gh-88057: in compile.c, assertion that stackdepth is alway >=0 is missing in one place

[iritkatriel]

I believe this assertion would have helped @larryhastings with the crash described in #88057.

• Issue: co_annotations branch caused a crash in stackeffect() in compile.c #88057

[larryhastings]

Hmm. Maybe I jumped the gun on approving the merge.

The thing is, in isolation the change is fine. But the code seems a bit weird. Why is there

assert(depth >= 0); /* invalid code or bug in stackdepth() */

on line 7081 in the original source (7082 in the changed source)? depth hasn't changed in the loop yet. Having the assertion here means we only assert the previous change in depth was okay when running the loop--and we don't run the assertion at all after the last instruction.

Without your one-line change, that assertion should really move below depth = new_depth; on line 7092. With your change, I think we can remove that assertion entirely.

I'd also like to see the assertion on target_depth moved up to immediately after the assignment. It's harmless, but pointless, to conditionally set max_depth before doing that check.

[iritkatriel]

Yes, exactly. The assertion on line 7082 was supposed to capture what my new assertion captures, and it does except for the last iteration. We could remove it now, though it's not wrong.