On Sat, Jan 17, 2026 at 11:56:20AM +0200, Eli Zaretskii wrote: > > From: Gavin Smith <[email protected]> > > Date: Thu, 15 Jan 2026 16:58:13 +0000 > > Cc: [email protected], [email protected] > > > > > IOW, this warning was expected, but was missing. I see that this > > > warning comes from texi2any.c:merge_opened_files, but I don't know how > > > to investigate the reason(s) -- any hints? > > > > It should not come from texi2any.c because that is the alternative > > ctexi2any implementation that is not used by default. > > Oh, so the "--enable-c-texi2any" configure-time option defaults to > "no", not to "detect"? I thought it defaults to "detect", because > "./configure --help" says "Use texi2any program implemented in C if > possible", and I see ctexi2any.exe was built as part of the build, so > evidently that _was_ possible. Then why isn't it used by the test > suite (and not installed, as I say in my previous message)?
I don't know why it's built when it's not used. If I run "./configure PERL_EXT_CC=failcc", this completes successfully and disables XS modules: configure: error: in '/home/g/src/texinfo/GIT/tta/perl/CheckXS': configure: error: C compiler cannot create executables See 'config.log' for more details configure: error: ./configure failed for perl/CheckXS configure: building Perl extension (XS) modules will be disabled "make" does appear to complete successully (which it didn't before), but only after a lot of time building C texi2any code. But then "make check" fails badly. $ perl -w t/02coverage.t ok 1 perl: symbol lookup error: t//../../C/.libs/libtexinfoxs.so.1: undefined symbol: newSVpv_utf8 According to the output of "TEXINFO_XS=debug perl -w t/02coverage.t", a large amount of C libraries and XS overrides are loaded anyway.
