[Pkg-utopia-maintainers] Bug#827473: ostree: FTBFS on mipsel: "ostree pull" sometimes gets SIGBUS, SIGSEGV

Simon McVittie smcv at debian.org
Wed Jun 29 13:12:58 UTC 2016


On Wed, 29 Jun 2016 at 12:30:40 +0000, Radovan Birdic wrote:
> I tried to build ostree package on three different machines for mipsel
> architecture: Cavium, Loongson and Broadcom.

Thanks, much appreciated!

What specific CPU/sub-architecture/whatever are these machines? I'm
assuming that saying you have a Broadcom mipsel CPU is like saying you
have an AMD x86 CPU, rather than like saying it's a specific model of
Athlon or whatever?

I've seen "ostree pull" failing on the Debian mipsel porterbox
etler.debian.org, which has /proc/cpuinfo like this:

system type		: loongson-ls3a-rs780e-1w
machine			: Unknown
processor		: 0
cpu model		: ICT Loongson-3 V0.5  FPU V0.1
BogoMIPS		: 718.84
wait instruction	: no
microsecond timers	: yes
tlb_entries		: 64
extra interrupt vector	: no
hardware watchpoint	: yes, count: 0, address/irw mask: []
isa			: mips1 mips2 mips3 mips4 mips5 mips32r1 mips64r1
ASEs implemented	:
shadow register sets	: 1
kscratch registers	: 0
package			: 0
core			: 0
VCED exceptions		: not available
VCEI exceptions		: not available

(and three more identical cores)

The official buildd where this has been failing is eberlin.debian.org,
which is described as "LS3A-RS780-1w (Quad Core Loongson 3A)" on
<https://db.debian.org/machines.cgi?host=eberlin>. Ordinary developers
can't log in to buildd machines, but the description on
<https://db.debian.org/machines.cgi?host=etler> also says LS3A-RS780-1w,
so eberlin's /proc/cpuinfo would probably look the same.

> On Broadcom machine, following tests fails:
> test-basic.sh - on every run
> > ERROR: tests/test-basic.sh - too few tests run (expected 57, got 55)
> > ERROR: tests/test-basic.sh - exited with status 1
> test-pull-c - occasionally
> > ERROR: tests/test-pull-c - too few tests run (expected 2, got 1)
> > ERROR: tests/test-pull-c - exited with status 138 (terminated by signal 10?)

Those errors are not enough information to be useful or diagnostic. Please
could you attach the detailed logs from a failing run? That either means
test-suite.log, or the individual test logs from the failing tests
(tests/test-basic.sh.log and tests/test-pull-c.log in this case).

If you run the tests with VERBOSE=1 in the environment, they'll write
test-suite.log to stdout/stderr, somewhat later than the ERROR lines
you quoted. That's what appears if the tests fail during a package build,
or in the official Debian buildd logs like
<https://buildd.debian.org/status/fetch.php?pkg=ostree&arch=mipsel&ver=2016.6-2&stamp=1467155890>
(search for ".. contents::" to find logs similar to the ones I'd
need to see).

(This is standard Automake behaviour, not OSTree-specific.)

If you can get anything useful from a core dump - for instance
a useful backtrace - that would also be really helpful.

Thanks,
    S



More information about the Pkg-utopia-maintainers mailing list