Re: O(n) set manipulation

[Ralf Wildenhues]

Hi Eric,

First, thanks a lot for your efficiency work on Autoconf!

* Eric Blake wrote on Mon, Jul 28, 2008 at 03:10:45PM CEST:
> I intentionally
> documented m4_set output as unspecified order, to allow a future change
> where the elements are maintained in sorted order rather than entry order,
> by making set manipulation an m4 2.0 module.

Does that mean that the Autoconf output will depend also on the M4
version used, and possibly even system-specifics like type sizes
or so?  That would be a major pain for developers checking in configure
scripts to their repositories, and breaking a (possibly unspoken)
Autoconf promise.

Also, it's a bit sad to see that neither of us has fixed any of the open
regressions.  As I would really like to see a stable next release, I'm
even wondering a bit whether we should consider releasing it without all
these new changes, to avoid the risk of introducing new regressions.
I for one would be less concerned if they had some more time in the git
tree before being in a stable release.  But I certainly would like to
hear your judgement about this (I unfortunately haven't had the time yet
to read them carefully).

Cheers,
Ralf