Dear Gavin, Eli & Patrice,
Yes, sorry, I should have done two different patches. Concerning the
MSYS2+MINGW64 console, this is only the first change that does the job,
ie handling "cygwin" same as "msys" when testing $OSTYPE.
> Am I correct in saying for this MSYS2 console, OSTYPE is "cygwin", but
> the output of "uname" does not contain the string "cygwin"? Does the
> output of "uname" contain anything else that we could check for in
> addition to "cygwin" and "msys"?
Yes, you are correct, in this MSYS2+MINGW64 console, uname outputs :
MINGW64_NT-10.0-19045
in the MSYS2+MSYS console, uname outputs :
MSYS_NT-10.0-19045
And in the MSYS2+MINGW32 console uname outputs :
MINGW32_NT-10.0-19045
Concerning the WSL2, you are right saying the the change is not as
simple, and we should probably let it boil down before doing anything. I
think that I will install a LaTeX distro for WSL2 too, and see what
happens, I have enough diskspace for that, the aim of using the native
distro was just avoiding installing LaTeX packages twice.
Actually, I am currently migrating from a MACOS to an MSW10 PC, and I
was struggling to compile from source some tool chain that is no longer
available as binary. The doc compilation was making the build fail, but
I just realized after making this patch that the top level script allows
to skip documentation generation.
To Karl :
> TBH, I'm not sure we should support this environment, unless you or
> someone else will come on board as its responsible person, and will
> actively participate in discussions and testing pre-releases. That's
> because there are quite a few subtleties involved in running Texinfo
> on non-Posix systems, and we need active experts on each such
> environment on board to make sure some changes don't break it or some
> other supported environment.
I agree that MSYS2 has some suttleties as far as TeX, notably when
someone does something like
tex '\def\somesetting{Hello World !}\input{some_file}'
then MSYS fails in interpreting some_file as a filename and apply its
filename conversion euristic.
IMHO the root cause has more to do with the tex command line syntax and
how tex handle filename.
Anyway, I expect that if you install TeXLive for WSL2 package, then
everything will run just as under Linux. TBH I have not yet done that, I
will let you know if I do it.
Vincent.
________________________________
De : Eli Zaretskii <e...@gnu.org>
Envoyé : vendredi 18 avril 2025 08:27
À : Vincent Belaïche <vincent....@hotmail.fr>
Cc : bug-texinfo@gnu.org <bug-texinfo@gnu.org>
Objet : Re: Making texi2dvi working better on Microsoft Windows 10
> From: Vincent Belaïche <vincent....@hotmail.fr>
> Date: Thu, 17 Apr 2025 20:46:45 +0000
>
> Just to make my point more clear, with the patch texi2dvi will still fail
> absolute paths under WSL2 if the TeX
> engine is a non-WSL2 executable (for which the .exe extension is needed). I
> have not yet tried to install a
> LaTeX distro Linux package for WSL2, I expect that in that case it would just
> be like under Linux.
TBH, I'm not sure we should support this environment, unless you or
someone else will come on board as its responsible person, and will
actively participate in discussions and testing pre-releases. That's
because there are quite a few subtleties involved in running Texinfo
on non-Posix systems, and we need active experts on each such
environment on board to make sure some changes don't break it or some
other supported environment.
But it's not my call, it's Gavin's and Patrice's.
