwards
|
||||
/// (correct, and doubles the disk a big upload costs), hold it in memory (correct until somebody uploads a
|
||||
/// disc image), or run the upload concurrently and hand back the writing half of a pipe.
|
||||
/// </para>
|
||||
/// <para>
|
||||
/// This is the third. <see cref="TransferUtility"/> reads the pipe and splits it into multipart chunks, so
|
||||
/// memory stays bounded by the part size however large the object is, and nothing lands on disk twice.
|
||||
/// </para>
|
||||
/// <para>
|
||||
/// <b>Completion is on <see cref="DisposeAsync"/>, and it is not optional.</b> The upload is only finished
|
||||
/// when the pipe is completed and the background task has been awaited — so a caller that abandons this
|
||||
/// stream without disposing it leaves an upload running against a bucket. That is the same contract every
|
||||
/// stream has; it is written down because the consequence here is remote rather than local.
|
||||
/// </para>
|
||||
/// <para>
|
||||
/// <b>A failed upload has to surface at the writer.</b> If the service refuses halfway, the reading half
|
||||
/// stops and this stream's next <c>WriteAsync</c> would otherwise block for ever — so the background task's
|
||||
/// completion also completes the pipe's reader with the exception, which is what makes the write throw with
|
||||
/// the real reason rather than hang.
|
||||
/// </para>
|
||||
/// </remarks>
|
||||
internal sealed class S3UploadStream : Stream
|
||||
{
|
||||
private readonly Pipe pipe = new();
|
||||
private readonly Task upload;
|
||||
private readonly CancellationToken cancellationToken;
|
||||
private int disposed;
|
||||
|
||||
internal S3UploadStream(
|
||||
IAmazonS3 client,
|
||||
string bucket,
|
||||
string key,
|
||||
CancellationToken cancellationToken)
|
||||
{
|
||||
this.cancellationToken = cancellationToken;
|
||||
|
||||
upload = UploadAsync(client, bucket, key);
|
||||
}
|
||||
|
||||
/// <inheritdoc />
|
||||
public override bool CanRead => false;
|
||||
|
||||
/// <inheritdoc />
|
||||
public override bool CanSeek => false;
|
||||
|
||||
/// <inheritdoc />
|
||||
public override bool CanWrite => Volatile.Read(ref disposed) == 0;
|
||||
|
||||
/// <summary>Not answerable: an object's length is not known until it has all been written.</summary>
|
||||
public override long Length => throw new NotSupportedException();
|
||||
|
||||
/// <inheritdoc cref="Length" />
|
||||
public override long Position
|
||||
{
|
||||
get => throw new NotSupportedException();
|
||||
set => throw new NotSupportedException();
|
||||
}
|
||||
|
||||
/// <inheritdoc />
|
||||
public override async ValueTask WriteAsync(
|
||||
ReadOnlyMemory<byte> buffer,
|
||||
CancellationToken cancellationToken = default)
|
||||
{
|
||||
var result = await pipe.Writer.WriteAsync(buffer, cancellationToken).ConfigureAwait(false);
|
||||
|
||||
if (result.IsCompleted)
|
||||
{
|
||||
// The reader has stopped, which means the upload ended — almost always because the service
|
||||
// refused it. Awaiting the task surfaces that exception here, at the write, instead of leaving
|
||||
// the caller to discover it at disposal after copying a whole file into nothing.
|
||||
await upload.ConfigureAwait(false);
|
||||
}
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Refused: this stream is asynchronous all the way down.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// Blocking on the pipe from a synchronous write is a deadlock waiting for a thread-pool starvation to
|
||||
/// find it — the other half of the pipe is being read by a task that needs a thread to run on. The only
|
||||
/// caller is the transfer queue, which copies asynchronously, so this is unreachable rather than
|
||||
/// inconvenient. Throwing says which; blocking would say nothing until a large upload hung.
|
||||
/// </remarks>
|
||||
public override void Write(byte[] buffer, int offset, int count) =>
|
||||
throw new NotSupportedException(
|
||||
"An upload to a bucket is written asynchronously; use WriteAsync.");
|
||||
|
||||
/// <summary>
|
||||
/// Nothing, deliberately.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// A flush cannot mean what a caller would want it to here — the object does not exist until the upload
|
||||
/// completes, so there is no partial state to make durable. The pipe's own writes are already handed to
|
||||
/// the reader as they arrive.
