Re: _AC_OUTPUT_SUBDIRS/configure.in

[Akim Demaille]

>>>>> "Ralf" == Ralf Corsepius <address@hidden> writes:

Ralf> Akim Demaille wrote:
>> You are very right, thanks.  May I ask you how you discovered this
>> problem?
Ralf> Well, I am using a specialized/extented (IMO: enhanced) variant
Ralf> of autoconf-2.13's AC_OUTPUT_SUBDIRS macro for toplevel
Ralf> configuration of a deep sourcetree. I am currently adapting this
Ralf> macro (full of autoconf-2.13 specific hacks) to the demands of
Ralf> autoconf-2.50. When comparing my macro against autoconf-2.49d,
Ralf> it was obvious something must have been broken.

>> It seems quiet benign, and I wonder if it can actually be triggered
>> in the real world.
Ralf> So do I. In my opinion the risk of bogusly picking up and trying
Ralf> to run an autoconf configure.in instead of actuallly finding a
Ralf> Cygnus-configure.in is pretty high.

Ralf> Therefore, IMHO, the check for the configure.in either should be
Ralf> extended to actually check for Cygnus configure.in, or should
Ralf> even be dropped entirely for the sake of simplicity, because
Ralf> probably nobody but the toplevel configury of the unified Cygnus
Ralf> source-tree (i.e. gcc, binutils and friends) nowadays applies
Ralf> Cygnus configure at all.

I'll need Alexandre's opinion on this.  I guess he will be able to
suggest a better Cygnusism test.