| Georg Brandl | b12fd63 | 2010-12-28 11:06:07 +0000 | [diff] [blame] | 1 | This directory only contains tests for outstanding bugs that cause the |
| 2 | interpreter to segfault. Ideally this directory should always be empty, but |
| 3 | sometimes it may not be easy to fix the underlying cause and the bug is deemed |
| 4 | too obscure to invest the effort. |
| Neal Norwitz | 8cc4ef5 | 2006-01-10 07:49:41 +0000 | [diff] [blame] | 5 | |
| 6 | Each test should fail when run from the command line: |
| 7 | |
| 8 | ./python Lib/test/crashers/weakref_in_del.py |
| 9 | |
| Georg Brandl | b12fd63 | 2010-12-28 11:06:07 +0000 | [diff] [blame] | 10 | Put as much info into a docstring or comments to help determine the cause of the |
| 11 | failure, as well as a bugs.python.org issue number if it exists. Particularly |
| 12 | note if the cause is system or environment dependent and what the variables are. |
| Neal Norwitz | 8cc4ef5 | 2006-01-10 07:49:41 +0000 | [diff] [blame] | 13 | |
| Georg Brandl | b12fd63 | 2010-12-28 11:06:07 +0000 | [diff] [blame] | 14 | Once the crash is fixed, the test case should be moved into an appropriate test |
| 15 | (even if it was originally from the test suite). This ensures the regression |
| 16 | doesn't happen again. And if it does, it should be easier to track down. |
| Nick Coghlan | c95ec3d | 2011-03-16 19:52:14 -0400 | [diff] [blame] | 17 | |
| 18 | Also see Lib/test_crashers.py which exercises the crashers in this directory. |
| 19 | In particular, make sure to add any new infinite loop crashers to the black |
| 20 | list so it doesn't try to run them. |