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
The documented dogfooding does not run: no workflow produces pinax-report-delegation, and runtests.jl never installs the capture #120
docs/src/test2pinax.md:11-17 presents a link -- "Open Pinax own test report ... it is the CI run output" -- and :38-46 a "three entry points, one suite" proof table.
Measured
.github/workflows/Documentation.yml:33 name: pinax-report-delegation # downloads it
grep -r 'upload-artifact' .github/workflows/ | grep pinax-report # 0 producers
grep -c 'install_test_capture' test/runtests.jl # 0
.github/workflows/CI.yml -> one job, `test`, which is
uses: QAtlasHub/TestShards.jl/.github/workflows/sharded-tests.yml@main
if_no_artifact_found: warn means the download quietly succeeds at nothing, docs/make.jl:100 logs "the Test -> Pinax link 404s", and the docs page ships with a dead headline link.
test2pinax.md:88-101 says install_test_capture!() is mandatory precisely when a reusable action owns Pkg.test -- which is this repository situation -- and runtests.jl does not call it.
Consequence
The CI runs the suite one way, not three. The dump -> merge -> render path, which is the reason dump_test_report, load_test_dump and _render_matrix exist, is exercised by no CI job.
This is the meta-issue behind several others filed alongside it: the wrong-margin bug, the zero-check PASS, the theme-blind cache and the dropped shard would all have surfaced from a report the project actually generated for itself.
Fix
Either make the claim true -- have CI produce pinax-report-delegation, and call install_test_capture!() from runtests.jl -- or remove the link and the proof table until it does.
#140 (73f2f77) took the second of the two options in this issue — the documentation no longer claims what CI does not do. This issue stays open for the wiring, and docs/src/test2pinax.md now names it as the tracking issue, so closing it would leave the page pointing at a closed issue.
What landed:
the ../test-report/ link and the "CI runs that same suite three ways" proof table are gone; the three entry points remain as a description of the API, under a note saying Pinax does not yet dogfood them
docs/make.jl, .github/workflows/Documentation.yml and test/test_report_demo.jl each still asserted a deployed self-report existed — all three corrected
test/test_readme.jl now ties the two ends together rather than pinning either: the docs may link ../test-report/only if a workflow uploads a pinax-report artifact. That assertion keeps holding when the wiring lands, so it does not have to be revisited then.
What remains is this issue's first option: install_test_capture!() in runtests.jl, and a CI job that produces pinax-report-delegation.
Activity
sotashimozono commentedon Sep 7, 2026
#140 (
73f2f77) took the second of the two options in this issue — the documentation no longer claims what CI does not do. This issue stays open for the wiring, anddocs/src/test2pinax.mdnow names it as the tracking issue, so closing it would leave the page pointing at a closed issue.What landed:
../test-report/link and the "CI runs that same suite three ways" proof table are gone; the three entry points remain as a description of the API, under a note saying Pinax does not yet dogfood themdocs/make.jl,.github/workflows/Documentation.ymlandtest/test_report_demo.jleach still asserted a deployed self-report existed — all three correctedtest/test_readme.jlnow ties the two ends together rather than pinning either: the docs may link../test-report/only if a workflow uploads apinax-reportartifact. That assertion keeps holding when the wiring lands, so it does not have to be revisited then.What remains is this issue's first option:
install_test_capture!()inruntests.jl, and a CI job that producespinax-report-delegation.