Sage 9.3

[edgarcosta]

LMFDB makes Sage 9.3 segfault on OSX when calling sage.libs.pari.convert_gmp.new_gen_from_mpz_t, i.e., converting from pari's data type to gmp's data type at the C level. Here is a traceback: https://gist.github.com/alexjbest/4058fcc8acee924142b46540bbe2c33d

@roed314 and @alexjbest observe the exact same issue.

Has anyone else observed this before?
@roed314 and I are trying to figure out the reason, but we are a bit afraid that it has to do with threading from werkzeurg and pari not being built threadsafe

[edgarcosta]

Here is perhaps a minimal example:

from concurrent.futures import ThreadPoolExecutor
with ThreadPoolExecutor() as e:
    j = e.submit(is_squarefree, 15)
j.result()

[roed314]

I asked on sage-devel: https://groups.google.com/g/sage-devel/c/NmYojQqWzII

[roed314]

And we've created a cypari2 issue.

[roed314]

Sage 9.2 doesn't currently build for me because a failure in fpylll, but I was able to get a working version by using a beta version. If you're looking for a version of Sage that will run the LMFDB, I would suggest Sage 9.3.beta5 (the last before the cypari upgrade). If you're on OS X using Homebrew, you may also need https://trac.sagemath.org/ticket/31127 (which was merged in 9.3.beta6, so you'll have to include it on your own).

[roed314]

Fixed by #4603