Ralf Wildenhues <Ralf.Wildenhues <at> gmx.de> writes:
> > It turns out that this patch exposed a cygwin bug - open("t\\", O_RDONLY |
> > O_CREAT) was creating the regular file 't' rather than failing with ENOENT
> > since the directory 't\' does not exist (remembering that \ is an alternate
> > directory separator, rather than a regular file name character). So I'm
> > checking in this, while waiting for the cygwin maintainers to fix the bug:
>
> Thanks for cleaning up after me. FWIW, I'm a bit surprised because this
> test did not error out for me on Cygwin.
It didn't error out for me, either, on a plain 'make check'. All it did was
create the bogus file 'tests/t'. But on 'make distcheck', it would have
errored out that you left junk in the directory (at least, I hope so - make
distcheck takes too long to run on cygwin, and I only noticed the problem
because git stat was no longer silent).
> I noted that the actual patch you checked in,
> <http://article.gmane.org/gmane.comp.sysutils.autoconf.cvs/838>,
> was good, but your post showed the other one.
Oops - I created the file, but forgot to refresh the editor to visit the new
file contents before I did a copy-n-paste. For the record:
http://git.sv.gnu.org/gitweb/?p=autoconf.git;a=commitdiff;h=53a1ce8
--
Eric Blake
