YJIT: all guard methods should update context's type info

[noahgibbs]

Right now we have various guard_object_is_array-type methods that do not update the context. Need to review all guard methods to make sure that if we do a guard, the context gets updated.

[maximecb]

Link to code for said guard methods: https://github.com/ruby/ruby/blob/9e067df76bda7384e8f3ddda2b348f3a742eb784/yjit/src/codegen.rs#L1295

And here, guard_two_fixnums is an example of another guard that does upgrade the types in the context: https://github.com/ruby/ruby/blob/9e067df76bda7384e8f3ddda2b348f3a742eb784/yjit/src/codegen.rs#L2286

[noahgibbs]

Previous, incomplete work: ruby#6797

[maximecb]

Previous, incomplete work: ruby#6797

Sorry Noah. I feel like I dropped the ball on this one. I wish there was an easy way to have our cake and eat it too (keep all the str_concat optimizations you made), particularly since we care so much about string concat performance. Maybe we can return to this in January, or find ways to merge some of the changes you worked on without touching str_concat.

[noahgibbs]

All good. You folks did great work on stability while I was away. This is a pretty small optimisation. But yeah, I could just add all the non-str_concat changes and leave a comment in str_concat saying why we don't do it the regular way. I'll plan on that, though after the 3.2.0 release. I'm OOO on the 20th/21st, and I don't think we want to sneak this in right before the deadline.

[noahgibbs]

Working on new draft PR: ruby#7056

[maximecb]

Since the last time it was difficult to get everything to pass checks with one larger PR, I'm trying to merge the missing bits from our last attempt in this PR: ruby#7489

[maximecb]

Looks like we've completed the work for this issue ✅ 🙂