commit 18fa445ffcad3abccb86136938965b2765c63c62
Author: Nick Craig-Wood <nick@craig-wood.com>
Date:   Mon Aug 3 17:45:42 2026 +0100

    compress: fix corrupted objects being created when the source ends early
    
    When the source supplied fewer bytes than its declared size, the
    compressed data file was stored under a name containing the declared
    size while the metadata recorded the actual number of bytes read.
    NewObject looks the data file up by the size in the metadata, so the
    resulting object could never be read again, and the upload reported
    success.
    
    Check that the number of bytes read matches the declared size after
    uploading the data and before writing the metadata, and remove the
    data file and return an error if it does not.
    
    This was found by the new FsPutShortEOF integration test.

commit e0701daea005c9bf65e573ad416c711d23d3a2b1
Author: Nick Craig-Wood <nick@craig-wood.com>
Date:   Sun Aug 2 18:25:46 2026 +0100

    dropbox: don't retry chunked upload requests when the upload has been cancelled
    
    The append loop retries everything once the upload session has
    started, so a cancelled context error was retried through all the low
    level retries with exponential backoff before the upload gave up.

commit bff17664ad2dc68956e4ed1099a754b0573c088f
Author: Nick Craig-Wood <nick@craig-wood.com>
Date:   Sun Aug 2 18:25:32 2026 +0100

    dropbox: fix chunked uploads of truncated files never finishing - fixes #9704
    
    A source which returned EOF before supplying as many bytes as it
    declared would either commit a truncated file (if the shortfall was
    within the final chunk) or loop forever appending empty chunks to the
    upload session. Return an error wrapping io.ErrUnexpectedEOF instead.
    
    Note that all dropbox uploads use the chunked upload path with the
    default batch_mode of sync, so this affected uploads of every size.

commit 03f8582a55f5ab090279b99a3cf552779744b263
Author: Nick Craig-Wood <nick@craig-wood.com>
Date:   Sun Aug 2 12:53:57 2026 +0100

    fstests: check uploads terminate when the source ends before its declared size
    
    A source reader which returns EOF before supplying as many bytes as it
    declared should not cause a backend Put to hang, panic or create an
    object claiming the declared size.
    
    The layers above detect a truncated transfer by comparing sizes after
    the upload, so a backend may either return an error or create a
    shorter object.
    
    See #9704

commit 5024425ba8bde8cf7f9ab152abc0d7979fd3650b
Author: Nick Craig-Wood <nick@craig-wood.com>
Date:   Mon Aug 3 20:57:53 2026 +0100

    Add alliasgher to contributors

commit ea9a64c751839f59f5b2be8f83080c6751529b61
Author: alliasgher <alliasgher123@gmail.com>
Date:   Sun Aug 2 22:50:21 2026 +0500

    smb: reuse the upload connection for SetModTime - fixes #9675
    
    Object.Update held its connection until the deferred putConnection ran at
    function exit, so the SetModTime it does at the end of every upload had to take
    a second connection from the pool, dialling a whole new SMB session when the
    pool was empty. With N transfers in flight the pool grew to roughly 2N sessions
    for no reason.
    
    Return the connection as soon as the file is closed. At that point the upload
    has succeeded and remove() can no longer be reached, so nothing else needs it,
    and SetModTime picks the same connection straight back out of the pool.
    putConnection nils the pointer, so the deferred putConnection becomes a no-op
    and the connection is not returned twice.

commit 065af8963550268a9ad426528638299e06333465
Author: alliasgher <alliasgher123@gmail.com>
Date:   Sun Aug 2 18:46:48 2026 +0500

    drive: don't list trashed files when removing a directory into the trash - fixes #9681
    
    purgeCheck listed a directory's children with includeAll set, so the Drive API
    returned trashed children as well as live ones. The only thing that came of
    that was the trashedFiles flag, which is used solely to decide whether to trash
    the directory or delete it permanently, and when use_trash is on the directory
    gets trashed either way.
    
    That made the common case pay for nothing: after trashing N files in a
    directory, removing the now empty directory paged through all N trashed entries,
    about one API call and pacer delay per 1000, instead of running a single query
    that returns nothing.
    
