Summary
#201 gave the t9 range a CI job. The 30 t0xxx files are still uncovered: nothing runs them, for the same reason t9 was invisible — ci.yml's select skips its jobs for a t/-only diff, and the only task that touches t/ is test-subdir, which now filters on t9.
I ran them to find out whether that is worth fixing, and all 30 pass, so there is no cleanup blocking it:
FRESH=$(mktemp -d)
env -i PATH="$PATH" HOME="$FRESH" bash -c '
git config --global user.name "Bit Test"
git config --global user.email "test@example.com"
bash t/run-tests.sh t0'
========================================
Total: 30
Passed: 30
Failed: 0
========================================
Run under env -i with an empty HOME and only the two config lines t9-suite.yml sets, so this reflects a stock runner rather than a developer machine.
Why it is worth doing
t9 went uncovered long enough for t9014 and t90010 to stay red for months (#191) and for a t9007 assertion to never pass at all since #139 (#196). Nothing structural stops the same thing happening to t0xxx; they pass today only because nobody has broken them yet.
Suggested approach
Widen the test-subdir task's filter from t9 to cover both ranges, and let .github/workflows/t9-suite.yml pick them up — renaming the job, since it would no longer be t9-specific.
Two things to weigh:
- Job duration. The
t9 job takes ~14 minutes in CI, most of it the release build that deps { build } triggers. t0xxx is 30 files against t9's 21, so one combined job is the simpler shape — the build is paid once either way — but it is worth checking the total stays comfortable before committing to it.
t0011, t0012, t0013, t0015 are randomized (t0011-random-ops.sh and friends). They pass on their seeds today; if they turn out to be flaky under CI's timing, they may want separating from the deterministic ones rather than destabilising the whole job.
Related: #202 (the harness supplies no git identity) would let the workflow drop its Configure git identity step either way.
Summary
#201 gave the
t9range a CI job. The 30t0xxxfiles are still uncovered: nothing runs them, for the same reasont9was invisible —ci.yml'sselectskips its jobs for at/-only diff, and the only task that touchest/istest-subdir, which now filters ont9.I ran them to find out whether that is worth fixing, and all 30 pass, so there is no cleanup blocking it:
Run under
env -iwith an emptyHOMEand only the two config linest9-suite.ymlsets, so this reflects a stock runner rather than a developer machine.Why it is worth doing
t9went uncovered long enough fort9014andt90010to stay red for months (#191) and for at9007assertion to never pass at all since #139 (#196). Nothing structural stops the same thing happening tot0xxx; they pass today only because nobody has broken them yet.Suggested approach
Widen the
test-subdirtask's filter fromt9to cover both ranges, and let.github/workflows/t9-suite.ymlpick them up — renaming the job, since it would no longer be t9-specific.Two things to weigh:
t9job takes ~14 minutes in CI, most of it the release build thatdeps { build }triggers.t0xxxis 30 files against t9's 21, so one combined job is the simpler shape — the build is paid once either way — but it is worth checking the total stays comfortable before committing to it.t0011,t0012,t0013,t0015are randomized (t0011-random-ops.shand friends). They pass on their seeds today; if they turn out to be flaky under CI's timing, they may want separating from the deterministic ones rather than destabilising the whole job.Related: #202 (the harness supplies no git identity) would let the workflow drop its
Configure git identitystep either way.