Bug#906796: [Debian-science-sagemath] Bug#906796: sagemath: Sage crashes at startup (segfault in pari modules)

[Doug Torrance]

On 08/21/2018 05:41 AM, Doug Torrance wrote:
> On 08/21/2018 04:13 AM, Tobias Hansen wrote:
>> Control: block -1 by 905434 905300
>>
>> This is expected after a major pari update. sagemath 8.3 is almost 
>> ready to be uploaded, major things we have to do:
>>
>> - fix build failures in fpylll and gfan
>> - fix at least the 24 failing sagemath doctests in 
>> sage/repl/ipython_kernel/install.py [1]
>>
>> If gfan 0.6.2 can't be fixed we could still use version 0.5...
> 
> I forwarded the gfan build failure upstream a few weeks ago but haven't 
> gotten a response yet.  I've tried debugging it on my own but haven't 
> figured anything out yet.
> 
> The build failure is from a test failing on 32-bit architectures.  We 
> could always just skip that test on those architectures for a short-term 
> fix.  I can work on that later.

I've pushed some commits to git doing the above [1].  Would anyone be 
able to take a look and sponsor?

A quick rundown of what I did: (Perhaps there's a better way?)

gfan gets the list of all the tests to run from the subdirectories in 
the testsuites directory.  Using dpkg-architecture, we remove the 
directories of the two failing tests only on 32-bit.

We were already skipping a test which failed on 32-bit, but removed the 
directory using quilt.  That wouldn't work with the new test because one 
of the files in the directory was empty and quilt wouldn't deal with it. 
  Plus it would be nice to run these tests on 64-bit anyway.  So I just 
got rid of that patch and now we remove both directories in d/rules.

Doug

[1] https://salsa.debian.org/science-team/gfan