|
||||
/// </remarks>
|
||||
public override void Flush()
|
||||
{
|
||||
}
|
||||
|
||||
/// <inheritdoc cref="Flush" />
|
||||
public override Task FlushAsync(CancellationToken cancellationToken) => Task.CompletedTask;
|
||||
|
||||
/// <inheritdoc />
|
||||
public override int Read(byte[] buffer, int offset, int count) => throw new NotSupportedException();
|
||||
|
||||
/// <inheritdoc />
|
||||
public override long Seek(long offset, SeekOrigin origin) => throw new NotSupportedException();
|
||||
|
||||
/// <inheritdoc />
|
||||
public override void SetLength(long value) => throw new NotSupportedException();
|
||||
|
||||
/// <inheritdoc />
|
||||
public override async ValueTask DisposeAsync()
|
||||
{
|
||||
if (Interlocked.Exchange(ref disposed, 1) == 1)
|
||||
{
|
||||
return;
|
||||
}
|
||||
|
||||
// Completing the writer is what tells the upload there is no more, so it must happen before the
|
||||
// await — and it must happen even when the caller is abandoning a failed transfer, or the background
|
||||
// task never ends.
|
||||
await pipe.Writer.CompleteAsync().ConfigureAwait(false);
|
||||
|
||||
try
|
||||
{
|
||||
await upload.ConfigureAwait(false);
|
||||
}
|
||||
finally
|
||||
{
|
||||
await base.DisposeAsync().ConfigureAwait(false);
|
||||
}
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Refused when it would have to finish an upload.
|
||||
/// </summary>
|
||||
/// <remarks>
|
||||
/// <para>
|
||||
/// Completing this stream means completing the pipe and awaiting the upload, and doing that from a
|
||||
/// synchronous <c>Dispose</c> is the deadlock the synchronous <c>Write</c> above avoids. The alternative
|
||||
/// — completing the writer and abandoning the task — silently drops whatever the service was about to
|
||||
/// say, including a refusal, and reports a transfer as finished that never landed.
|
||||
/// </para>
|
||||
/// <para>
|
||||
/// So a <c>using</c> rather than an <c>await using</c> throws, which is loud, immediate and correct. The
|
||||
/// only caller already uses <c>await using</c>; this is what stops a second one being written by
|
||||
/// accident.
|
||||
/// </para>
|
||||
/// </remarks>
|
||||
protected override void Dispose(bool disposing)
|
||||
{
|
||||
if (disposing && Volatile.Read(ref disposed) == 0)
|
||||
{
|
||||
throw new NotSupportedException(
|
||||
"An upload to a bucket finishes asynchronously; use await using rather than using.");
|
||||
}
|
||||
|
||||
base.Dispose(disposing);
|
||||
}
|
||||
|
||||
private async Task UploadAsync(IAmazonS3 client, string bucket, string key)
|
||||
{
|
||||
using var transfer = new TransferUtility(client);
|
||||
|
||||
try
|
||||
{
|
||||
await transfer.UploadAsync(
|
||||
new TransferUtilityUploadRequest
|
||||
{
|
||||
BucketName = bucket,
|
||||
Key = key,
|
||||
InputStream = pipe.Reader.AsStream(),
|
||||
|
||||
// The stream has no length, so the utility has to be told not to look for one. It reads
|
||||
// until the pipe completes and splits what it read into parts.
|
||||
AutoCloseStream = false,
|
||||
},
|
||||
cancellationToken).ConfigureAwait(false);
|
||||
|
||||
await pipe.Reader.CompleteAsync().ConfigureAwait(false);
|
||||
}
|
||||
catch (Exception exception)
|
||||
{
|
||||
// Completing the reader *with* the exception is what unblocks a writer that is still copying:
|
||||
// its next write sees a completed pipe and awaits this task, which rethrows this.
