On 07 Aug 2001 22:19:25 +0100, Nix wrote:
>On Tue, 7 Aug 2001, Alexander Mai gibbered:
>> autoconf 2.50 in turn breaks the vast
>> majority of all configure.in setups of various packages on my harddisk.
>
>It breaks maybe twenty on my system (out of >300). Not `the vast
>majority' by any means.
>
>I'm wondering how you can have managed such an unlucky selection :)
Well, in physics people sometimes tend to claim 3 to be
a large number, so I didn't heistate to do statistics
on a small multiple of that :-)
300 is an impressive number indeed.
>> Also it's clear that I can neither debug all configure.in's nor
>> autoconf itself - that's not an option.
>
>Build autoconf-2.13 or 2.50 and rename the programs to autoconf-2.13 and
>autoheader-2.13, then run them when needed and set AUTOHEADER and
>AUTOCONF to the right binaries when running such a project's makefile.
>
>> So what's left? libtool people insist on going ahead, I think 1.4 even
>> could be built using autoconf 2.13, CVS version no longer does so.
>> I don't like the idea of having two sets of auto* installed on my
>> systems just to work on LessTif.
>
>Why not? They're not very large.
>
Things tend to be very subtle ...
e.g. when working on LessTif life is not only done with
running "make" which might be replaced by calls like
"set AUTOCONF=foo; make".
One is running test scripts, release scripts, etc.
I really wonder what the auto* maintainers are going to do,
once they would listen to the comments they will get from
the community. (now for sure someone will claim here that they do:
they don't. And be asure that only a small percentage of all
affected projects have already complained)
Perhaps we should go back to imake :-)
---
Alexander Mai
[EMAIL PROTECTED]
