Reinstate ragel_subproject=true so CI continues exercising the bundled
Ragel build and generated-source regeneration. This was the only CI job
covering that path, as noted in the PR discussion.
Keep the configs depend-mode fix, which handles Ragel line-marker
pseudo-files independently of whether the generator is rebuilt.
Validation: actionlint, git diff --check, and meson test -C build
(302 passed, one skipped).
Assisted-by: OpenAI Codex
Restore compiler caches through a shared setup action and save them
explicitly only when compilation reports cache misses. Keep existing
restore prefixes and make cache-save failures nonfatal. Before saving
ccache, prune entries that were not used during the job.
This avoids duplicate snapshots on fully cached runs and removes
unrelated restored objects from changed snapshots. Key the MSVC
installed prefix by its dependency build inputs rather than unrelated
workflow edits; dependency options remain in the hashed helper project.
Validation: actual ccache 4.9.1 cold/warm and pruning checks; sccache
statistics and failure-path checks; nine workflow setup/save pairs;
actionlint, git diff --check, and meson test -C build
(302 passed, one skipped). Remote uploads remain for CI.
Assisted-by: OpenAI Codex
Key Cargo snapshots by compiler, build configuration, Rust sources,
and public headers instead of run ID. An exact cache hit can now skip
the upload rather than storing another copy on each successful run.
Retain the existing restore prefix for incremental rebuilds.
Validation: actionlint, git diff --check, and meson test -C build
(302 passed, one skipped). Remote snapshot reuse remains for CI.
Assisted-by: OpenAI Codex
ccache 4.9 treats Ragel line-marker filenames as inputs and fails to
stat paths such as hb-ot-shaper-indic-machine.rl and the pseudo-file
NONE when compiling the amalgamations. Compilation succeeds through
fallback, but every cache request fails with bad_input_file.
Enable depend mode and emit full compiler dependencies with -MD,
avoiding those preprocessor line-marker paths.
Validation: ccache 4.9.1 with GCC 13.5 builds all six configurations;
the repeat build has 12 direct hits and zero errors. All executables
run, and concurrency/failure checks pass. Actionlint, diff checks, and
meson test -C build pass (302 passed, one skipped).
Assisted-by: OpenAI Codex
Use the checked-in generated headers when Ragel is unavailable instead
of building the generator subproject for every clean CI checkout.
Leave normal system-Ragel discovery and developer regeneration intact.
Validation: actionlint, git diff --check, and meson test -C build
(302 passed, one skipped). The latest runner log confirms no system
Ragel is installed.
Assisted-by: OpenAI Codex
Use Meson optimization=1 (/O1) for x86 while retaining optimization=3
(/O2 /Gw) for amd64. Keep the two existing jobs and their tests, and set
b_ndebug=true explicitly to preserve the previous release assertion
setting when the x86 build type becomes custom.
This is a CI compile-time experiment; dependency-prefix builds remain
release builds and the amd64 job retains optimized MSVC coverage.
Validation: actionlint, option/matrix checks, and meson test -C build
(302 passed, one skipped). Native compiler timing remains for CI.
Assisted-by: OpenAI Codex
Build GLib, FreeType, PNG, and zlib in a small Meson dependency project
and cache the installed prefix. Key it by architecture, compiler, SDK,
Python version, wrap definitions, and CI build configuration.
Use pkg-config and nofallback for HarfBuzz so cached dependencies avoid
repeated fallback configuration and linking. Keep generated Meson state
out of the cache and retain GLib API and native Windows backend tests.
Validation: actionlint, git diff --check, and meson test -C build
(302 passed, one skipped). Pinned fallback dependencies built and installed
on Linux. A relocated-prefix nofallback consumer compiled, linked,
and passed its runtime smoke test. Native Windows behavior remains
for CI.
Assisted-by: OpenAI Codex
Run two configurations concurrently within the existing runner. Use
separate outputs and logs per configuration, then print grouped logs
and collect failures after waiting for each pair. Retain all six core
link checks and subset compilations without adding matrix jobs.
Validation: actionlint, git diff --check, and meson test -C build
(302 passed, one skipped). All six configurations compile and link
with the real compiler. A stub-compiler check confirms the two-worker
limit and failure propagation.
Assisted-by: OpenAI Codex
The forcefallback wrap mode also builds ICU for its default auto feature.
Disable it explicitly to avoid building the large optional dependency.
Keep GLib and the native Windows backends so API and DirectWrite tests
remain enabled. Linux and MSYS2 retain ICU integration coverage.
Validation: actionlint, git diff --check, and meson test -C build
(302 passed, one skipped). Native MSVC validation remains in CI.
Assisted-by: OpenAI Codex
Use ccache for each configuration's core and subset compilation. Split
core compilation from linking so ccache can store its object while
retaining the existing link check and accumulated failure status.
Validation: actionlint, git diff --check, and meson test -C build
(302 passed, one skipped). The HB_TINY core/subset commands compile
and link successfully and produce cache hits on a repeat build.
Assisted-by: OpenAI Codex
The Valgrind job leaves Graphite support disabled, so its development
package is unused. Remove it without changing the build or test options.
Validation: actionlint and git diff --check. Tests were not rerun for
this CI-only package-list change.
