[Debichem-devel] Bug#666110: Bug#666110: mpqc: FTBFS on armel (tests hang?)

[Michael Banck]

reassign 666110 openmpi-bin
thanks

On Wed, Mar 28, 2012 at 09:31:36PM +0200, Julien Cristau wrote:
> Source: mpqc
> Version: 2.3.1-14
> Severity: serious
> Justification: fails to build from source (but built successfully in the past)
 
Well, it built successfully cause openmpi was just crashing/not
executing the test suite in the past, see #494046.

> See the build log at
> https://buildd.debian.org/status/fetch.php?pkg=mpqc&arch=armel&ver=2.3.1-14&stamp=1332702986

The runtime of that test is 0.05 seconds on my T400 on amd64, so it
seems to be a more general problem with openmpi on arm.  Indeed if I
extract the -13 .debs and set LD_LIBRARY_PATH, mpqc spins at 100%
without output.  Even if I just run /usr/bin/ompi-ps without arguments,
it eats 100% cpu and spins at sched_yield() forever accordings to
strace:

socket(PF_INET, SOCK_STREAM, IPPROTO_IP) = 7
setsockopt(7, SOL_TCP, TCP_NODELAY, [1], 4) = 0
setsockopt(7, SOL_SOCKET, SO_SNDBUF, [131072], 4) = 0
setsockopt(7, SOL_SOCKET, SO_RCVBUF, [131072], 4) = 0
fcntl64(7, F_GETFL)                     = 0x2 (flags O_RDWR)
fcntl64(7, F_SETFL, O_RDWR|O_NONBLOCK)  = 0
connect(7, {sa_family=AF_INET, sin_port=htons(37855),
sin_addr=inet_addr("172.28.17.11")}, 16) = -1 EINPROGRESS (Operation now
in progress)
sched_yield()                           = 0
sched_yield()                           = 0
sched_yield()                           = 0
sched_yield()                           = 0
sched_yield()                           = 0
sched_yield()                           = 0
[...]

Reassigning to openmpi-bin for now, I am of course open to suggestions
for any work-arounds or other things to try.


Thanks,

Michael