configure.php: recover failed XIncludes inside classsynopsis

[lacatoire]

xinclude_residual_fixup() recovers failed XIncludes on translations per
parent element, but had no classsynopsis case, so a failed include targeting
a methodsynopsis or fieldsynopsis inside a class synopsis hard-failed the
build (Unknown parent of failed XInclude: classsynopsis).

Adds a classsynopsis case injecting a <classsynopsisinfo role="comment">
marker, a valid child per the DocBook 5.2 RNG.

Related:

• [skip-revcheck] Remove empty xi:fallback elements doc-en#5699 removes empty <xi:fallback/> that masked such failures
• imagick: fix broken methodsynopsis xpointers in exception classes doc-en#5700 the Imagick fix whose failure first hit this path

@alfsb

[alfsb]

Got curious about the role="comment" part. I tested without the attribute (and after removing all `xi:fallback/locally), and it passes RNG 5.2.1 validation, on PHP 8.1.2-1ubuntu2.25, libxml 2.9.13. Checked the output to be sure that it was generated.

Does it break something locally, or does the missing attribute cause omitions/problems on the PhD side?

I prefer to keep fixups as minimal as possible. so no attribute or role='comment' (without the escape) is preferable. That said, it is a very ok change, that I may merge immediately.

[lacatoire]

Good question. Dropping the attribute passes RNG, but it changes the PhD rendering. In format_classsynopsisinfo (Generic/XHTML.php:1145), a role-less classsynopsisinfo is treated as the class-body opener: it appends {, sets classsynopsis close=true and asserts legacy rendering. With the comment role the marker renders as /* ... */, which is what the fixup wants.

I've switched to single-quoted role='comment' to drop the escaping, as you suggested.

[alfsb]

Nice. I really do not know these PhD specific behaviors.

Plan to merge this by monday, before any xi:fallback deletions.