    Ask the server to filter out trashed children when the directory is going to be
    trashed anyway. Hard deletes still enumerate them, since there the distinction
    decides whether the directory can be removed permanently (#1040), and
    --drive-trashed-only is left alone because it deliberately lists trashed items.

commit 3f9d583cffd94105f3ae2cde2cc6c0d5a2c9d631
Author: Nick Craig-Wood <nick@craig-wood.com>
Date:   Sat Aug 1 19:42:58 2026 +0100

    gitannex: reduce subprocess launches in end to end test setup
    
    Write the git configuration for each test's fake home directory as a
    file instead of running three "git config --global" commands, and drop
    the "git annex version" invocation from repository setup. This removes
    four subprocess launches from each of the fifteen test cases.

commit b6a82ee56c4379695a35e1136ac0c6d25578bc2e
Author: Nick Craig-Wood <nick@craig-wood.com>
Date:   Sat Aug 1 19:42:17 2026 +0100

    gitannex: run the end to end test functions in parallel
    
    The subtests within each end to end test function already run in
    parallel, but the three test functions themselves ran one after
    another.
    
    Marking the functions parallel lets all their subtests overlap.

commit c9cd065458e81c9165a708155194ee90130812c9
Author: Nick Craig-Wood <nick@craig-wood.com>
Date:   Sat Aug 1 19:41:38 2026 +0100

    gitannex: skip end to end tests under the race detector
    
    The end to end tests exercise rclone via a separate subprocess spawned
    by git-annex, and that subprocess is not built with race
    instrumentation.
    
    Running them in the race test therefore adds several minutes to CI
    without providing any race coverage. The unit tests in this package
    still run under the race detector.

commit b3168b3ad6ef4dc24ffd2970a8bde8aae25c6946
Author: Nick Craig-Wood <nick@craig-wood.com>
Date:   Sat Aug 1 19:41:00 2026 +0100

    gitannex: speed up the end to end tests with testremote --fast
    
    Run the full "git annex testremote" suite for a single layout mode and
    use "testremote --fast" for the rest.
    
    The full suite repeats the same protocol operations across a matrix of
    key sizes and chunk configurations, which exercises client side
    git-annex behaviour rather than rclone.

commit c4e82700cb9fef16ae1b41354fd2c42abb7dbe76
Author: Nick Craig-Wood <nick@craig-wood.com>
Date:   Sat Aug 1 19:40:23 2026 +0100

    gitannex: speed up the end to end tests by removing redundant checks
    
    The migration test ran the full "git annex testremote" suite for every
    layout mode after verifying the migration with "git annex fsck".
    
    The fsck calls already prove the migrated data is accessible via the
    builtin special remote, and TestEndToEnd covers the special remote
    protocol with testremote, so the extra five full testremote runs
    duplicated coverage at a cost of tens of seconds each on CI.

commit 01d32dbde5b429970687d0254e514eecc3d9cec7
Author: Nick Craig-Wood <nick@craig-wood.com>
Date:   Sat Aug 1 13:45:45 2026 +0100

    build: fix intermittent CI failures by raising the test timeout to 20m
    
    The cmd/gitannex end to end tests routinely take 230-530s and on slow
    CI runners the package exceeds the go test default timeout of 10m.
    
    Raise the per-package timeout to 20m in quicktest and racequicktest to
    give slow runners headroom.

commit c59083f3b45058ef82cbecb1a14548515baff457
Author: Nick Craig-Wood <nick@craig-wood.com>
Date:   Sun Aug 2 18:25:01 2026 +0100

    Add debaditya to contributors

commit a15551a838ae7ca41c0685cdcb43a53d51f329de
Author: Nick Craig-Wood <nick@craig-wood.com>
Date:   Sun Aug 2 18:25:01 2026 +0100

    Add Loi Nguyen to contributors

commit 40dbc9ba229d456b46a07ad09bd518aca01c2ba0
Author: Nick Craig-Wood <nick@craig-wood.com>
Date:   Sun Aug 2 18:25:01 2026 +0100

    Add am-at-enrollvb to contributors

commit ce8caf0daa1110621fa5543d7e025b03bae43446
Author: Nick Craig-Wood <nick@craig-wood.com>
Date:   Sun Aug 2 18:25:01 2026 +0100

    Add SillyZir to contributors

commit 49e41e7f0be70438827c7453f08e061545c23d7c
Author: Nick Craig-Wood <nick@craig-wood.com>
Date:   Sun Aug 2 18:25:01 2026 +0100

    Add acoeur to contributors

commit 23f1dbfe503de46d4fe92eb9eccf6c735e480f59
Author: Loi Nguyen <vanloi999@gmail.com>
Date:   Sun Aug 2 17:53:25 2026 +0700

    smb: test closing failed dial connections
    
    Followup to:
    
    74f9f182aa0b3050 smb: fix TCP connection leak when connection setup fails - #9678

commit c97ab80ca1c803db47bda45d1b02736a44010245
Author: debaditya <82204129+DebadityaHait@users.noreply.github.com>
Date:   Sun Aug 2 16:16:27 2026 +0530

    dropbox: fix context cancellation (e.g. --max-duration limit) not stopping in-flight requests
    
    This switches the calls to the dropbox SDK over to use the context taking
    variants. This allows context cancellation to propagate properly.
    
