On Wed, 2004-02-18 at 19:33, david daney wrote:
In message <1077159735.17315.136.camel@dzur.sfbay.redhat.com>, Eric Christopher
>writes:
>>
>>> *looks* ... no, on reflection, as long as we have to keep this
>>> bletcherous feature around, we might as well make sure it works.
>>> Go ahead and commit your original patch.
>>
>>Other than all the pch regressions. I'd prefer he figure those out
>>before committing. :)
Is it possible that David is working on a system with exec-shield and
memory layout randomization? In the past when I've seen PCH stuff
randomly pass/fail it's always been solved by turning off the exec-shield
and related bits.
It is a cross compiler running of Fedora 1. This may have exec-shield, I am not sure.
It does and very likely is causing the problems given:
Target is plain Linux 2.4.25-pre6 from linux-mips.org w/ glibc 2.2.5 running on a MIPS 4Kc core. But this is not important because the tests that fail are all run on the compiler host.
I have the following in /etc/sysctl.conf
# Turning off exec shield.
kernel.exec-shield = 0
kernel.exec-shield-randomize = 0
If you turn off exec shield and rerun the tests an they pass go ahead
and commit. Otherwise I'd appreciate you looking into it.
Thanks.
-eric