I apologize for the vagueness of the message, but for all ye TiBook users,
over the last couple months of kernels, I've noticed gcc (various versions
in the 3.0 series randomly), non-deterministically crashing on large builds.
The builds tend to be fine and complete immediately after a reboot.
I've replaced my RAM recently, and the problem happened before and after
the replacement so I don't *think* it's the RAM.
Has anyone seen this sort of weird corruption behavior? I don't know
where or how to start debugging this. Could be anything... bad drivers,
bad builds of gcc.. Any ideas? (and if you suggest d/ling a stock
compiler, instructions for doing this in gentoo would be appreciated ;-) )
-David
Hi,
On 13. Jan 2005, at 19:54 Uhr, David Eger wrote:
> I apologize for the vagueness of the message, but for all ye TiBook
> users,
> over the last couple months of kernels, I've noticed gcc (various
> versions
> in the 3.0 series randomly), non-deterministically crashing on large
> builds.
>
> The builds tend to be fine and complete immediately after a reboot.
> I've replaced my RAM recently, and the problem happened before and
> after
> the replacement so I don't *think* it's the RAM.
>
> Has anyone seen this sort of weird corruption behavior? I don't know
> where or how to start debugging this. Could be anything... bad
> drivers,
> bad builds of gcc.. Any ideas? (and if you suggest d/ling a stock
> compiler, instructions for doing this in gentoo would be appreciated
> ;-) )
My T2 (http://www.t2-project.org) PowerPC systems tend to be rock solid - I
only once managed to get random memory corruption when I tried out
PREEMPTION. BenH mentioned PREEMPTION and ReiserFS (I use) might not
play that well - at least not on PowerPC.
I had yet no time to review the affected code myself - however without
PREEMPTION all is well.
Yours,
--
Ren? Rebe - Rubensstr. 64 - 12157 Berlin (Europe / Germany)
http://www.exactcode.de | http://www.exactcode.de/t2
+49 (0)30 255 897 45
I do not run said machine, but I did have GCC crashes under similar large
builds (notably the kernel) on one of my boxes.
It turned out the CPU was overheating and investigation revealed the heatsink
paste was dried out (it is an old box, circa 1998 since paste was applied).
As I only loaded the thing during GCC builds is why the problem revealed it's
self. A new heatsink and paste fixed it.
Nick
--
"When you're chewing on life's gristle,
Don't grumble, Give a whistle..."
On Thursday 13 January 2005 20:54, David Eger wrote:
> I apologize for the vagueness of the message, but for all ye TiBook users,
> over the last couple months of kernels, I've noticed gcc (various versions
> in the 3.0 series randomly), non-deterministically crashing on large builds.
gcc is a very good hardware checker. :)
I had a gcc crashing problem. Turned out to be RAM
timing problem. memtest86 wasn't able to trigger it
for me in overnight runs. gcc does it in seconds -
if you find a "magic" set of compile options and source file.
As you correctly noticed, yu have more chances of hitting
magic in large compiles.
I grabbed crashed gcc's command line from make log
and ran it in endless loop.
I had a crash (SIGSEGV due to mangled pointer) with ~10%
probability.
--
vda
On Sun, Jan 16, 2005 at 01:34:42AM +0100, Ren? Rebe wrote:
> On 16. Jan 2005, at 1:07 Uhr, David Eger wrote:
>
> >>BenH mentioned PREEMPTION and ReiserFS (I use) might not
> >>play that well - at least not on PowerPC.
> >
> >Interesting. I do have Pre-emption on (and I have Reiser compiled,
> >but not in use). Sounds like time to kill it.
>
> Could you reply if you have some recompiled kernel results?
Recompiling my kernel without PREEMPT seems to have fixed my problem...
I haven't had a gcc crash badly since...
-dte