Message238827
| Author |
vstinner |
| Recipients |
berker.peksag, pitrou, python-dev, serhiy.storchaka, vstinner |
| Date |
2015年03月21日.19:24:05 |
| SpamBayes Score |
-1.0 |
| Marked as misclassified |
Yes |
| Message-id |
<CAMpsgwZz2RpjqkJcqiRZjCQk66vRY1qdhoyVFVcnTf_q39FEtQ@mail.gmail.com> |
| In-reply-to |
<1426957549.28.0.411172634494.issue23571@psf.upfronthosting.co.za> |
| Content |
Le 21 mars 2015 18:05, "Serhiy Storchaka" <report@bugs.python.org> a écrit :
> But an assertion itself provides less information than an exception.
Debug build is less informative than release build.
I like Antoine's idea to replace the assertion with Py_FatalError() in
debug mode.
I'm more concerned by bugs in Python itself. A fatal error/assertion should
be noticed quickly on buildbots which compile python in debug mode.
It's just fine if other people use the release mode and get an exception.
> May be add a runtime flag to control the reaction on system errors? If it
set, all raised SystemError will be converted to fatal errors.
It's already what I do when running Python test suite with pyfailmalloc. I
modify manually SystemError constructor in the C code. A flag may help, but
you have to know that many tests expect SystemError (I don't remember which
ones).
Victor |
|