|
||||
await pipe.Reader.CompleteAsync(exception).ConfigureAwait(false);
|
||||
throw;
|
||||
}
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,88 @@
|
||||
{
|
||||
"version": 2,
|
||||
"dependencies": {
|
||||
"net10.0": {
|
||||
"AWSSDK.Core": {
|
||||
"type": "Direct",
|
||||
"requested": "[4.0.100.9, )",
|
||||
"resolved": "4.0.100.9",
|
||||
"contentHash": "OPYy41jZjXwxxcYRotaq24HDrwUnVtBB/mvg1IwB9D1ICXAtHqMa1sp2hpmlVJCZwjlcrPcTCJIejInvV1vp5g=="
|
||||
},
|
||||
"AWSSDK.S3": {
|
||||
"type": "Direct",
|
||||
"requested": "[4.0.101.6, )",
|
||||
"resolved": "4.0.101.6",
|
||||
"contentHash": "LsVXGc3lyJuUJe+EbGubkFeR0cVmtmj4YdMChqsqSIsjZtCMzPg2BXR7cJqcrIBGoHab3q3RS6K8T9QD2tbhhQ==",
|
||||
"dependencies": {
|
||||
"AWSSDK.Core": "[4.0.100.9, 5.0.0)"
|
||||
}
|
||||
},
|
||||
"Meziantou.Analyzer": {
|
||||
"type": "Direct",
|
||||
"requested": "[3.0.137, )",
|
||||
"resolved": "3.0.137",
|
||||
"contentHash": "48n8WsokyefLOZt2qGPpk+hjfF9rQRa7MGK+8hMS6sO22QuIDtBztfbXV7RtVfX7JKjUwDI9e06I8aeCc5XZzQ=="
|
||||
},
|
||||
"Microsoft.CodeAnalysis.BannedApiAnalyzers": {
|
||||
"type": "Direct",
|
||||
"requested": "[5.6.0, )",
|
||||
"resolved": "5.6.0",
|
||||
"contentHash": "Kcobt3pnOdO0A+6CKiMHZdTEluJpsfxiV20axtZdmfBQnDmiWTKPJADlgAfdTuKNAnVarrkJa0UEGwuOo91muw=="
|
||||
},
|
||||
"Microsoft.Extensions.DependencyInjection.Abstractions": {
|
||||
"type": "Transitive",
|
||||
"resolved": "8.0.2",
|
||||
"contentHash": "3iE7UF7MQkCv1cxzCahz+Y/guQbTqieyxyaWKhrRO91itI9cOKO76OHeQDahqG4MmW5umr3CcCvGmK92lWNlbg=="
|
||||
},
|
||||
"Microsoft.Extensions.Logging.Abstractions": {
|
||||
"type": "Transitive",
|
||||
"resolved": "8.0.3",
|
||||
"contentHash": "dL0QGToTxggRLMYY4ZYX5AMwBb+byQBd/5dMiZE07Nv73o6I5Are3C7eQTh7K2+A4ct0PVISSr7TZANbiNb2yQ==",
|
||||
"dependencies": {
|
||||
"Microsoft.Extensions.DependencyInjection.Abstractions": "8.0.2"
|
||||
}
|
||||
},
|
||||
"dodossh.client.domain": {
|
||||
"type": "Project"
|
||||
},
|
||||
"dodossh.client.ssh": {
|
||||
"type": "Project",
|
||||
"dependencies": {
|
||||
"NSec.Cryptography": "[26.4.0, )",
|
||||
"SSH.NET": "[2025.1.0, )"
|
||||
}
|
||||
},
|
||||
"BouncyCastle.Cryptography": {
|
||||
"type": "CentralTransitive",
|
||||
"requested": "[2.6.2, )",
|
||||
"resolved": "2.6.2",
|
||||
"contentHash": "7oWOcvnntmMKNzDLsdxAYqApt+AjpRpP2CShjMfIa3umZ42UQMvH0tl1qAliYPNYO6vTdcGMqnRrCPmsfzTI1w=="
|
||||
},
|
||||
"libsodium": {
|
||||
"type": "CentralTransitive",
|
||||
"requested": "[1.0.22, )",
|
||||
"resolved": "1.0.22",
|
||||
"contentHash": "KPD9SloJFclrsjnhABu7dzWrcyYkwPbvx5l1gRSPAX/0n+OBtSiVCKtGFv4n+ecWUHU0tCG9LSSwoZZx673zBQ=="
|
||||
},
|
||||
"NSec.Cryptography": {
|
||||
"type": "CentralTransitive",
|
||||
"requested": "[26.4.0, )",
|
||||
"resolved": "26.4.0",
|
||||
"contentHash": "0vsCtY5f+YgQROiWNqzgWp+l2pddfk9FkWoGV/bEo0MuEYPKlJWuoA8aOfO6qp3f+EnObKE3zSJhn1PspJeJVg==",
|
||||
"dependencies": {
|
||||
"libsodium": "[1.0.22, 1.0.23)"
|
||||
}
|
||||
},
|
||||
"SSH.NET": {
|
||||
"type": "CentralTransitive",
|
||||
"requested": "[2025.1.0, )",
|
||||
"resolved": "2025.1.0",
|
||||
"contentHash": "jrnbtf0ItVaXAe6jE8X/kSLa6uC+0C+7W1vepcnRQB/rD88qy4IxG7Lf1FIbWmkoc4iVXv0pKrz+Wc6J4ngmHw==",
|
||||
"dependencies": {
|
||||
"BouncyCastle.Cryptography": "2.6.2",
|
||||
"Microsoft.Extensions.Logging.Abstractions": "8.0.3"
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user