Re: hppa in danger of being ignored for testing migration and eventual removal
On Wed, Apr 29, 2009 at 12:57:28AM +0200, Thibaut VARENE wrote:
> On Tue, Apr 28, 2009 at 11:34 PM, dann frazier <email@example.com> wrote:
> > On Tue, Apr 28, 2009 at 05:09:25PM -0400, Carlos O'Donell wrote:
> >> On Tue, Apr 28, 2009 at 4:12 PM, Luk Claes <firstname.lastname@example.org> wrote:
> >> > * The machines that host the buildds still seem to have a very
> >> > unreliable kernel. Is there any update on this?
> >> I can't comment on this.
> > Thibaut had planned to setup a second buildd (and I think had it up
> > for a while?) but that box experienced a hardware failure. We're also
> > working on moving one of the buildds to a different platform
> > (rp2470). We have no specific reason to believe that will be any
> > better, but its worth a shot.
> Yes the supplementary buildd had a major hardware failure today
> (SBA/LBA failures during last POST and now the GSP can't load PDC
> anymore. The machine is basically dead). I'm switching the hard drives
> to another box I have that I used elsewhere (not yet back online as
> the Debian kernel seems to have major trouble coping with PCI addon
> NICs - tulip and tg3 HPMC the machine on driver load), and I'm still
> hoping to get feedback on some hardware donation requests I've made a
> month or so ago.
> I hereby take the opportunity to say that I would gladly welcome any
> rackable parisc system in my server room. :-)
> >> I run stock: linux-image-2.6.26-1-parisc64-smp (2.6.26-13)
> >> on my SMP 2x PA8700 system without any problems.
> > There are several reports of stability on various mixtures of
> > kernel/platform - and the non-buildd debian.org hppa machine seems to
> > be quite stable as well. But, once we start running a buildd on
> > something, instability issues abound.
> The only issue I've been aware of so far was the ruby build problem.
> If there are others, they need more publicity I think. OTOH, ISTR
> Carlos said most of the problems could go away with the transition to
> NPTL. Might be worth a try...
I hand-built an NPTL environment before and it didn't help, but maybe
something has gotten fixed since. But this certainly isn't the only
stability issue - both buildds regulary hang hard and need
rebooting. The hangs are fairly random - there doesn't appear to be
anything consistent about them (not building the same package, running
the same command, etc).
> >> > * The debian-installer dailies that are now built again, but seem to
> >> > fail to build most of the time. Is there any particular reason for this?
> >> No idea. Do you have a log?
Problems all appear to be w/ dpkg.
build_netboot.log has this:
Unpacking console-keymaps-at (from udebs/console-keymaps-at.udeb) ...
*** glibc detected *** dpkg: corrupted double-linked list: 0x00100c58
dpkg: error processing udebs/dhcp3-client-udeb.udeb (--unpack):
subprocess dpkg-split killed by signal (Aborted)
build_miniiso.log has this:
Unpacking busybox-udeb (from udebs/busybox-udeb.udeb) ...
dpkg: error processing udebs/cdebconf-newt-terminal.udeb (--unpack):
subprocess dpkg-split killed by signal (Segmentation fault)
Errors were encountered while processing:
make: *** [stamps/tree-unpack-miniiso-stamp] Error 1
make: *** [_build] Error 2
make: *** [build_miniiso] Error 2
> I would blame recent failures on failure to build kernel, maybe?
> AFAICT there was the phonet issue (fixed since then) and it seems
> recent kernel builds fail to link the btrfs module...
These have been fixed in svn and an upload is scheduled for tomorrow.