fix: docs pages

[jjmaestro]

fix: deprecated versions of actions/upload-artifact

The pages workflow is failing due to deprecated versions of actions/upload-artifact.

I believe this is why the docs are not updating (e.g. the Meson docs are still blank after #1060 landed).

For more info, see:

• https://github.blog/changelog/2024-04-16-deprecation-notice-v3-of-the-artifact-actions/
• https://github.blog/changelog/2024-02-13-deprecation-notice-v1-and-v2-of-the-artifact-actions/
• https://github.com/actions/upload-artifact

chore: bump version.bzl to the current released version

This is also needed for docs generation, see docs/tools/workspace_status.sh

docs: update releases

[jjmaestro]

I was hoping this would work... but there's a bunch of checks failing for the generation of the docs.

I tried to test the workflow in my fork but I wasn't able to :S @jsharpe if you know how to do it, I'll be happy to try.

[jsharpe]

I don't actually know how the doc generation works - @UebelAndre put it together. I believe the issue though is that its trying to run the latest version of stardoc on older branches to generate the docs for older releases and so its failing because bzlmod syntax such as @@ didn't exist when the earlier releases were done.

[jjmaestro]

Yeah, I was checking the checks and it does seem like it was doing that, e.g. running Bazel 7.4 with Bzlmod on the 0.7.0 branch and failing... it's a pity cos it stops the other checks from running with more recent tags, e.g. 0.12.0 :-/

For the time being, I'm removing the old releases to see if the checks pass for the new stuff, and if so, I'll revert the change and we can at least know that it works for the new stuff 😅

[jjmaestro]

OK, so this worked for the latest releases 🎉 Now, there's some other stuff failing...

Huh! some checksum is failing? 😮

(...)
Error downloading https://download.gnome.org/sources/glib/2.77/glib-2.77.0.tar.xz (...)
Checksum was 7cc9c60664653ddfca4d1cb016b756d054123c112116079f0a97ada510e1b07b 
but wanted 1897fd8ad4ebb523c32fabe7508c3b0b039c089661ae1e7917df0956a320ac4d
(...)

But... the checksum in

rules_foreign_cc/examples/third_party/glib/glib_repositories.bzl

Line 12 in d9367d2

sha256 = "1897fd8ad4ebb523c32fabe7508c3b0b039c089661ae1e7917df0956a320ac4d",

is right!?

curl -sL https://download.gnome.org/sources/glib/2.77/glib-2.77.0.tar.xz | shasum -a256 -
1897fd8ad4ebb523c32fabe7508c3b0b039c089661ae1e7917df0956a320ac4d  -

Also, the checksum fails for a few jobs and the checksums these jobs get are all different o.0

So... maybe some weird temporary error?

[jjmaestro]

Aaaah, I found this!

rules_foreign_cc/.github/workflows/pages.yaml

Lines 42 to 47 in 9edaefe

	- name: Patch older branches
	run: |
	mkdir -p ${{ github.workspace }}/.github
	curl https://raw.githubusercontent.com/bazelbuild/rules_foreign_cc/main/.github/docs-${{ matrix.ref }}.patch > ${{ github.workspace }}/.github/docs-${{ matrix.ref }}.patch
	git apply ${{ github.workspace }}/.github/docs-${{ matrix.ref }}.patch
	if: ${{ matrix.ref == '0.4.0' || matrix.ref == '0.3.0' || matrix.ref == '0.2.0' || matrix.ref == '0.1.0' }}

So, I guess the correct approach is to produce patches for all the old branches! A bit cumbersome but yeah, it's a good escape hatch @UebelAndre 😄

I'm going to try to patch docs/.bazelrc for those branches forcing to run without Bzlmod...

[jjmaestro]

OK, I think this last commit will fix the issue, but the checks fail because the curl is trying to get the patches from main and the patches don't exist for a bunch of releases that I'm now expecting to be patched :-/

The only way out I see is to create a separate PR with the patches and land it first, then rebase this PR. But that would also mean landing the PR with the patches "blind", without much testing so... maybe we could land this PR and see if it fixes the issues.

@jsharpe let me know what you want to do!

P.S. in case I need to re-do this in the future, I generated the patches with a quick-n-dirty shell script that could be easily adapted to generate other patches.

[jjmaestro]

I've tried reproducing the checksum error with another script but I can't, the checksum of the downloaded artifact always matches the expected checksum... so I'm not sure what's happening with those first errors in the checks.

@jsharpe since the doc generation is currently broken, I don't think that landing this PR would create further issues, so... maybe we can try and see if the CI jobs finally go green? :-?

[jjmaestro]

@jsharpe ping here too, I think this PR should land, the checks breaking seem completely unrelated (the weird GNOME glib checksum error that I haven't been able to reproduce).

[jsharpe]

Thank you for looking into why this was failing - I believe what you've done should work - will merge it and check!

[jsharpe]

This is almost there - but think some of the patches don't quite apply e.g. 0.3.0 - github cancels all the actions as soon as one fails so its a bit annoying to check whether any of the others are also bad..

[jjmaestro]

Ok, I'll give it another thought and will also try to modify the workflow to try and run all of the jobs so at least there's more info to debug!