GH-91079: Implement C stack limits using addresses, not counters.#130007
Conversation
…m independent constants
encukou
left a comment
There was a problem hiding this comment.
Looks great now, thank you!
Would you be OK with me asking the SC to remove Py_C_RECURSION_LIMIT (and c_recursion_remaining) without deprecation? IMO it's the right thing to do here.
|
|
@markshannon if you look at line 427 of https://buildbot.python.org/#/builders/1046/builds/7562/steps/8/logs/stdio you will see that |
|
I opened #130319 and assigned it to @markshannon . |
|
This also appears to have caused a failure on the iOS buildbot (masked because there was an existing failure with the colorise tests). However, this build, triggered by the merge of this PR is the first one that has a failure in test_sys |
|
@markshannon Looking into the iOS failure - the comment before the test that is failing says: What constitutes "normal" here? Is it possible a sequential, single process run (i.e., how iOS runs the test suite) might trigger this? I'm a bit out my depth here on diagnosing the cause of this one, so any pointers or suggestions for possible causes would be most welcome. |
|
This PR does no refcounting, allocation or deallocation. |
… counters. (pythonGH-130007)" for now Unfortunatlely, the change broke some buildbots. This reverts commit 2498c22.
…its using addresses, not counters. (pythonGH-130007)" for now (GH130413)" This reverts commit ef29104.

Hopefully, this will also fix #113655.