Open zhaoli-IHEP opened 4 years ago
I don't know what causes this. Can you try it with valgrind?
Anyone else able to reproduce the problem?
With what versions of arb/flint should I try this? Could not reproduce with the latest of both.
With what versions of arb/flint should I try this? Could not reproduce with the latest of both.
flint version is 2.5.2 arb is the git latest version
I can say https://github.com/fredrik-johansson/arb/blob/master/partitions/test/t-partitions_fmpz_ui_threaded.c#L146 should probably be flint_cleanup_master and that @wbhart would have a better idea of the final specifications of the threading in flint. This is nothing to do the crash, though.
@tthsqe12 @wbhart Yes, I noticed that the threaded tests in Arb now leak a small amount of memory according valgrind, presumably because the thread pool is not cleaned up. Changing all instances of flint_cleanup() to flint_cleanup_master() would unforunately not be backward compatible with previous versions of flint, which for the time being might be desirable. I wonder if the new interface in flint could be tweaked in some way for better backward compatibility. Alternatively, I could perhaps replace flint_cleanup() with an arb_cleanup() that detects the version of flint and calls flint_cleanup() / flint_cleanup_master() (though as of the current git version of flint I don't know if there's a way to check for this).
"make" is okay but not for "make check". flint has successfully passed "make" and "make check". Here is the error message
gcc version is