Updating top-level autoconf to 2.59

Mark Mitchell mark@codesourcery.com
Tue Feb 13 00:25:00 GMT 2007

Alexandre Oliva wrote:
> On Feb  8, 2007, Paul Brook <paul@codesourcery.com> wrote:
>> Instead of just saying what you want, you have to reverse-engineer
>> the autoconf guessing logic.
> As with any form of communication, it takes some background knowledge
> to be able to express oneself.  For example, if you don't know C
> aliasing rules and violate them, you're up for surprises when the
> compiler optimizes guessing you know about them.  This is really no
> different.  And you don't have to reverse-engineer anything, the rules
> are documented in both cases.

I think the discussion of whether autoconf-2.59's option handling makes
sense or not belongs somewhere else: the autoconf list.  Therefore, I'll
not state an opinion as to whether or not I think this is a good change.

As CodeSourcery builds all manner of cross toolchains from many
different versions of GCC, the changes in autoconf are certainly going
to make work for us.  But, also, updating to autoconf-2.59 brings with
it some improvements.

Just as we don't want the autoconf people checking in
-fno-strict-aliasing as a default option, so, presumably, they don't
want us to make GCC behave differently than all other GNU programs.
And, if other GNU programs are using the new option semantics, then we
should too.

I think we should just document the change, and move on.  We'll all get
used to it (or write scripts to handle it) and there are lots of bigger
fish to fry.

Mark Mitchell
(650) 331-3385 x713

More information about the Newlib mailing list