On Jan 22, 2006, at 2:34 AM, Ralf Wildenhues wrote:
> Hi Brian,
> * Brian Barrett wrote on Sat, Jan 21, 2006 at 11:53:41PM CET:
>> Ralf, I was wondering if you could take a look at what I actually
>> committed (r8779) and let me know what you think. It's mostly your
>> patch, but I rewrote a good chunk of the m4 macro to take better
>> advantage of m4 and ditch all the extra variables that get eval'ed at
>> the end so that a prefix can be passed. We never pass in anything
>> other than a literal, so there's no problems with that. Seems to be
>> a lot shorter, although slightly more complicated to understand.
> Yes, much nicer indeed. Seems fine to me.
> I'd appreciate my name being spelled correctly, though.
Sorry about that (said with very red face). The svn history has been
> A mostly unrelated question, to fulfill this: ;-)
>> It's the thread that never ends... :)
> Which systems' m4 is capable of translating the configury but does not
> ignore unquoted [eval]? (See comment in ompi_get_version.m4.)
I was actually going to ask you about that -- that comment came from
your original patch. I was under the impression that AC actually
required GNU m4 these days. At least, Autoconf's configure complains
quite loudly when I build on Solaris or AIX and forget to have a GNU
m4 build in my path. I've always operated under the assumption
developers had to have GNU m4, as I couldn't see how they built AC