2000-11-21 05:49:35

by Robert Love

[permalink] [raw]
Subject: [WEIRD] working kernel off RH7's gcc-2.96!?

i dont want to revisit the flame fest (at all, please) but it seems i have been
using a kernel that successfully compiled under RedHat 7's gcc snapshot (2.96).
i normally use gcc-2.91.66 for everything (mv kgcc gcc) but just synced my
system with rawhide, so the gcc/kgcc pair is back on my system and i forgot. so
i recompiled to test11 yesterday, and:

[00:05:51]rml@phantasy:~# cat /proc/version
Linux version 2.4.0-test11 (rml@phantasy) (gcc version 2.96 20000731 (Red Hat
Linux 7.0)) #1 Mon Nov 20 19:06:06 EST 2000

the odd thing is, not only did it compile, but my machine has been up for a day
with heavy use in X with a full-featured kernel! not only no OOPSs, but no bugs!

the only thing i thought of was that alan's kgcc detection routine was in-place,
and using the detected kgcc to compile but then doing a hardcoded use of "gcc"
to grab the version ... but i could not find the code by grepping.

so i went ahead and removed kgcc from my path and recompiled, and sure enough --
it compiled fine.

just fyi, but someone tell me i am on crack... regardless, i am about to reboot
into a gcc-2.91.66 kernel.

--
Robert M. Love
[email protected]
[email protected]


2000-11-21 15:27:36

by Jes Sorensen

[permalink] [raw]
Subject: Re: [WEIRD] working kernel off RH7's gcc-2.96!?

>>>>> "Robert" == rml <[email protected]> writes:

Robert> i dont want to revisit the flame fest (at all, please) but it
Robert> seems i have been using a kernel that successfully compiled
Robert> under RedHat 7's gcc snapshot (2.96). i normally use
Robert> gcc-2.91.66 for everything (mv kgcc gcc) but just synced my
Robert> system with rawhide, so the gcc/kgcc pair is back on my system
Robert> and i forgot. so i recompiled to test11 yesterday, and:

[snip]

Robert> the odd thing is, not only did it compile, but my machine has
Robert> been up for a day with heavy use in X with a full-featured
Robert> kernel! not only no OOPSs, but no bugs!

Just keep in mind that 'seems to run fine' doesn't clearly show cases
like something in the signal handling code got miscompiled for a
special case or a bit error in the file system code. You could be
lucky, it might also show itself as more frequent crashes later etc.

Jes