Fix thread safety in atexit(f): Lock access to atexit_hooks

[NHDaly]

• atexit(f) mutates global shared state.
• atexit(f) can be called anytime by any thread.
• Accesses & mutations to global shared state must be locked if they can be accessed from multiple threads.

Fixes #49746

[kpamnany]

Good find!

I'm fine with the one-liner.

Guessing you checked whether the test fails on current master?

[NHDaly]

Guessing you checked whether the test fails on current master?

@kpamnany: Yes, and boy do they fail!...

julia> using Test

julia> @testset "atexit thread safety" begin
           f = () -> nothing
           before_len = length(Base.atexit_hooks)
           @sync begin
               for _ in 1:1_000_000
                   Threads.@spawn begin
                       atexit(f)
                   end
               end
           end
           @test length(Base.atexit_hooks) == before_len + 1_000_000
           @test all(hook -> hook === f, Base.atexit_hooks[1 : 1_000_000])

           # cleanup
          # Base.@lock Base._atexit_lock begin
          #     deleteat!(Base.atexit_hooks, 1:1_000_000)
          # end
       end

[57702] signal (11.2): Segmentation fault: 11
in expression starting at REPL[2]:1
sweep_big_list at /Users/nathandaly/src/julia/src/gc.c:1094 [inlined]
sweep_big at /Users/nathandaly/src/julia/src/gc.c:1116 [inlined]
gc_sweep_other at /Users/nathandaly/src/julia/src/gc.c:1608 [inlined]
_jl_gc_collect at /Users/nathandaly/src/julia/src/gc.c:3315
ijl_gc_collect at /Users/nathandaly/src/julia/src/gc.c:3486
maybe_collect at /Users/nathandaly/src/julia/src/gc.c:954 [inlined]
jl_gc_pool_alloc_inner at /Users/nathandaly/src/julia/src/gc.c:1317
ijl_gc_pool_alloc at /Users/nathandaly/src/julia/src/gc.c:1368
jfptr_iterate_34613 at /Users/nathandaly/src/julia/usr/lib/julia/sys.dylib (unknown line)
_jl_invoke at /Users/nathandaly/src/julia/src/gf.c:0 [inlined]

... here's how it failed on 1.8.2 😂:

signal (11): Segmentation fault: 11

signal (11): Segmentation fault: 11
in expression starting at REPL[2]:1

signal (11): Segmentation fault: 11
in expression starting at REPL[2]:1

signal (11): Segmentation fault: 11
in expression starting at REPL[2]:1
_ZNK5dyld311MachOLoaded17findClosestSymbolEyPPKcPy at /usr/lib/dyld (unknown line)
Allocations: 2961458 (Pool: 2959697; Big: 1761); GC: 3

signal (11): Segmentation fault: 11
in expression starting at REPL[2]:1
write at /usr/lib/system/libsystem_kernel.dylib (unknown line)
Allocations: 2961458 (Pool: 2959697; Big: 1761); GC: 3

signal (11): Segmentation fault: 11
in expression starting at REPL[2]:1

signal (11): Segmentation fault: 11
in expression starting at REPL[2]:1
_platform_strlen at /usr/lib/system/libsystem_platform.dylib (unknown line)
Allocations: 2961458 (Pool: 2959697; Big: 1761); GC: 3

...

lol there were like 30 segfaults reported 😂