Context
action-link-checker v1.1.0 (released 2026-08-03) added ignore-patterns, default-on issue deduplication, and a fix for the timeout path that made the existing false-positive handling unreachable. lecture-python-programming adopted it in QuantEcon/lecture-python-programming#592: pinned the action to @v1 and declared the repo's known false positives (FRED and friends) as ignore-patterns, verified against the published HTML.
The other repos using the action have not been updated (checked 2026-08-04):
| Repo |
Current ref |
ignore-patterns |
lecture-python-intro |
@main |
none |
lecture-python-advanced.myst |
@main |
none |
continuous_time_mcs |
@v1.0.0 (predates dedup and ignore-patterns) |
none |
Why this matters
The two repos on @main already inherited v1.1.0's default-on deduplication ungated the moment it merged — an unpinned action ref means every future behaviour change also ships to them ungated. That is the same failure mode that broke the execution workflows in lecture-python-programming (unpinned jupyter-book → QuantEcon/lecture-python-programming#569). Meanwhile continuous_time_mcs is pinned to a release that predates the false-positive fixes entirely.
Each repo also needs its own ignore-patterns before its recurring link-check false-positive reports stop — lecture-python-programming was averaging one duplicate FRED report per weekly run before #592, and the same FRED-style links exist across the series.
A series split across pinning schemes (@main, @v1.0.0, @v1) is worse than any single choice: the same weekly report can behave differently in every repo.
Suggested change (per repo)
Port QuantEcon/lecture-python-programming#592: pin uses: QuantEcon/action-link-checker@v1, and add ignore-patterns mirroring each repo's existing linkcheck_ignore list in lectures/_config.yml. lecture-python-intro shares linkcheck.yml near-verbatim with lecture-python-programming, so the port there is almost mechanical; the other two need their own pattern lists derived from their configs.
Filed here rather than as three sibling PRs so the rollout can be owned as one piece of work across the series.
🤖 Generated with Claude Code
Context
action-link-checkerv1.1.0 (released 2026-08-03) addedignore-patterns, default-on issue deduplication, and a fix for the timeout path that made the existing false-positive handling unreachable.lecture-python-programmingadopted it in QuantEcon/lecture-python-programming#592: pinned the action to@v1and declared the repo's known false positives (FRED and friends) asignore-patterns, verified against the published HTML.The other repos using the action have not been updated (checked 2026-08-04):
ignore-patternslecture-python-intro@mainlecture-python-advanced.myst@maincontinuous_time_mcs@v1.0.0(predates dedup andignore-patterns)Why this matters
The two repos on
@mainalready inherited v1.1.0's default-on deduplication ungated the moment it merged — an unpinned action ref means every future behaviour change also ships to them ungated. That is the same failure mode that broke the execution workflows inlecture-python-programming(unpinnedjupyter-book→ QuantEcon/lecture-python-programming#569). Meanwhilecontinuous_time_mcsis pinned to a release that predates the false-positive fixes entirely.Each repo also needs its own
ignore-patternsbefore its recurring link-check false-positive reports stop —lecture-python-programmingwas averaging one duplicate FRED report per weekly run before #592, and the same FRED-style links exist across the series.A series split across pinning schemes (
@main,@v1.0.0,@v1) is worse than any single choice: the same weekly report can behave differently in every repo.Suggested change (per repo)
Port QuantEcon/lecture-python-programming#592: pin
uses: QuantEcon/action-link-checker@v1, and addignore-patternsmirroring each repo's existinglinkcheck_ignorelist inlectures/_config.yml.lecture-python-introshareslinkcheck.ymlnear-verbatim withlecture-python-programming, so the port there is almost mechanical; the other two need their own pattern lists derived from their configs.Filed here rather than as three sibling PRs so the rollout can be owned as one piece of work across the series.
🤖 Generated with Claude Code