> Date: Sun, 23 Oct 2022 16:07:00 +0200
> From: [email protected]
> Cc: [email protected], [email protected]
> 
> > I've run "make check" in many_input_files, and it reported 3 failed
> > tests: input_dir_nonascii, output_dir_nonascii, and
> > output_dir_file_nonascii.  Given the "nonascii" part, should I look
> > into the failures?
> 
> There are file/dir names encoded in UTF-8 given as arguments.  We
> assume that the shell pass the file names as binary strings (and that
> the corresponding directories and files can be created), if I understood
> well your previous messages, this is not true in Windows.

Yes, Windows "interprets" the bytes according to its (incorrect, in
this case) assumptions.

> We also set LC_ALL=C and LANGUAGE=en in the tests for portability, I do
> not know if some other value could be relevant in Windows.
> 
> I am not sure that we can do anything to have those tests work for
> native windows setups.  Maybe the best would be to skip if on windows as
> determined by the HOST_IS_WINDOWS conditional sets in configure.ac.

Yes, I think we should skip them.

Reply via email to