Re: automake 1.7.2b, tests/depcomp3.test and DJGPP

[Alexandre Duret-Lutz]

>>> "Richard" == Richard Dawe <address@hidden> writes:

 Richard> Hello.
 Richard> Alexandre Duret-Lutz wrote:
 >> 
 >> >>> "Richard" == Richard Dawe <address@hidden> writes:
 >> 
 Richard> Some of the tests fail, because they use DESTDIR.
 >> 
 >> Huh?  What's wrong with DESTDIR?

 Richard> Sorry, I should been more verbose. I sent a mail about
 Richard> this problem to bug-automake before (10th January
 Richard> 2003) - I can forward it to you or resend it, if you'd
 Richard> like. Below is a summary of the problem:

Thanks, I've just found this in the archives.  Now I remember I
already read it.  However the thread discussed the user side of
this issue, and I didn't realized it had an impact on the test
suite and most likely on distcheck.


 Richard> DESTDIR does not work with DJGPP. automake tries to
 Richard> create paths like $(DESTDIR)/$(binprefix). $(DESTDIR)
 Richard> and $(binprefix), etc. usually contain fully-qualified
 Richard> DOS-style paths (c:/some/file). When two
 Richard> fully-qualified DOS-style paths are concatenated, they
 Richard> give a filename that doesn't make any sense -
 Richard> c:/foo/c:/bar. This causes the DESTDIR-related stuff
 Richard> to fail.

What are the failing tests?  All those that use `pwd` (either as
an argument to --prefix or when setting DESTDIR)?  And those
that run distcheck (that in turn uses `pwd` and DESTDIR)?

Would it be right just to drop the leading drive letter as
follows?

  workdir=`pwd | sed -e 
's,^[ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz]:,,`

I think we could define this in tests/defs.in, and adjust all
relevant tests to use $workdir instead of `pwd`.  This should
also go in the distcheck target.  All these places use `pwd` to
later designate a subdirectory, so the drive shouldn't matter.

[...]

 >> How about requiring the target to be at least two letters?  (Not
 >> that I find this particularly cunning, but I can't think of
 >> anything else right now.)

 Richard> I think that's reasonable, as long as the target file
 Richard> has a file extension, so that it copes with, say,
 Richard> dependencies in a.c.

Yep, we always run depcomp on objects, and objects always ends
with `.$(OBJEXT)' or `.lo'.

[...]

 Richard> Below is a patch that works for me. 

Thank you! I checked it in.

 Richard> This should also cope with filenames that have spaces
 Richard> in them.

Actually make can't support filenames with spaces :)

[...]

 >> It take the same amount of time here.  Unfortunately.

 Richard> What kind of box do you have? I'm running an Athlon
 Richard> 850MHz. I've wondered for a while if Linux is much
 Richard> faster than DJGPP, so I'd be interested to know what
 Richard> kind of box you are using.

Mobile Pentium II at 366Mhz here.  Sorry for the bad news.
-- 
Alexandre Duret-Lutz