On Sat, May 30, 2026 at 09:14:57AM +0100, Gavin Smith wrote: > On Fri, May 29, 2026 at 07:28:39PM +0200, Patrice Dumas wrote: > > Hello, > > > > Now that there is always unicode collation for C, I would propose to > > have the same requirement for Perl, that is require Unicode::Collate > > to be present. Nowadays, texi2any in pure Perl is a different option, > > mainly relevant for users in case of issues with C, I think that it > > would be better if the pure Perl option had the same features as the C > > command. > > That sounds like a good idea. We have had issues in the past of the > fallback code (Texinfo::CollateStub) not being tested and having problems.
I also remember that, but I believe that it is the fallback itself that was problematic, as it there is at least some testing of Texinfo::CollateStub through setting USE_UNICODE_COLLATION=0 when one test with TEXINFO_XS=omit, something I try to do for all changes. > Can I confirm that if Unicode::Collate is not present, but the C index > code is used instead of the Perl code in Texinfo/Indices.pm, then there > will be no error? Yes. There seems to be no difference between the Unicode::Collate and your code sorting orders. The situation of sorting in C is actually the one tested in the default case now if XS is enabled, which is the common situation, as in the default case there are XS overrides for index sorting for all the converters, not only for HTML. However, I still test often with TEXINFO_XS=omit and some platforms in the github CI also do not have XS enabled, I think. -- Pat
