Update for pandoc 3.1.7

[hseg]

Tests updated in view of jgm/pandoc@f7035d0

[hseg]

Unsure why nix build is failing, though

[lierdakil]

Unsure why nix build is failing, though

In all likelihood, you just need to bump hackage.nix in flake.lock (e.g. nix flake update). But feel free to ignore it.

This Pandoc update, however, is a problem, see #402. We need to remove label generation from equations, apparently 🤔

[hseg]

Ah, missed that. That will require better understanding of these codepaths than I possess, hope I at least gave something of value.

[lierdakil]

The update required non-trivial changes to the output, so I've opted to do those manually. In any case, thanks for the PR, sorry I didn't merge it.

[hseg]

On Sat, Sep 23, 2023 at 10:23:51AM -0700, Nikolay Yakimov wrote: The update required non-trivial changes to the output, so I've opted to do those manually. In any case, thanks for the PR, sorry I didn't merge it.

No problem -- in view of #402, my PR wasn't very useful. Though, it does raise the question -- it took quite a bit of effort to write, and it's something I'd rather not duplicate next time there's a breaking change. Is there interest in eg teaching the testsuite to generate new golden tests when pandoc's changes don't break stuff for us? That way, this PR could've just been a single script invocation in the release engineering.

[lierdakil]

Golden tests in test/m2m are relatively easy to regenerate, essentially find -iname input.md | while read i; do pandoc $i -F pandoc-crossref -t markdown-raw_attribute-raw_html -o ${i%/input.md}/expect.md; done and similar for -t latex.

Golden tests in test/test-pandoc-crossref.hs kinda have to be done manually. It may be worth converting at least some of those to something more manageable -- certainly at least the LaTeX group (which is statistically the most problematic in terms of Pandoc updates) can be converted to files. The code there is pretty gnarly, though.