Repository navigation
Re-probe the eight published hosts and record each host's newest publish run #135
Description
#127 §6.2. The 2026-09-07 comment on QuantEcon/workspace-lectures#40 reported 0 of 60 Track X rows cleared and attributed it to no publish since 2026-09-01. The validating session could not reach any of the eight hosts (python-programming.quantecon.org, python.quantecon.org, and the six quantecon.github.io/… sites) or the Actions API, so the box is open.
Partial evidence in hand
Newest publish* tag by creatordate in each host's repository: lecture-python-programming publish-2026aug20, lecture-python.myst publish-2026aug14, lecture-dp publish-2026jul07 (this repo also publishes on workflow_dispatch, which leaves no tag), lecture-python.zh-cn publish-2026aug24b, lecture-python-programming.{zh-cn,fr,fa} publish-2026aug19; .ml publishes on push, last gh-pages deploy 2026-09-04. No publish tag after 2026-09-01 anywhere, consistent with the ws#40 pass.
To do, from a machine with a browser or wget and gh
- For each of the eight repositories:
gh run list --workflow publish.yml --limit 3 --json createdAt,event,conclusion,headBranch— the newest run, not the tag list (tags are not date-ordered). - Re-probe five rows per host from the Track X tables on ws#40 with a different client from the original
curlloop (wget --spider --server-response, or a browser): expect 200 at the deletion-day byte count while nothing has published, 404 with the 9,379-byte Pages body once a publish follows a clean rebuild. Run the two controls per host first (a lecture page → 200, a never-existed path → 404). - Post the per-host table on ws#40 (that issue owns the rows) and the verdict on VALIDATION: independent review of the 2026-09-07 schema-decisions, rename and manifest-driven validator work #127.
If any host published after 2026-09-01 and still serves a deleted row, the settle-policy cache theory needs revisiting — say so on ws#40.
Exit criterion
Per-host publish runs and the 40 probe results are recorded on ws#40, the verdict is on #127 and its §6.2 box is ticked.
Activity
mmcky commentedon Sep 7, 2026
Re-probed 2026-09-07. The per-host table and the full method are on QuantEcon/workspace-lectures#40, which owns the rows: https://github.com/QuantEcon/workspace-lectures/issues/40#issuecomment-5568105442
A census of all 60 rows, not the 5-per-host sample this issue asked for, and 0 cleared. Two clients that are not
curl(GNU Wget, then Python urllib), full GETs with the body byte-counted. Controls discriminated on all eight hosts — eight distinct/intro.htmlbodies, so no host was on a wrong base URL, and one identical 9,379 B stock Pages 404 on a never-existed path, so no host has an SPA 200-fallback. Every deleted path answers 200 with a body byte-identical by sha256 to the pre-deletion git blob, fetched at the parent of each Track X commit. Nothing cleared, and nothing changed either. The kept control is still 200 at its original size, so Track X removed exactly what it intended.The row universe is now derived rather than asserted: the eight Track X commits removed 64 files, 60 of them under
_static/lecture_specific/.python.quantecon.orggenuinely has one row, not five.The attribution holds, and is now measured rather than inferred. Cache-buster GETs on a fresh CDN key returned identical bytes; for the three
gh-pageshosts the deleted file is still physically in the deployed tree. Seven hosts simply have not published since the deletions..mlhas deployed three times since — including on the Track X commit itself — and still serves everything, which is positive proof of thekeep_files: trueexemption rather than an inference from the workflow file.Two things moved forward. The rebuild half of the settle gate is now met on all seven cache-bearing hosts (this morning's
cache.ymlruns), so only the tag is outstanding anywhere. And the deploy mechanism is verified for all eight rather than assumed —quantecon/actions/publish-gh-pagesturns out to beupload-pages-artifact+deploy-pages, a full replace, so every non-.mlhost will prune on its next publish.Method note for the next pass:
gh run listcreatedAtis the run start, not the deploy —.fr's last publish ran 3.5 hours.gh api repos/<r>/deployments?environment=github-pagesis authoritative and works for all eight, including the artifact-based repos wherepages/builds/latest404s.#127 §6.2 is ticked. The rows themselves stay open on ws#40.
Generated by Claude Code