Works OK with UP kernel; best I can get with SMP is single-user mode
(still after lots of errors). Even then commands randomly (but often) fail
with fatal exception. Compiled fresh 2.2.17 kernel, still had problem.
Even recompiled bash from source (just guessing), still not working.
System is Tyan Tomcat IVD mobo, dual Pentium 200 (classic) CPU's, 512K
cache, 256M RAM, Diamond Fireport 40 SCSI.
Worked fine up through Redhat 6.1, and WinNT 4 earlier.
assigning to kernel
A quick guess would have something to do with flakey IOAPIC support in the old
pentiums. The MMX's were the first ones to have reliable IOAPIC support. Since
UP uses the chipset PIC, it sidesteps the IOAPIC on CPU.
What kernel rev worked before, what doesn't now?
First I ran the kernel that came with the distro, (so long ago now I forget -
2.2.16 I think). I grabbed a copy of 2.2.16 from kernel.org, built and
installed it with SMP support and had the exact same experience. Downloaded &
built 2.2.17 from kernel.org, same thing again.
I've downgraded to Redhat 6.2. The XFree that came with it is flakey under SMP;
random X crashes and console lockups, but rest of system was fine (BTW I did try
two different video cards - a Permedia2-based one and am S3-virge gbased one,
same thing with both cards). I installed the binary of XFree 4.01 from
xfree.org and it's been perfectly stable since, both with the distro's kernel
and with 2.2.17 which I am running now with SMP.
BTW, I had the same X problems with Mandrake 7.1 on the same system, again cured
by the binaries from xfree.org.
So in a nutshell, the same kernel will work or not work depending on the
version of Redhat, plus RH6.2 (and Mandrake 7.1 which looks to me to be based on
RH6.2) introduced X instability cured by installing XFree's own binaries. Looks
to me like a userspace problem.
Commit pushed to master at https://github.com/openshift/origin
Merge pull request #18857 from soltysh/issue18753
Automatic merge from submit-queue (batch tested with PRs 18835, 18857, 18641, 18656, 18837).
Warn when AuditFilePath is relative
Fixes #18753. I'll open an issue for 3.11 to turn change this entirely into a requirement.
David, you ain't gonna like it, so be warned :wink: but I couldn't figure out a better approach than the one in this PR, b/c we modify all the paths to be absolute before we reach validation. Lemme know what you think.