Re: mingw conflict between mkdtemp and sys_stat

[Eric Blake]

Eric Blake <ebb9 <at> byu.net> writes:

> 
> Any chance we can fix the mkdtemp module depend on the sys_stat module, which 
> already takes care of mkdir for mingw, rather than open-coding its own mkdir 
> replacement?  And would this be any closer to the glibc implementation?

For that matter, it is rather enlightening to compare tmpdir.c/mkdtemp.c 
against tempname.c/mkstemp.c - both are based on the same upstream glibc 
files.  Bruno and Paul, would you be willing to accept a patch that does the 
following things?

- Break tempname.c into a full-fledged module, and make the mkstemp module 
depend on the new module rather than including the file itself.
- Carve direxists/__path_search out of tempname.c, and make the (new) tempname 
module depend on the tmpdir module instead.
- Carve gen_tempname out of mkdtemp.c, and make the mkdtemp module depend on 
the new tempname module instead.
- Double check with glibc to make sure we have all the appropriate upstream bug 
fixes ported to gnulib.
- Make sure the final result behaves well on mingw.

-- 
Eric Blake