gh-145000: find correct merge base in check-html-ids

[StanFromIreland]

github.event.pull_request.base.sha is not the same as the git merge-base, per https://github.com/orgs/community/discussions/39880. Instead compare the commits and use merge_base_commit (this also however has quite sparse docs :-( at https://docs.github.com/en/rest/commits/commits?apiVersion=2026-03-10&versionId=free-pro-team%40latest&category=commits&subcategory=commits#compare-two-commits).

• Issue: CI: Hash-pin all actions #146488

• Issue: Check removed HTML IDs #145000

[webknjaz]

I'm not sure what GH puts into that response object.. But here's something I've been playing with on the Git level for gradually fetching older commits given a shallow clone, if you'd like to experiment with that: https://github.com/ansible/ansible/pull/84375/changes#diff-dfbae2cdf0eee1583a3eea53c82843bb9d1dfde27e6a3832e4ee48ab8aa2a1d3R226-R294.

---

Instead of having two git checkout steps, I'd go with one, then load more commits gradually until enough of the Git tree is in cache and use a git worktree to produce the second folder with older files. I'd maybe also factor-in the number of commits in the PR (as returned by the API) to start with.

---

Though, your solution might be functional, I don't have any experience with this specific API response key. I don't know if they'll produce different responses when there's multiple commits merging main into the topic branch.

[StanFromIreland]

if you'd like to experiment with that: https://github.com/ansible/ansible/pull/84375/changes#diff-dfbae2cdf0eee1583a3eea53c82843bb9d1dfde27e6a3832e4ee48ab8aa2a1d3R226-R294.

Cool! I implemented it in bash.

[StanFromIreland]

It calculated the correct merge base :-)

MERGE_BASE: 9e5b8383724211d14165a32c0e7682e56e13843a

Edit: And it also calculated it correctly after updating the PR.