Issue1733973
Created on 2007-06-09 08:04 by peaker, last changed 2022-04-11 14:56 by admin. This issue is now closed.
| Files | ||||
|---|---|---|---|---|
| File name | Uploaded | Description | Edit | |
| cProfile.patch | peaker, 2007-06-09 08:04 | A patch to the most recent svn | ||
| cProfile.patch | peaker, 2007-06-09 08:30 | Removed redundant PyErr_Clear (already done by PyErr_Fetch) | ||
| Messages (4) | |||
|---|---|---|---|
| msg52738 - (view) | Author: Eyal Lotem (peaker) | Date: 2007-06-09 08:04 | |
_lsprof.c:ptrace_enter_call assumes PyErr_* has no errors. It assumes this by feeling free to call functions that PyErr_Clear and overwrite PyErr. This assumption is broken when throwing into a generator. In that case gen_send_ex(.., 1) calls the ptrace_enter_call with an exception already set. This assumption causes two bugs: A. The exception thrown into the generator is overwritten and lost. Later it was also cleared. B. gen_send_ex(..) still returns error-indication, but since the exception was cleared, you get a SystemError (error indication returned but no exception was set). To fix this, ptrace_enter_call now saves/restores the current PyErr. The attached patch was applied on the newest trunk svn. I guess it should be applicable without problems on the 2.5 bugfix branch as well. |
|||
| msg52739 - (view) | Author: Eyal Lotem (peaker) | Date: 2007-06-09 08:24 | |
It seems that bug #1733757 encountered the same bug. See comments on that bug. |
|||
| msg52740 - (view) | Author: Eyal Lotem (peaker) | Date: 2007-06-09 08:30 | |
File Added: cProfile.patch |
|||
| msg55681 - (view) | Author: Armin Rigo (arigo) *  |
Date: 2007-09-06 08:36 | |
Thanks for the patch. Checked in: * r58004 (trunk) * r58005 (release25-maint) |
|||
| History | |||
|---|---|---|---|
| Date | User | Action | Args |
| 2022-04-11 14:56:24 | admin | set | github: 45071 |
| 2007-09-06 08:36:01 | arigo | set | status: open -> closed resolution: accepted messages: + msg55681 nosy: + arigo |
| 2007-06-09 08:04:30 | peaker | create | |