Skip to content

Conversation

@dginev
Copy link
Collaborator

@dginev dginev commented Oct 15, 2025

Work in progress:

  • CI texlive docker images have been completed with the great aid of @tkw1536
    • and a tiny build+push of the final runner images from me.
  • Discovery: perl 5.42 has a regression in the test suite, due to \p{Latin} regex changes.
  • Discovery: The glossary test is failing under texlive 2024 (I believe @brucemiller had mentioned this in a meeting - please see PR remove non-portable glossaries test #2609)
  • Discovery: The texlive-enabled linux tests take forever for texlive 2023 and 2024 (55 min).
    • We probably want to only run those under make formats
  • Discovery: The latest choco package for miktex is also failing to install, needs more care.
  • TODO: Add strategic make formats targets to the CI matrix, ideally after we fix the dumper in the master branch.

Also, thanks to @xworld21 for suggesting the latest perl versions available via choco for Strawberry Perl.

@dginev
Copy link
Collaborator Author

dginev commented Oct 16, 2025

The (first?) perl v5.42 regression is traceable to Font.pm's tests, and in particular what reduces to:

"\x{02C7}" =~ /\p{Latin}/
  • true in perl v5.42
  • false in perl v5.40 and earlier

I am not sure what is the full range of characters affected yet, but I now know \p{Latin} has been adjusted.

If this only affects modifier letters we may get lucky, as the Font.pm check can use a negative check for \p{Lm}.

Update: The negative Lm filter worked! Regression resolved.
Update 2: Made a tracking issue in #2670 and switched here to \p{LC} positive match and \p{Script=Common} negative match, to avoid the card suits and letterlike math symbols.

@dginev dginev force-pushed the updated-ci-matrix branch 2 times, most recently from 01d097f to 0cad01d Compare December 4, 2025 17:25
@dginev dginev force-pushed the updated-ci-matrix branch from 0cad01d to c7fd3b4 Compare December 4, 2025 17:29
@brucemiller
Copy link
Owner

Interestingly 2023 & 2024 work with make formats, but 2025 fails. Probably should fail the build immediately if formats fails (and save the pointless tests)?

@dginev dginev marked this pull request as ready for review December 4, 2025 19:33
@dginev dginev requested a review from brucemiller as a code owner December 4, 2025 19:33
@dginev
Copy link
Collaborator Author

dginev commented Dec 4, 2025

Done. Let me know if you want to merge this as-is, specifically if I should undo the Font.pm changes. Now that we allow merging with texlive 2025 + perl 5.42 failing, it can also get patched elsewhere.

I will take yet-another look at the MikTeX setup, see if something can be salvaged.

Edit: Checked the choco avenue, not really usable. We may be better off trying to install texlive on windows. But that is outside of the scope of this PR, which just wanted to update the main texlive and perl targets.

@dginev dginev changed the title towards updated CI for v0.9 release in 2025 updated CI for v0.9 release in 2025 Dec 4, 2025
Copy link
Owner

@brucemiller brucemiller left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not perfect, but very close. Thanks for pulling the pieces together!

@brucemiller brucemiller merged commit 4175d31 into brucemiller:master Dec 4, 2025
14 of 15 checks passed
@brucemiller brucemiller deleted the updated-ci-matrix branch December 4, 2025 23:51
dginev added a commit to dginev/LaTeXML that referenced this pull request Dec 11, 2025
* towards updated CI for v0.9 release in 2025

* update the CI image list - published to gchr.io

* allow perl 5.42 to fail for now

* avoid the miktex install question for now

* attempt to guard \p{Latin} change in perl v5.42 by negative filter on \p{Lm}

* performance tweak: 3 regex at the cost of 1

* catch more cases in PCRE Latin Letter hunting

* [CI] run make formats on texlive 2023 and later

* [CI] fail fully on errors during make formats

* strict only in CI
@nschloe
Copy link

nschloe commented Jan 2, 2026

where v0.9 release in 2025

@dginev dginev changed the title updated CI for v0.9 release in 2025 updated CI for v0.9 release in 2026 Jan 2, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants