On Sun, Oct 04, 2026 at 10:13:30AM +0100, Gavin Smith wrote: > FWIW I agree with both points above and my understanding was that texi2any > already behaved like this on MS-Windows.
It did in the past, and then at some point I changed the input to be read in binary mode (probably by mistake), when I did a relevant change, that is do not change the encoding by using binmode, but by converting upon reading. Right now, there is an inconsistency between Perl and C, C does like Perl did before (not sure if it is by chance or on purpose, though). Actually testing native build, however, makes it possible to actually test what is going on there, and this is a relatively new setup, although Eli already tested something similar. Lastly, by chance, we have in the github CI a platform where native Windows on MSYS2 is tested with 1) pure Perl tested (msys2 mingw ucrt clang) 2) with Perl+XS tested (msys2 mingw ucrt gcc) The difference between those two allowed me to find the discrepancy between Perl and C code regarding input and CRLF. -- Pat
