You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit b8405ed
Browse filesBrowse the repository at this point in the historyBrowse files
Every workflow now pins the newest runner images: ubuntu-26.04 and
ubuntu-26.04-arm replace ubuntu-24.04 and ubuntu-24.04-arm, macos-26 and
macos-26-intel replace macos-15 and macos-15-intel, and the stale and
branch cleanup jobs name ubuntu-26.04 instead of ubuntu-latest. Windows
stays on windows-2025. The job that builds with the oldest supported OCaml
(5.0.0) stays on ubuntu-24.04, because OCaml 5.0's runtime does not compile
with GCC 15, which defaults to C23.
The new images needed three CI and test fixes. The APT package lookup
reports apt's output and fetches the package lists once before giving up.
The watch-exit lock holder in rewatch-ocaml/tests/run.sh is a copied bash
instead of a copied sleep, which uutils coreutils (Ubuntu 26.04) refuses
to run under the name rescript. The --ppx-wait helper of the rewatch unit
tests publishes its PID with a rename, so a test that polls for the marker
never reads an empty file.
Review: Is there a newer Windows runner? No. windows-2025 is the newest
Windows Server image; windows-2025-vs2026 is the same image with Visual
Studio 2026, which the MinGW build does not use.
Review: Why did the APT lookup fail? On the first ubuntu-26.04 run,
apt-cache policy exited with status 100 and no output in all four Linux
jobs; the rerun on the same image (20261012.163.1) resolved every package
on the first try. The cause is unknown: apt 3.0.3 does not reproduce it
with or without package lists, nor when awk closes the pipe early. The
lookup now prints apt's output if it recurs.
Review: Does the lock holder still count as a rescript process? Yes. The
owner check reads the basename of /proc/<pid>/exe on Linux and of
ps -o comm= on macOS; both give rescript for the copied bash, which runs
its loop in its own process instead of exec'ing sleep.
Review: Why did escaped_exception fail on macos-26-intel? The test polls
for the marker file and then reads the PID from it. The PPX created the
file before writing to it, so a poll in between read an empty string
(Failure "int_of_string"). This is a race in the test helper, not a
macOS 26 change.
Review: Do the APT and OPAM caches need new keys? No. The APT key hashes
/etc/os-release and the OPAM key includes the runner label, so both
rebuild on the new images.
Area: ci rewatch tests
Output: unchanged
Tests: ~rewatch-ocaml/tests/run.sh ~tests/rewatch_ounit_tests/unit_tests.ml
PR: #8891
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014rR6jNpVHh7sx7L62iEf4u
Signed-off-by: Christoph Knittel <ck@cca.io>
0 commit comments