[MPlayer-cvslog] r27288 - trunk/configure

[Reimar Döffinger]

> "Exactly what ao_alsa.c needs"? How so? How is <sys/time.h> related at
> all except that you saw a broken header conflict in one case (and a case
> with broken compilation settings anyway)?

Well, ao_alsa needs the alsa header to work with sys/time.h.
Though I do not see why it is needed at all right now, none of the stuff from time.h
is used and current alsa headers include time.h. Does someone have some
really old alsa versions around to test if it is ever needed?

> At least the alloca uses I noticed were in macro definitions only, and
> so only matter if the source file uses the snd_*_alloca() macros. Since
> the configure test does not use those macros it does not need alloca.h.

Well, obviously we have different ideas of what configure is supposed to
do. I think it should check which features actually _work_ and enable
those and not just enable those that somehow if you are not too unlucky
could work.
Of course there are corner cases that are too unlikely and too much
effort to test and/or maintain to be worth testing, but that is a
different point.
That said, does someone remember why we are using that alloca mess
instead of the proper malloc variants? Weren't they available in earlier
versions or something?