The gate restores .test-inputs/photocraft/corpus before its setup, keyed on
scripts/ci-test-inputs.sh (which holds the PhotoCraft pin), and the setup uses a
restored corpus as it is. A hit skips the PhotoCraft checkout and its xtask
build: 2:17 of Windows setup, about 1:00 on Linux and macOS. Uses
craft-common-actions' new cache-paths, cache-key and cache-key-files inputs.
The Workspace gate already runs fmt, both clippy runs and the tests of every
crate on Linux, Windows and macOS. The per-crate workflows drop those steps and
keep their own checks: wasm32 builds and tests, no-default and single-feature
clippy and tests, cross-target clippy, docs, release and sample tests,
--ignored tests, ffmpeg oracle tests, the Language-data proofing builds, the PDF
Unicode fixtures, read-only contract and vendored backends. All of them run on
Linux. docpath.yml, print.yml, i18n.yml and ui.yml had no check of their own and
are removed.
They also stop caching dependencies: about 25 rust-cache entries no longer
compete with the gate's for the repository's 10 GB, which the gate's main run
missed on every OS. The data caches (Opus vectors, CRW samples) stay.
CARGO_PROFILE_DEV_DEBUG=0 on Windows, as PhotoCraft's Windows legs do: writing
PDBs for every test binary is a large part of the Windows build. The gate
writes windows-env after its dev-debug default, so the Linux and macOS legs keep
line-tables-only.
Every Workspace leg runs nextest with --all-features and the doctests with
cargo test --doc. The default-feature test run is gone: it added exactly four
tests, all in craft-image-codecs and craft-text (feature-off behaviour), so
those two crates run their tests with default features as well. The
default-feature compile stays covered by the first clippy run.
The resource groups from craft-libs#2058's audit move into the ci profile with
platform = 'cfg(unix)', so Linux and macOS share one profile and Windows keeps
the groups it already passes with. No separate example build: nextest builds
examples as cargo test does, and clippy --all-targets checks them.
scripts/test-workspace-tests.py follows the single command set.
craft-testkit's hostile hammer reads CRAFT_HOSTILE_DIVISOR=N: the unmodified
seed, every N-th truncation and 1/N of the rounds. hostile::divisor() gives the
same value to sweeps written without the hammer; craft-image-codecs' TIFF zoo
and malformed-header sweeps use it, with a different offset per input so the
inputs together still reach every position. Proptests without their own case
count read proptest's PROPTEST_CASES.
The Workspace gate and the media, audio, video and PDF workflows set
CRAFT_HOSTILE_DIVISOR=8 and PROPTEST_CASES=64. The longest of these tests
(craft-media's MXF hostile input, 1,310 s on Windows; craft-pdf's crypt
proptests, up to 257 s; the image-codecs sweeps, 140-162 s) drop to about an
eighth or a quarter. nightly.yml runs the whole workspace on Linux without
either variable, so every case still runs once a day.
The gate's test-input setup moves to scripts/ci-test-inputs.sh, shared by both
workflows; it uses a corpus that is already present.
* har: preserve source reader history
Port Craft Anything crates/formats/src/har.rs from f64ae3bf3c8863a982af1304e06de20653803968.
* har: preserve source reader history
Port Craft Anything crates/formats/src/har.rs from cb0bb9d8fa4536ba44ef2c86f5ed759b879cc2d3.
* Expose shared HAR reader with explicit input admission
---------
Co-authored-by: Brandon Thomas <echelon@gmail.com>
Co-authored-by: BFlat <bombayapirat@gmail.com>
* tar: port old GNU sparse extension traversal from Craft Anything
Source: storytold/craft-anything#1973, 8ea2ec66a1bc80f0420f3695baac877a0cb9b9a5. Retains non-extractable sparse member kind and regression tests.
* tar: preserve GNU CRC symlink checksum compatibility
Source: storytold/craft-anything#1984, b226a475f7dce9710e3af7a069386fe7a9720df7; preserves nonzero checks and regular-file checksum validation.
* tar: apply valid PAX sizes before legacy member size fields
Source: storytold/craft-anything#2007, d29465c20b0d570ea4973be512a0cb4e212bb60f; metadata records still retain their own framing.
* Record provenance of current TAR and CPIO compatibility fixes
---------
Co-authored-by: RRFZ <247457381+Plastic-time@users.noreply.github.com>
craft-audio names its VMD decoder feature `vmd-decode`, and craft-video
names a codec's decoder `<codec>-decode` wherever the codec may also have
an encoder (mpeg2, mpeg4, vp8…). The video decoder now follows that:
`vmd-decode` gates the module and its VideoFrame conversion, and `vmd`
turns it on, so `vmd` keeps working and an encoder can be added later
without renaming.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
benandez's re-review reproduced three escapes the first confinement fix still
left open. A dangling symlink: canonicalize reported it as missing, so the
old confine_to_root walked up to root and accepted it, and write created the
symlink's target for real, outside root. A check-then-use race: confine_to_root
checked one path string and a later fs::read or fs::write resolved that same
string again, a second lookup a concurrently swapped link could land in
between. A hard link: write truncated the shared inode in place, with nothing
in a canonicalized path to show that another name pointed at the same file
outside root.
list, read, write and create now resolve every already segment-checked
relative path through a cap_std::fs::Dir capability opened on root, instead
of joining an absolute OS path and handing it to plain std::fs. cap_std
resolves one path component at a time against real, already-open directory
handles, so a symlink or a Windows junction cannot walk resolution above
root, dangling or not, and there is no separate check with a window for the
directory tree to change underneath it: the one resolution is the open. This
is the same family of guarantee craft-font-store already depends on through
cap-primitives, which cap_std is built on.
write no longer truncates its target in place either way: it writes a new
sibling file and renames it over the target, so a hard link's directory
entry is replaced rather than its shared content being overwritten.
Per that review's non-blocking notes: a name that already exists on disk
with an ordinary ':' or '\' byte (macOS stores a typed '/' as ':') stays
reachable by list and read, since neither character is a path separator on
any platform's filesystem; the stricter refusal now applies only to a name
being chosen by Create. Create also goes back to bringing a fresh root into
existence, matching its behaviour before the first confinement fix. A
Windows reserved device name (CON, COM1, CONIN$, with or without an
extension) is refused in a chosen name on every platform.
Regression tests reproduce all three escapes (dangling symlink, swapped-link
race, hard link), plus the previously missing coverage: a link in the folder
a Create targets, and a write test that actually reaches disk instead of
being refused for an unrelated reason before confinement ever ran. Verified
by hand that each of the three new write-path tests fails against the
pre-fix pattern (an absolute OS path joined by string concatenation) before
trusting them.