It actually breaks compiling altogether, because the
old egcs (kgcc) on Red Hat doesn't work with the
"-march=k6" switch.
The current gcc (from 7.0 updates or from beta/raw
hide) actually works very well, and all that the
current "configure" script does is screw up.
Even more annoying is that you've broken "CC=gcc
./configure ..." too, so I can't override it. I
consider this a bug in the current build system for
quakeforge. I have a uquake-glx from the "old" tree
built with RH gcc-2.96 that works quite happily, so I
am fairly sure that the test you have in place is both
heavy-handed and in many cases unnecessary. If you
must, at least check to see if it the original
RH7.0-shipped compiler, rather than completely failing
for those of us with more recent versions, please?
Sorry to rant, but I actually find RH7.x very usable, I
have had no major problems compiling anything, I have
working DRI-accelerated 3D graphics, UT and Q3A work,
quakeforge (oldtree) works a treat, and I would like to
test the current quakeforge with it.
Logged In: YES
user_id=2761
THIS IS NOT A BUG!
The disabling of gcc-2.96 was NOT done lightly. It produces
broken code for quakeforge and has thus been banned (most
noticable in the server). gcc-2.95.4 (even though it's not
an official release), available from rh's contrib section,
is reported to work well.
Logged In: YES
user_id=14762
Can you point me to an example (please) because I really am
surprised by this ... I have the old tree built and working
fine, giving me quite impressive FPS with my G400. It's one
of my test cases :)
Has this been reported to Red Hat yet?