gh-114828: parenthesize non-atomic macro definitions in pycore_symtable.h

[carljm]

In #115139, I used the expression ~DEF_FREE as a mask to clear the DEF_FREE bit in a symbol table entry.

Unfortunately, DEF_FREE is defined as #define DEF_FREE 2<<4, so this macro-expands to ~2<<4, and the bitwise negation binds more tightly than the bit-shift, so this actually meant (~2)<<4 rather than the ~(2<<4) which I intended.

This is why non-atomic macro bodies should always be parenthesized!

Fix it by parenthesizing all the non-atomic definitions in pycore_symtable.h, which not only fixes my ~DEF_FREE, but also prevents anyone else making the same mistake in future.

• Issue: Python/flowgraph.c:528: _Bool all_exits_have_lineno(basicblock *): Assertion `0' failed #114828

[JelleZijlstra]

Thanks for figuring this out!

[JelleZijlstra]

It might be worth backporting this to 3.11, though there will be some conflicts and it's unlikely that we'll make a lot of changes to the 3.11 symtable at this point. Up to you.

[AlexWaygood]

Thanks!

[carljm]

Hmm. I think my (weak) inclination is not to backport to 3.11, given that AFAIK there were major symtable changes in 3.12 but not in 3.11, so it seems unlikely we'll be backporting many changes to 3.11, and even less likely that any such changes would be affected by this. This PR is actually a bugfix for 3.12 (given that my ~DEF_FREE exists there), but for 3.11 it would be purely preventative. So I'd err on the side of being conservative with backports.

But like I say, this is a weak preference, and I'll happily backport to 3.11 if anyone feels strongly that that's the thing to do.

[miss-islington-app]

Thanks @carljm for the PR 🌮🎉.. I'm working now to backport this PR to: 3.12.
🐍🍒⛏🤖

[bedevere-app]

GH-115149 is a backport of this pull request to the 3.12 branch.