gh-67877: Fix memory leaks in terminated RE matching

[serhiy-storchaka]

If SRE(match) function terminates abruptly, either because of a signal or because memory allocation fails, allocated SRE_REPEAT blocks might be never released.

• Issue: Memory leak in Modules/sre_lib.h #67877

[serhiy-storchaka]

See also #126843.

[wjssz]

Looks good to me.
You may also remove the outdated SRE_ERROR_RECURSION_LIMIT at the same time, or remove it later.

[serhiy-storchaka]

Setting state->fail_after_count does not affect subsequent calls of the Pattern methods, but it still affects the finditer() iterator which reuses state between iterations.

[wjssz]

I deleted my previous reply.
Maybe setting state->fail_after_count = -1 doesn't make sense, it's better to make the finditer() iterator stop the subsequent matchings.

[wjssz]

I remember another thing, if this patch is going to be backported, you may remove the changes related to this:

-    PyObject* pattern;	
+    PatternObject* pattern;

The patch will be smaller and clearer.
Polishing work can be done only on the main branch, such as change the type, remove the outdated SRE_ERROR_RECURSION_LIMIT.

[miss-islington-app]

Thanks @serhiy-storchaka for the PR 🌮🎉.. I'm working now to backport this PR to: 3.12, 3.13.
🐍🍒⛏🤖

[miss-islington-app]

Sorry, @serhiy-storchaka, I could not cleanly backport this to 3.13 due to a conflict.
Please backport using cherry_picker on command line.

cherry_picker 7538e7f5696408fa0aa02fce8a413a7dfac76a04 3.13

[miss-islington-app]

Sorry, @serhiy-storchaka, I could not cleanly backport this to 3.12 due to a conflict.
Please backport using cherry_picker on command line.

cherry_picker 7538e7f5696408fa0aa02fce8a413a7dfac76a04 3.12

[bedevere-app]

GH-126960 is a backport of this pull request to the 3.13 branch.

[bedevere-app]

GH-126961 is a backport of this pull request to the 3.12 branch.

[serhiy-storchaka]

I remember another thing, if this patch is going to be backported, you may remove the changes related to this:

It is easier to not do. I can introduce new bugs when trying to remove these changes in backports.

Initially it was needed to pass the second argument to state_fini, after reverting that change it is just a clean up, but it can still be helpful if in future we will need to pass the second argument to state_fini.

[wjssz]

sre only has the “\B can't match empty string” bug as far as I know. It's easy to fix, if decide to fix it.
Maybe people won't touch sre's code for many years.