GC/Boehm: Silence GC warnings about big allocations.

[yxhuvud]

These warnings can happen for example when repeatedly adding elements
to big arrays or hashes. Unfortunately these warnings are not very
helpful and usually just irritating.

Fixes #11274

[straight-shoota]

Can you please add a comment that explains the reasoning directly in the code?

[asterite]

Maybe add a comment explaining why we are doing this?

[sdogruyol]

Thank you @yxhuvud 🙏

[luislavena]

Hello folks,

I would like to add two issues that were connected to this:

• Request: show backtrace for GC allocation warnings when using debug builds #2110
• GC Warning: Repeated allocation of very large block #2104

Merging this one will close those two.

Cheers.

[yxhuvud]

Hmm. #11324 seems to be getting warnings like GC Warning: Finalization cycle involving 0x7f6d1fa9c900.

This make me think we should only swallow the repeated allocation warning, and not all other possible warnings. In this case it looks to be an issue about some miswritten finalizer. That seems a lot more helpful than not warning at all.

Thoughts?

[yxhuvud]

Thoughts, would you prefer the check for the above to use a regexp to match against /Repeated allocation/ or would you prefer to compare against the full message, ie GC Warning: Repeated allocation of very large block (appr. size %ld):\n\tMay lead to memory leak and poor performance\n". The latter seems more prone to break on minor changes to me. It shouldn't happen often enough to be a performance impacting choice in any case.

[Sija]

Maybe some middle ground (without using regex):

msg.starts_with?("GC Warning: Repeated allocation of very large block")

[yxhuvud]

Opinions? It felt cleaner to use printf than to rewrite the format string, but 🤷‍♂️