Re: [Bug 375] New - Fix for autoconf >2.5

----- Original Message -----
From: "Henrik Nordstrom" <hno@marasystems.com>
To: "Robert Collins" <robert.collins@syncretize.net>;
<squid-dev@squid-cache.org>
Sent: Saturday, July 06, 2002 1:10 AM

> Hmm.. I think there is a way to make this specific test in a
> backwards/forward compatible way.. I assume autoconf 2.5x uses the
> same kind of results cache as earlier versions? If so then the
> following should work fine:
>
> 1. Instead of modifying the LIBOBJS variable, modify the result cache.

Erk. The problem is changing LIBOBJS at all outside the AC_LIBOBJ macro.
What we need to do is reproduce that macro for 2.13, without colliding with
a 2.5x definition.

> 2. Then use the existing AC_REPLACE_FUNCS macro to modify LIBOBJS. As
> the cache is already populated it won't run the test again and simply
> modifies LIBOBJS if needed as indicated by our custom test.

IIRC LIBOBJS isn't modified. Thats the point of the change. There is some
reasoning in the changelog/manual. I'm writing this offline however, so
can't be specific at this time. I'll review and get back to the list.

> What is the other autoconf 2.5x compability issues we might need to
> deal with?

The handling of object extensions has changed. Automake has begun depending
on autoconf 2.52 and above. the quoting rules for m4 macros have been
clarified and made consistent. Many macros have been replaced with more
flexible forms.

I really don't think it's worth putting time into trying to have a 2.5x and
2.13 configure.in when there are what - 6 people? that regulary need to
create configure, and as the platform where configure is created doesn't
matter, we can do that anywhere!

Rob
Received on Sat Jul 06 2002 - 01:34:50 MDT