Re: Texinfo 7.2.90 build with mingw.org's MinGW: testing

> Date: Sun, 18 Jan 2026 23:03:42 +0100
> From: [email protected]
> 
> On Sun, Jan 18, 2026 at 09:37:36PM +0000, Gavin Smith wrote:
> > On Sun, Jan 18, 2026 at 10:26:45PM +0200, Eli Zaretskii wrote:
> > > > Date: Sun, 18 Jan 2026 21:01:51 +0100
> > > > From: [email protected]
> > > > Cc: [email protected], [email protected]
> > > > 
> > > > I couldn't find anything obvious, here is a patch that shows directly
> > > > the option value, please apply and report the value.
> > > 
> > > This code from Converter.pm:
> > > 
> > >      if (defined($self->get_conf('SUBDIR')) and $output_file ne '') {
> > >        my $dir = File::Spec->canonpath($self->get_conf('SUBDIR'));
> > >        print STDERR "DFD SUBDIR: '$dir'\n";
> > >        $output_file = join('/', ($dir, $output_file));
> > > 
> > > calls File::Spec->canonpath.  I'm pretty sure it returns file names
> > > with backslashes on Windows, evidently because that's the "canonical"
> > > form of file names there.
> > > 
> > > So we could have a customized version of canonpath, which on Windows
> > > mirrors all backslashes to forward slashes.
> > 
> > We could solve the problem in two steps.  In texi2any:merge_open_files,
> > we could call File::Spec->canonpath.  As the failing test is testing
> > whether the error message is output, this should make the error message
> > output.
> > 
> > Further, for consistent reference test results, we could normalise
> > the error message (but not delete it, as I suggested before).  We
> > could delete the filename after the message "overwriting file".
> 
> I prefer replacing backslashes by forward slashes in code, otherwise the
> file will compare differently and we will miss overwritten files.

I agree: using forward slashes everywhere is a safer approach.