    Fixes #9688

commit 1d03a771727b6b158d4675442c41d3280d9eee88
Author: Loi Nguyen <vanloi999@gmail.com>
Date:   Fri Jul 31 13:58:04 2026 +0700

    vfs: make poll shutdown lifecycle deterministic

commit d5275c4ebb82ef7ffe8406c2049f29c78a8b6446
Author: Loi Nguyen <vanloi999@gmail.com>
Date:   Fri Jul 31 12:39:05 2026 +0700

    vfs: test poll interval update lifecycle

commit f132aef2958f21d127d614db8ea150d33eef4412
Author: Loi Nguyen <vanloi999@gmail.com>
Date:   Fri Jul 31 12:38:22 2026 +0700

    vfs: synchronize poll updates with shutdown - fixes #9689

commit f4e231b094f9bb26b2d8cc3c081c8601ab4fa8fd
Author: Loi Nguyen <vanloi999@gmail.com>
Date:   Sat Aug 1 18:36:59 2026 +0700

    drive: fetch metadata permissions concurrently - fixes #9682
    
    Before this change permission metadata fetches were serialized by the permission cache mutex.

commit 8c8062da782e7cb5e6dca977736350aee6b5a7be
Author: Loi Nguyen <vanloi999@gmail.com>
Date:   Sat Aug 1 18:33:32 2026 +0700

    dropbox: preserve Paper export paths on lookup - fixes #9691
    
    Direct lookups of exported Dropbox Paper files retained the
    caller-visible extension before export metadata processing appended it
    again. Track when metadata was resolved through an export path so the
    object keeps the requested remote name while listing behavior remains
    unchanged.

commit 7f6207fae2386d747347199936d56a105cbdf557
Author: Loi Nguyen <vanloi999@gmail.com>
Date:   Fri Jul 31 00:06:00 2026 +0700

    lib/batcher: prevent commits racing shutdown - fixes #9687

commit 5dd34275dcbb485de2f6f7ce5d52242237c191f3
Author: am-at-enrollvb <arthur.melton@enrollvb.com>
Date:   Sat Aug 1 11:25:06 2026 +0000

    serve: pass the client IP address to the auth proxy - fixes #4499
    
    The auth proxy was only given the user and their password or public
    key, so a proxy program had no way to restrict logins to particular
    networks, or to record where an authentication attempt came from.
    
    The JSON sent to the program now has a client_ip key holding the bare
    IP the client connected from, with the port stripped so IPv6 arrives
    as 2001:db8::1 rather than [2001:db8::1]:52344. An IPv4-mapped IPv6
    address is reported as plain IPv4 so that a client arriving over a
    dual-stack listener still matches IPv4 networks. The key is omitted
    when the client has no IP address.
    
    The IP is also mixed into the backend cache key. That is needed as the
    program is only run on a cache miss, so a client from a
    non-allowlisted address presenting valid credentials within the 5
    minute cache lifetime would get a cache hit and be let in without the
    program being consulted at all.

commit 7804c1b315a5e74378611aada900c791a9fa7cd8
Author: SillyZir <269283839+SillyZir@users.noreply.github.com>
Date:   Mon Jul 20 22:49:57 2026 -0400

    serve nfs: fix EIO when creating symlinks with --vfs-links
    
    The macOS NFS client sends SETATTR after SYMLINK, which arrives as
    Chmod/Chown on the link path. These opened the target with vfs.Open,
    which follows symlinks - a freshly created symlink usually dangles, so
    the open failed with ENOENT, surfaced to the client as NFS3ERR_IO even
    though the link was created.
    
    Add path-based VFS.Chmod and VFS.Chown mirroring VFS.Chtimes. They do
    not follow symlinks (lstat semantics, matching VFS.Stat) and return
    ENOSYS when the node exists, since the VFS stores neither permissions
    nor ownership; serve nfs calls them and masks ENOSYS as before.
    
    Fixes #9627

commit 060b9975951d1d9dfa9f576384934d46b02d99f6
Author: acoeur <7315110+forestrock@users.noreply.github.com>
Date:   Fri Jul 31 19:41:46 2026 -0400

    azureblob: enable on Solaris
    
    The azure SDK didn't used to compile on Solaris, but now it does.

commit 39d8e83a12c0d265ceee981cd4495cf6ac4890f5
Author: Nick Craig-Wood <nick@craig-wood.com>
Date:   Fri Jul 31 18:21:36 2026 +0100

    Start v1.76.0-DEV development
