Upgrading Sage from 9.2 to 9.3 breaks the LMFDB <http://www.lmfdb.org>
(segfaults on startup). We believe that the problem arises from an
interaction between Pari and threads, and were able to duplicate it in Sage
with the following:
from concurrent.futures import ThreadPoolExecutor
sage: with ThreadPoolExecutor() as e:
....: j = e.submit(is_squarefree, 15)
....: j.result()
...
~/sage/sage-9.4.beta0/local/lib/python3.9/site-packages/sage/rings/integer.pyx
in sage.rings.integer.Integer.is_squarefree
(build/cythonized/sage/rings/integer.c:37634)()
6007 False
6008 """
-> 6009 return self.__pari__().issquarefree()
6010
6011 cpdef __pari__(self):
~/sage/sage-9.4.beta0/local/lib/python3.9/site-packages/sage/rings/integer.pyx
in sage.rings.integer.Integer.__pari__
(build/cythonized/sage/rings/integer.c:37754)()
6029
6030 """
-> 6031 return new_gen_from_mpz_t(self.value)
6032
6033 def _interface_init_(self, I=None):
~/sage/sage-9.4.beta0/local/lib/python3.9/site-packages/sage/libs/pari/convert_gmp.pyx
in sage.libs.pari.convert_gmp.new_gen_from_mpz_t
(build/cythonized/sage/libs/pari/convert_gmp.c:1298)()
50 True
51 """
---> 52 sig_on()
53 return new_gen(_new_GEN_from_mpz_t(value))
54
SignalError: Segmentation fault
Searching for cypari issues, #94
<https://github.com/sagemath/cypari2/issues/94> suggests that there may be
a problem with "relatively recent decision of maintaining pointers on the
PARI stack."
At this point, we're a little lost. We understand that sage overall is not
threadsafe. But we hope that cypari (and the parts of Sage that we use in
the LMFDB) can be, since it was in 9.2. Any tips on where to look next (or
fixes to suggest!) would be welcome.
David
--
You received this message because you are subscribed to the Google Groups
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to sage-devel+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/sage-devel/CAChs6_nAw%3D7B6yqjFL5A-V%2Bb5gjVUG_%3DipzqYHkUn8Vt0gvkYw%40mail.gmail.com.
