`opaque_closure` test segfaulting on aarch64-linux-gnu

[giordano]

julia> Base.runtests("opaque_closure")                                                                                                                       
Running parallel tests with:                                                                                                                                 
  getpid() = 390433                                                                                                                                          
  nworkers() = 1                                                                                                                                             
  nthreads() = 1                                                                                                                                             
  Sys.CPU_THREADS = 36                                                                                                                                       
  Sys.total_memory() = 477.556 GiB                                                                                                                           
  Sys.free_memory() = 442.172 GiB                                                                                                                            
                                                                                                                                                             
Test       (Worker) | Time (s) | GC (s) | GC % | Alloc (MB) | RSS (MB)                                                                                       
opaque_closure  (1) |        started at 2024-04-11T21:05:15.712                                                                                              
                                                                                                                                                             
[390433] signal 11 (2): Segmentation fault                                                                                                                   
in expression starting at /home/cceamgi/.julia/juliaup/julia-nightly/share/julia/test/opaque_closure.jl:369                                                  
jl_system_image_data at /home/cceamgi/.julia/juliaup/julia-nightly/lib/julia/sys.so (unknown line)                                                           
jl_system_image_data at /home/cceamgi/.julia/juliaup/julia-nightly/lib/julia/sys.so (unknown line)                                                           
Allocations: 14847564 (Pool: 14847364; Big: 200); GC: 14                                                                                                     
ERROR: A test has failed. Please submit a bug report (https://github.com/JuliaLang/julia/issues)                                                             
including error messages above and the output of versioninfo():                                                                                              
Julia Version 1.12.0-DEV.325                                                                                                                                 
Commit e9a24d4cee4 (2024-04-10 13:11 UTC)                                                                                                                    
Build Info:                                                                                                                                                  
  Official https://julialang.org/ release                                                                                                                    
Platform Info:                                                                                                                                               
  OS: Linux (aarch64-linux-gnu)                                                                                                                              
  CPU: 72 × unknown                                                                                                                                          
  WORD_SIZE: 64                                                                                                                                              
  LLVM: libLLVM-16.0.6 (ORCJIT, neoverse-v2)                                                                                                                 
Threads: 1 default, 0 interactive, 1 GC (on 72 virtual cores)                                                                                                

This is failing also in all jobs in CI, but aarch64-linux-gnu is allowed to fail tests because of #52434, so no one noticed it.

[giordano]

Doing a quick search through buildkite logs, I think it was introduced by 4ee1022 (#53878). CC: @Keno.

[giordano]

With a debug+assertions build of be3bc9a I got just barely more information:

Test       (Worker) | Time (s) | GC (s) | GC % | Alloc (MB) | RSS (MB)
opaque_closure  (1) |        started at 2024-04-12T17:37:36.108
fatal error: stack corruption detected

[482916] signal 6 (-6): Aborted
in expression starting at /home/cceamgi/repo/julia/usr/share/julia/test/opaque_closure.jl:369
__pthread_kill_implementation at /lib64/libc.so.6 (unknown line)
raise at /lib64/libc.so.6 (unknown line)
abort at /lib64/libc.so.6 (unknown line)
__stack_chk_fail at /home/cceamgi/repo/julia/src/rtutils.c:236
eval_value at /home/cceamgi/repo/julia/src/interpreter.c:346
jl_system_image_data at /home/cceamgi/repo/julia/usr/lib/julia/sys-debug.so (unknown line)
jl_system_image_data at /home/cceamgi/repo/julia/usr/lib/julia/sys-debug.so (unknown line)
Allocations: 16190713 (Pool: 16190388; Big: 325); GC: 13

Note the

fatal error: stack corruption detected

from

julia/src/rtutils.c

Line 234 in be3bc9a

fprintf(stderr, "fatal error: stack corruption detected\n");

The previous frame refers to src/interpreter.c:346, but that's weird since it's just the closing brace of a function definition:

julia/src/interpreter.c

Line 346 in be3bc9a

}

In GDB I get

(gdb) where
#0  0x0000fffff7e34650 in __pthread_kill_implementation () from /lib64/libc.so.6
#1  0x0000fffff7def86c in raise () from /lib64/libc.so.6
#2  0x0000fffff7dd7030 in abort () from /lib64/libc.so.6
#3  0x0000fffff73cf60c in __stack_chk_fail () at /home/cceamgi/repo/julia/src/rtutils.c:236
#4  0x0000fffff7349be0 in eval_value (e=0x0, s=0x0) at /home/cceamgi/repo/julia/src/interpreter.c:346
Backtrace stopped: previous frame inner to this frame (corrupt stack?)

also complaining about corrupted stack.

[gbaraldi]

I think this might be a wrong gc_pop

[gbaraldi]

This code reproduces it

ir = first(only(Base.code_ircode(sin, (Int,))))
oc = Core.OpaqueClosure(ir, do_compile=false)
oc(1)

but no stacktrace in gdb as well, this is a full blown corrupt stack

[gbaraldi]

This got fixed in #54443