Re: [AD] broken dat and grabber?

[ Thread Index | Date Index | More lists.liballeg.org/allegro-developers Archives ]


Peter Wang <tjaden@xxxxxxxxxx> wrote:

> Anyway, now that we've figured out the problem, what to do about it?
> Certainly submit a bug report, but 2.95.x is becoming standard in
> newer distributions.  Should `configure' test for something and
> disable the `-funroll-loops' option for `blit.c' only?

yes! but not only for the unix build... we have to make a script or program
that finds out what gcc version we have and use that to enable workarounds.
ideally i would like to see the following solution:

IIRC gcc has an internal variable for version we can make a program simmilar
to mmxtest which then causes workarounds to be enabled.
this could directly append an extra target in makefile.all for blit.c which
overrides the normal one. or it could be nice and return an errorlevel (*)
according to what gcc version we have, this is a bit more powerful and could
be used to enable other workarounds as well.

hmm, i wonder if it would be best to put all this (mmx capability checking,
gcc version checking and other ugly shit we don't want in the makefiles)
into one single program and call it alutil.c or so?

*) errorlevel - the DOS way of returning a value to the command shell from a
program / the return value of main()

-henrik




Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/