Assisted-by: OpenAI Codex
Cache downloaded Cargo sources and Meson's Rust debugoptimized target
outputs, including the rebuilt standard library. Scope the cache to the
compiler version and build configuration, and save updated entries for
subsequent runs without caching generated bindings or Meson state.
Retain the Cargo invocation so it checks source freshness after restore.
Validation: actionlint, git diff --check, and meson test -C build.
Remote cache restoration remains to be exercised in Actions.
Assisted-by: OpenAI Codex
Disable documentation setup and GIR/typelib generation, and remove their
build dependencies. The docs job covers these outputs. Retain the full
Rust integration test suite, Clippy, and rustfmt.
Validation: actionlint, git diff --check, and meson test -C build.
Assisted-by: OpenAI Codex
Set fuzz-seconds to 300 instead of 600 to shorten pull-request feedback.
Validation: actionlint and git diff --check. Tests were not rerun for
this CI-only duration change.
Assisted-by: OpenAI Codex
Group pull-request runs by workflow and PR number so new commits cancel
older queued or running checks. Give non-PR runs unique groups, leaving
main-branch builds and tag releases independent.
Validation: actionlint; workflow concurrency assertions;
meson test -C build (302 passed, one skipped).
Assisted-by: OpenAI Codex
Linux and Rust shared a cache key, while Valgrind used the unscoped
default. Include the workflow name in cache keys consistently and give
Linux and Rust 1G each, since their 500MB caches were full in PR #6637.
Validation: actionlint; meson test -C build
(302 passed, one skipped).
Assisted-by: OpenAI Codex
Setting CC and CXX prevents Meson from automatically enabling ccache.
Wrap both compilers explicitly and isolate each sanitizer's cache.
Validation: actionlint; standalone Meson ASan/UBSan C/C++ build and
cache hits on rebuild; meson test -C build
(302 passed, one skipped).
Assisted-by: OpenAI Codex
Install the native ccache package for each MSYSTEM and persist its
workspace cache separately for MINGW64 and CLANG64. Check compiler
contents across package updates and report compiler-cache statistics.
Validation: actionlint; YAML assertions; meson test -C build
(302 passed, one skipped). Native MSYS2 validation remains in CI.
Assisted-by: OpenAI Codex
Disable Cairo and dependency tests for GLib and libffi. Keep GLib so
HarfBuzz's API tests, including DirectWrite, still build and run under
MSVC. MSYS2 continues to cover the Cairo integration on Windows.
Validation: actionlint; meson test -C build
(302 passed, one skipped). Native MSVC validation remains in CI.
Assisted-by: OpenAI Codex
Meson does not automatically prepend ccache to compilers selected in a
cross file. Generate a CI-only overlay with explicit compiler wrappers
and isolate the cache by target architecture. Keep the regular cross
files usable without ccache installed.
Validation: actionlint; C/C++ Meson cross-builds for i686 and x86_64,
including cache hits on rebuild; meson test -C build
(302 passed, one skipped).
Assisted-by: OpenAI Codex
Run the six configuration variants in one job and the four C++
compiler/standard variants in another. This removes eight runner
requests per push or pull request while retaining all compile checks.
Group logs by variant and accumulate failures so later checks still run.
Validated all 16 compile/link checks locally, workflow Bash syntax,
compiler-failure propagation, and actionlint with the existing
ubuntu-26.04 runner label allowed.
Fixes: https://github.com/harfbuzz/harfbuzz/issues/6649
Assisted-by: OpenAI Codex
GitHub's Ubuntu runner carries Chrome and Microsoft apt sources, but every
package installed by these workflows comes from Ubuntu. A third-party
repository publishing while apt refreshes its indexes can therefore fail
otherwise unrelated jobs with a hash-sum mismatch.
Remove both unused sources before updating package indexes in every affected
workflow.
Assisted-by: OpenAI Codex <openai-codex@users.noreply.github.com>
* [cmake] Compile core sources individually
CMake built libharfbuzz from the amalgamated harfbuzz.cc source,
placing all core symbols in one archive member. Static consumers
linking an auxiliary library before libharfbuzz then pulled duplicate
internal definitions into the final link.
Derive the core source list from harfbuzz.cc and compile each source
separately, matching the Meson archive layout.
Fixes: https://github.com/harfbuzz/harfbuzz/issues/5683
Assisted-by: OpenAI Codex <codex@openai.com>
* [ci] Make CMake static build explicit
Set BUILD_SHARED_LIBS=OFF on the ARM CMake job so changes to static
archive construction have intentional CI coverage.
Assisted-by: OpenAI Codex <codex@openai.com>
The configs job only compiled harfbuzz.cc, now it compiles
harfbuzz-subset.cc too to catch config breakage in subset code.
Assisted-by: Claude Opus 5 <noreply@anthropic.com>
* [test/fuzzing] Skip hb-depend-closure-parity without GLib
It depends on GLib unlike the other fuzzers.
* [ci] Build the msan bot without external dependencies
We don’t care about msan errors in code that is not ours, and some
libraries like GLib are not built with memory sanitizers in mind.
* [ci] Don't pipe sanitizer test output
The pipe was masking meson exit status, and it was redundant anyway
since the sanitizer output is is already symbolized and demangled.
* [COLR] Fix integer overflow in get_extents
Use to_glyph_extents(), which guards against this.
Update test expectation since we now round instead of truncate.