[Debian-science-sagemath] Nearly released; but ell_rational_field has lfunction/sympow failures

Ximin Luo infinity0 at debian.org
Wed Dec 14 07:46:00 UTC 2016


Jerome BENOIT:
> 
> 
> On 14/12/16 03:25, François Bissey wrote:
> 
>>> On 14/12/2016, at 16:23, Ximin Luo <infinity0 at debian.org> wrote:
>>>
>>> Ximin Luo:
>>>> I built 7.4-1 for Debian experimental, it failed with 89 test failures. However, I think I have fixed 12 of them, which takes it below our arbitrary "80 failed tests" limit and I am now rebuilding it. I will upload this to NEW in the morning.
>>>>
>>>> However, we do need to fix the following before we re-upload to unstable:
>>>>
>>>> 30+ of those test failures are from ell_rational_field.py, the direct failures in that file (17) are attached. It seems related to lfunction or sympow. 
>>>>
>>>> [..]
>>>
>>> Maaaybe this is because sympow craps out when it can't write to ~/.sympow. I've logged into the schroot session where sbuild is working and manually created this directory now, so `sympow` at least works. I'm going to bed, will report back with the results in the morning.
>>>
> 
>> Sounds likely. simpow will create a number of data files
>> and in debian and Gentoo we create them in ~/.sympow.
> 
> For Debian, sympow was patched to create data on the fly in such
> a way it is transparent for the final user [1].
> I have no idea if it was applied to Gentoo.
> 
> Jerome
> 
> [1] http://sources.debian.net/src/sympow/1.023-7/debian/patches/upstream-on_fly_new_data.patch/
> 

Thanks for the input guys. It looks like the non-existent (and non-writable) $HOME was indeed the reason for the failures.

But unfortunately my work-around failed because I did the wrong thing - I created /home/infinity0 inside the chroot but I should have instead done `su -c 'chown infinity0: /sbuild-nonexistent'` because that is what sbuild sets $HOME to - and it already exists because something creates /sbuild-nonexistent/.jupyter/migrated with root-owned permissions during sbuild's setup phase.

That is pretty annoying :/ We will be forced to fix it for unstable, because the buildd network uses sbuild.

Perhaps we can set HOME=$DOT_SAGE for all sympow child processes, in Sage?

X


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



More information about the Debian-science-sagemath mailing list