[Debian-science-sagemath] sagemath build getting stuck on armel buildds

[Ximin Luo]

Tobias Hansen:
> On 02/21/2018 12:56 PM, Ximin Luo wrote:
>> Tobias Hansen:
>>> On 02/21/2018 11:37 AM, Julien Puydt wrote:
>>>> Hi,
>>>>
>>>> Le 21/02/2018 à 11:25, Tobias Hansen a écrit :
>>>>
>>>>> the build of sagemath regularly gets stuck on armel buildds. It stays in building status for weeks, probably until someone stops it. The successful builds on armel took about 12 hours. Any idea how to find the cause of this? Should we try restricting armel builds to 2 threads? Or 1? Currently they seem to use four.
>>>> Is there a log to know at which point it gets stuck?
>>>>
>>>> Snark on #debian-science
>>> No, it seems the logs are not kept if it never finishes and is aborted manually (or whatever happens):
>>> https://buildd.debian.org/status/logs.php?pkg=sagemath&arch=armel
>>>
>>> Oh, and sagemath is not the only package getting stuck on armel. Right now guile-2.0 and guile-2.2 are building for 77 and 53 days and I have seen this happen for other packages on https://buildd.debian.org/status/architecture.php?a=armel&suite=sid
>>>
>> I hear from #debian-buildd that the build kills the buildd, not sure if logs contain anything obvious. Sledge (Steve McIntyre) has to manually restart the machine every now and again. Same issue with guile.
>>
>> I suggested building with DEB_BUILD_OPTIONS=parallel=1 but not yet heard back on whether that succeeds.
>>
>> X
>>
> The same thing happened with sagemath 8.1-5 and 8.1-6. Should we just give up on armel for now and remove it from sagemath's architecture list?
> 

I think we can just leave it, it will get AUTORMd anyways in a few days due to the missing build on mips64el. Then the next time it enters testing, it won't be blocked by lack-of-build on armel, because there would be no existing build.

The segfault on mips64el is reproducible on eller, I'll try to debug it in the next week or so. I probably don't have enough time to do this before sagemath gets AUTORMd from testing, but this will help the armel issue without having to add (IMO dirty) architecture-specific exceptions.

X

-- 
GPG: ed25519/56034877E1F87C35
GPG: rsa4096/1318EFAC5FBBDBCE
https://github.com/infinity0/pubkeys.git