[gentoo-amd64] Re: Modular-X: bug 121394 (-fweb) apply on amd64?

[Duncan]

Simon Strandman posted <43F18FBC.1000107@×××××.com>, excerpted below,  on
Tue, 14 Feb 2006 09:07:24 +0100:

> Duncan skrev:
>> For those using modular-x, has anyone run into bug 121394 [1], window
>> corruption if CFLAGS include -fweb?
> I guess this is a gcc 4.0 only bug. -fweb is broken in gcc 4.0 and also
> causes bad performance regressions in many cases. xorg-xserver is
> probably not the only affected package.
>
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20743

You didn't read the (Gentoo) bug I referenced, and the related forum posts
there linked, did you?  The forum reports are all or almost all x86
compiled with gcc 3.x, so it's not the gcc 4.0 -fweb bug, but something
unrelated.

I was aware of a gcc 4.0 -fweb bug (tho I never bothered looking up the
details), and had removed that from my cflags when 4.0 was my
system default. However, the gentoo-xorg-modular bug I referenced is
apparently entirely unrelated to that.  The bug references several pages
on the xorg-modular thread on the forums, with quite a few folks there
reporting the issue, most of which were running gcc 3.x on x86 according
to their posted emerge info.

I hadn't experienced the bug here, having done my compiling with the gcc
4.1 snapshots with -fweb again enabled, and was asking if anyone else had,
either, who was running xorg-modular on amd64 but had compiled it with gcc
3.4.  I suspected, apparently correctly from the replies, that  the
problem didn't hit amd64 and was likely x86 only.  Thus, the filterflag
-fpic that's in the latest xorg-server ebuild should be conditional on
x86, or at least conditional on NOT amd64, since it doesn't appear that
the flag causes issues here and therefore needn't be filtered.

Back to the gcc 4.0 -fweb bug, AFAIK, it has been fixed for the 4.1
snapshots.  Whether it's fixed for 4.0.2, I'm not sure, but I've certainly
experienced no issues with the 4.1 snapshots, and that's when I reenabled
it.

In fact, the 4.1 snapshots have had *NO* regressions here at all
(remember, YMMV, I'm just reporting /my/ experience to date), and seem to
finally achieve the promise of the 4.x series rewrite -- a good portion of
all the theoretical improvements that the rewrite was supposed to bring,
/way/ better than 4.0, without /any/ of the troubles 4.0 naturally had as
the first in the series. Of course part of that is that the tree is now
better prepared for the 4.x series, but still, the 4.1 snapshots have been
if anything more stable than 4.0.2 is even now. It's really eerie just
/how/ stable, to the point I'm almost crossing my fingers as I'm expecting
it to all come down around me at once, with a kernel miscompile that
screws my filesystem or something equally drastic!

Back to xorg-modular and this -fweb bug.  As it doesn't seem to apply to
us, based on the thread, I /would/ like to get that filterflags put in an
x86 or at  least !amd64 test.  If I file a bug toward that end, no doubt
the Gentoo xorg folks will consult with you, so I might as well ask what
your answer will be before cloning that bug and asking for such a test. 
Or, if you agree, perhaps it'd be better if you handled it, arch-team to
x-team, since the first thing the x-folks would do would be to contact you
anyway?

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman in
http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html


-- 
gentoo-amd64@g.o mailing list