Re: Removal of libtool1.4
Henning Makholm <henning@makholm.net> writes:
> Some arguments are given on
> <http://www.invisible-island.net/autoconf/autoconf.html>. I hasten to
> add that these are not *my* arguments, and I haven't even tried to
> decide whether I think they are valid or not. But it is clear that
> this particular upstream author would reject a patch that upgraded his
> configure.in to 2.5x. I assume that there are other authors out there
> who feel like him.
Autoconf 2.5x has been unstable, with various bugs and syntax changes
between versions and some experimentation with completely new techniques
that's broken backward compatibility on some platforms. The rate of
change seems to have finally slowed down some with Autoconf 2.57, though.
Note that there are some regressions in portability relative to Autoconf
2.13; for example, Autoconf 2.57 is more picky about what shell it wants
to use and sometimes has significant speed problems running on old,
non-Linux systems that don't have bash.
The internals have changed *completely*, so any configure script with
hooks deep into the insides of Autoconf (XEmacs is one of the best
examples) will fail miserably and porting to 2.5x will be extremely
difficult.
--
Russ Allbery (rra@stanford.edu) <http://www.eyrie.org/~eagle/>
Reply to: