Re: sparc64 vs. gcc 4.8

[BERTRAND Joël <joel.bertrand%systella.fr@localhost>]

Michael a écrit :
> Hello,

	Hello,

> I didn't really notice the switch since all my build scripts had
> HAVE_GCC=45 left over from the previous round of updates. Since
> removing them I ran into a bunch of problems. Userland seems to be
> fine, kernels not so much:
> - recent SMP kernels on my U60, built with gcc 4.5 are rock stable. The
>    same built with gcc 4.8 ( same hardware, same source tree ) will
>    spontaneously reboot ( with nothing in the message buffer so no
>    panic message ), virtual console switching won't work right ( console
>    just freezes, no updates occur ) and X won't start. I didn't get
>    around to investigate in detail but as I said, each and any of these
>    issues go away by switching compilers and rebuilding the kernel.

	I can't test on my U60 as its power supply is dying. But I have notice 
some strange reboot (fast data mmu...) on USIII (Blade2k). As I have 
four Blade2k, I have checked that, with the same 7.99.1 kernel built 
with the same gcc (4.8), USIII randmly reboots and USIII Cu seems to be 
rock solid.

	There are a lot of different USII's silicon with a lot of erratas. Can 
you indicate your module P/N ? My U2E seems to be stable with dual 300 
MHz USII.

> - similar problems seem to exist on the Blade 2500, although it's more
>    complicated by other issues that crept in lately. I've seen the
>    machine do both deadlocks and spontaneous reboots, still trying to
>    figure out which is triggered by what.
>
> No idea if any of this happens on UP machines, the only one I've got is
> off limits for this kind of experiments.

	Regards,

	JKB