Re: Skip filename recoding tests on MS-Windows

[pertusus]

On Mon, Oct 24, 2022 at 10:19:57PM +0300, Eli Zaretskii wrote:
> > Date: Mon, 24 Oct 2022 21:03:38 +0200
> > From: pertusus@free.fr
> > Cc: GavinSmith0123@gmail.com, bug-texinfo@gnu.org
> > 
> > > I've added debugging printouts and ran the above command.  What I see
> > > is that encode_file_name doesn't encode the file name at all, so it
> > > stays in its UTF-8 encoding, and the 'stat' in locate_include_file
> > > cannot find a file by that name.  The reason encode_file_name doesn't
> > > encode is that reverse_iconv is zero at entry to encode_file_name, and
> > > all of input_file_name_encoding, doc_encoding_for_input_file_name, and
> > 
> > input_file_name_encoding should be the string "ISO-8859-1", I checked
> > that it is the case for me.
> 
> Where is it set?  I mean, where is the original setting from which it
> is taken the first time?  Is it some file?

It is on the command line, see  -c INPUT_FILE_NAME_ENCODING=ISO-8859-1:
perl -w ./..//texi2any.pl -c TEXI2HTML --force --conf-dir ./../t/init/ 
--conf-dir ./../init --conf-dir ./../ext -I ./formatting/ -I formatting// -I ./ 
-I .  --error-limit=1000 -c TEST=1  --output 
formatting//out_parser/manual_include_accented_file_name_latin1_explicit_encoding/
 --info -c INPUT_FILE_NAME_ENCODING=ISO-8859-1 -D 'needrecodedfilenames Need 
recoded file names' ./formatting//manual_include_accented_file_name_latin1.texi
> 
> > > locale_encoding are NULL.  So none of the calls to iconv are made,
> > > reverse_iconv stays zero, and the file name is returned in its
> > > original encoding.
> > > 
> > > Let me know if you want me to debug this further.
> > 
> > Yes, that'd probably be good, although I have no clue as what could be
> > wrong.  You could add a printout in tp/Texinfo/XS/parsetex/Parsetexi.pm
> > like
> > 
> >       } elsif ($key eq 'INPUT_FILE_NAME_ENCODING' and 
> > defined($conf->{$key})) {
> >         print STDERR "III $key $conf->{$key}\n";
> >         conf_set_input_file_name_encoding ($conf->{$key});
> > 
> > If it is not set, then check earlier in texi2any.pl, maybe l 1111,
> > after procesing of the command line:
> > print STDERR "INPUT_FILE_NAME_ENCODING: 
> > ".get_conf('INPUT_FILE_NAME_ENCODING')."\n";
> > 
> > Also a possible location to check would be just before calling the
> > parser, near l 1364:
> > print STDERR "INPUT_FILE_NAME_ENCODING: 
> > $parser_file_options->{'INPUT_FILE_NAME_ENCODING'}\n";
> > 
> > 
> > If the INPUT_FILE_NAME_ENCODING is passed to the Parser, then the best
> > would probably be to check, in input.c at the end of
> > set_input_file_name_encoding whether input_file_name_encoding is set
> > to "ISO-8859-1".  If it is the case, I don't know what's happening as
> > the variable is not used much in the code besides
> > set_input_file_name_encoding.
> 
> None of these is set, and set_input_file_name_encoding ans
> set_locale_encoding are called with value of NULL.
> 
> > If it is not set, maybe a printout in
> > conf_set_locale_encoding in api.c cound be relevant
> 
> What should I print there?
> 
> > or looking at Parsetexi.c to see if something is strange.
> 
> Where to look there?