Re: [PATCH] New test: distributed broken symlinks cause failure.

[Ralf Wildenhues]

Hi Stefano,

* Stefano Lattarini wrote on Sun, Jun 20, 2010 at 10:27:26PM CEST:
> At Sunday 20 June 2010, Ralf Wildenhues wrote:
> > The test doesn't look safe for MinGW.  Its 'ln -s' is emulated by
> >  'cp -p' which, from looking at the test, will cause a failure not
> >  a skip.
> Why? The test is skipped if `ln -s NONEXISTING-FILE SYMLINK-NAME' fails,
> and this should be the case if `ln -s' is equivalent to `cp -p'.
> Am I missing something?

Ah, right, I overlooked that.  Thanks.

> > I'm wondering whether we should be throwing short file name support
> >  out of the window completely with Automake; this is an issue for
> >  this test because distlinksbrk.test shares >= 8 characters before
> >  the dot with distlinks.test; info Autoconf 'File System
> >  Conventions' has details.
> Apparently, by what I read there, the only systems that might be affected
> are:
>  1. plain DOS systems, and
>  2. Win32 systems with the registry edited to match the old DOS
>     behaviour w.r.t. file names (which is a really poor decision if one
>     plans to run Posixy software on those Win systems).

Please see <http://www.delorie.com/djgpp/v2faq/faq22_19.html>.
<http://www.delorie.com/djgpp/doc/kb/kb_10.html> might also be
interesting.

> Do we still care about these systems?

That is the question, right.  There have been bug reports against
Autoconf about a year ago, so someone is still interested in this.
Not many people, I presume.

Oh well.  Let's leave it at a "when introducing new file names, prefer
those that are unique within 8.3 characters" recommendation, so that
when someone wants DJGPP support, we don't have to do too much clean up
work, ok?

Thanks, I'm adding your patch to maint.

Cheers,
Ralf