>>>>> "Alexandre" == Alexandre Oliva <[EMAIL PROTECTED]> writes:
Alexandre> On Aug 9, 2000, Akim Demaille <[EMAIL PROTECTED]> wrote:
>> Why not, it's just too early to make a decision on this IMHO.
Alexandre> Ok, since you're desperate to make this snapshot before
Alexandre> leaving for a long trip, let's keep this out of the
Alexandre> snapshot, and decide about it later.
I really mean for 2.50, not just the snapshot.
>> What if someone AC_PROG_CC_NO_LINK and then AC_PROG_CC?
Alexandre> AC_PROG_CC is run-once, isn't it?
Nope.
>> One problem is that your proposal addresses a single problem. We
>> know there are plenty of problems to address in this area. We must
>> come with a better design of all the compiler looking macros.
>> Addressing each problem individually might produce a monster which
>> is unmaintainable, not simple to understand and use, not to mention
>> backwardisms.
Alexandre> I see. AC_NO_EXECUTABLES is indeed better in this sense.
Yes, I agree. And maybe options set before calling the macro is even
better. I don't know. We need a clear picture of iot before changing
anything. We need specs too, for instance it is not obvious we really
want to check the compiler one after the other, or if we do, to what
extent we should check them etc.
Really, I'm in favor of something which makes it possible for you to
address your issues, but the real nice clean answer will probably be
brought by 2.51, not 2.50.
Alexandre> If you feel like checking it in before the snapshot, I'm
Alexandre> all for it, but I'd rather have it fully disable linking,
Alexandre> instead of just setting up the magic for per-language
Alexandre> disabling.
OK, I will do that.
