Re: Release sprint results - team changes, auto-rm and arch status
On Fri, Nov 29, 2013 at 01:57:39PM +0000, Ben Hutchings wrote:
> On Fri, 2013-11-29 at 09:22 +0100, Aurelien Jarno wrote:
> > Some precision about the MIPS machines:
> >
> > On Thu, Nov 28, 2013 at 09:04:56PM +0100, Niels Thykier wrote:
> > > * [mips, mipsel] buildds/porterboxes run on hardware which is very old or has known defects:
> > > - mips octeon is unstable
> >
> > Better me more precise there, Octeon machines in general are very
> > stable. That said out of the three machines we have in Debian, two are
> > unstable, the other one is very stable. We never really understood why,
> > we only have remarked they have different CPU revision number.
>
> Whatever, mips has one reliable buildd which is not enough.
Two actually with ball.d.o which is not an octeon.
My point is to say that Octeon boxes are not unstable, but the Octeon
boxes *we have* as part of the Debian infrastructure are unstable. This
of course has to be fixed.
> > > - mipsel loongson have CPU bugs
> >
> > I see on http://release.debian.org/jessie/arch_qualify.html that it
> > concerns the "NOP implementation error" from Loongson 2F. Debian is
> > using Loongson 2E for buildds and porters machines, which are not
> > affected by this issue.
>
> They are affected by that or a very similar issue, as demonstrated by Jo
> Shields recently: http://apebox.org/wordpress/rants/545/
>
MUL is a MIPS32 instruction, which is not present on MIPS3 CPUs like the
Loongson 2, MULT + MFLO should be used instead. There is no CPU bug
there, it's like trying to build x86 code with SSE4 instructions, and
then saying that all x86 CPUs which do not support the SSE4 instructions
are buggy.
--
Aurelien Jarno GPG: 1024D/F1BCDB73
aurelien@aurel32.net http://www.aurel32.net
Reply to: