runtime: mapassign_fast64ptr crashes in malloc

[adonovan]

This stack cysh5A was reported by telemetry:

crash/crash
runtime.systemstack_switch:+4
runtime.persistentalloc:+2
runtime.newBucket:+11
runtime.stkbucket:+53
runtime.mProf_Malloc:+13
runtime.profilealloc:+6
runtime.mallocgc:+270
runtime.newarray:+8
runtime.makeBucketArray:+18
runtime.hashGrow:+10
runtime.mapassign_fast64ptr:+64
go/types.(*Checker).recordSelection:+4
go/types.(*Checker).selector:+232
go/types.(*Checker).exprInternal:+308
go/types.(*Checker).rawExpr:+10
go/types.(*Checker).exprOrType:+1

golang.org/x/tools/gopls@v0.16.1 go1.23.0 windows/amd64 neovim (1)

Issue created by golang.org/x/tools/gopls/internal/telemetry/cmd/stacks.

[adonovan]

@prattmic @mknyszek

[prattmic]

It's unfortunate that this only captures the stack from systemstack_switch.

My guess is that this is most likely "runtime: cannot allocate memory" from https://cs.opensource.google/go/go/+/master:src/runtime/malloc.go;l=1609;drc=54fe0fd43fcf8609666c16ae6d15ed92873b1564

[cherrymui]

If it just runs out of memory, I'm not sure if there is anything particular we can do...

[cherrymui]

@randall77 mentioned that if it is the profiler fail to allocate memory, we can ignore it and just not profile, but it will probably fail soon in some other allocations...

It would be nice if we can get any information about the memory, but that would be hard for telemetry...