Re: 07-fyi-deterministic-cond3.patch

[Akim Demaille]

>>>>> "Tom" == Tom Tromey <address@hidden> writes:

>>>>> "Akim" == Akim Demaille <address@hidden> writes:
Akim> It contains only bug fixes, the combinatorial explosion can be
Akim> considered as such, and anyway, as far as I'm concerned, it
Akim> works, I use it.

Tom> The trunk needs autoconf 2.52, right?  So it isn't just bug
Tom> fixes.

Correct.  In this regad it is different from 1.5.1.

Akim> So why not releasing 1.6 +/- now?

Tom> That's a different question.  Maybe this is the thing to do.  I
Tom> haven't been reading my automake email lately.  I don't know if
Tom> there is a big demand for a 1.5.1 bug fix release that works with
Tom> the old autoconf.  If not, I agree -- 1.6.

The thing is that most bug reports are related to CDPATH and zsh: on
Darwin they use zsh as sh, and Automake macros break the job done by
2.50, which copes with these issues.  At that time it was just because
I was working with Zsh, so I didn't realize Automake would _need_ it,
but in the meantime Darwin came to select good macros from bad ones :)

I am using amtraces=autoconf, and it works great.  There are still a
few tests failures, but (i) because some tests use LIBOBJ assignments,
which is now wrong with Autoconf, uses AC_LIBOBJ (precisely introduced
so that Automake could trace them), and (ii), because I am not tracing
the obsolete macros, hence it doesn't get the complaints.

About (ii), (a) I still don't think this is Automake's job, that's a
job for Autoconf, in particular AU_DEFUN and AU_ALIAS, and (b) anyway
aclocal already complains, so why should automake double this job.
Are there really people using automake and not aclocal?