Do not add checksums to css files if building using the htmlhelp builder

[mkay-sh]

Subject: The windows help compiler requires that css links don't have a query component. This PR disables the checksum for css files if the htmlhelp builder is used.

Feature or Bugfix

• Bugfix

Relates

• Fixes HTMLHELP Relative URL strings must not have a query component #11889

[AA-Turner]

Please could you add an entry to CHANGES?

[jayaddison]

This looks like it affects Sphinx 7.1.0 onwards currently, and I do think that this would solve the problem.

However: it's not great to introduce builder-specific fixes in Sphinx itself, and I'd suggest that maybe adding a crc32_checksum attribute (default True, overridden as False in the sphinxcontrib-htmlhelp builder) to the base HTML builder -- since that's the top-level at which this feature is relevant -- could be an alternative. (I considered a generic checksum attribute name -- but I think that there are other checksum/integrity methods we might want to consider in future, hence trying to avoid a future naming conflict)

That said: that would require some release co-ordination (addition + release in Sphinx, then a corresponding usage + baseline Sphinx dependency bump + release of sphinxcontrib-htmlhelp) - so so it's possible that I'm suggesting more work than is practically worthwhile.

[jayaddison]

Ok: now I understand that as of recently, many of the sphinxcontrib-* extensions no longer have Python package dependency references back to Sphinx itself. So in the short term, my alternative suggestion (implement an attribute in Sphinx, release, update the baseline dep in the extension, then release that) is not valid.

I'll try to learn a bit more about the reasoning behind the dependency versioning change (#11567); to my mind, the ability to list supported Sphinx versions from each extensions seems quite important.

[jayaddison]

I think my previous comments here were too hasty and cautious about this - I now think it's a reasonable solution to a genuine problem, and that unless another approach arrives (that I could and should offer myself if I find time), on balance it's better to fix the existing issue. My apologies for delay caused.

[mkay-sh]

Please could you add an entry to CHANGES?

I added an entry to the changes.

[picnixz]

Thank you! I'll merge it when the tests are done

[picnixz]

Thank you!