You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Inconsistently, and at "random" points during testing, I experience segmentation faults on Python 3.7.0. It happens no matter the OS (Mac, Windows, Linux). I strongly believe that this is not a bug in fastnumbers but rather with Python itself. My logic for this conclusion is
There is an open issue on bugs.python.org (https://bugs.python.org/issue34087, called "int(s), float(s) and others may cause segmentation fault") that points out that a change to the unicode handling internal function _PyUnicode_TransformDecimalAndSpaceToASCII that is called by int and float can introduce segfaults.
There is no difference in code within fastnumbers between Python 3.6 and 3.7, and no segfaults occur on Python 3.6.
I will leave this issue open till Python 3.7.1 is released (which has a patch for the segfault) and I can verify that this resolves the issue.
In interesting coincidence - it is the fact that _PyUnicode_TransformDecimalAndSpaceToASCII changed in 3.7 that started PR #16, but it was because its behavior changed, not because of segfaults.
The text was updated successfully, but these errors were encountered:
Inconsistently, and at "random" points during testing, I experience segmentation faults on Python 3.7.0. It happens no matter the OS (Mac, Windows, Linux). I strongly believe that this is not a bug in
fastnumbers
but rather with Python itself. My logic for this conclusion is_PyUnicode_TransformDecimalAndSpaceToASCII
that is called byint
andfloat
can introduce segfaults.fastnumbers
between Python 3.6 and 3.7, and no segfaults occur on Python 3.6.I will leave this issue open till Python 3.7.1 is released (which has a patch for the segfault) and I can verify that this resolves the issue.
In interesting coincidence - it is the fact that
_PyUnicode_TransformDecimalAndSpaceToASCII
changed in 3.7 that started PR #16, but it was because its behavior changed, not because of segfaults.The text was updated successfully, but these errors were encountered: