[ruby-core:97437] [Ruby master Bug#16682] Ruby 2.7.0p0 crash on exit if there is an active RUBY_INTERNAL_EVENT_GC_EXIT tracepoint

Issue #16682 has been updated by methodmissing (Lourens Naud薊.


byroot (Jean Boussier) wrote in #note-6:
> ko1 (Koichi Sasada) wrote in #note-5:
> > how about to disable this TP?
> 
> `at_exit { tracepoint.disable }` indeed prevent the bug from firing in my reproduction repo, and I've used this workaround a while ago successfully.
> 
> But it's merely hiding the bug not fixing it. Our application is segfauling a lot on 2.7.0p0, and not just during the exit sequence, so I suspect this can also happen after a regular GC run.

A cleaner workaround (and yes I agree this doesn't fix the segfault) would be to define a finalizer on the tracepoint instead, which disables it.

The downside of NOT emitting the tracepoint for GC_EXIT could be undefined behaviour for profilers or users of the Tracepoint API where the GC_ENTER event is not paired with a GC_EXIT one anymore. Although this may not matter on shutdown as the process is exiting anyways. Filtering tracepoint objects of type T_NONE would also result in this behaviour.

Jean, the crashes references other TP events too, or just the GC ones?

----------------------------------------
Bug #16682: Ruby 2.7.0p0 crash on exit if there is an active RUBY_INTERNAL_EVENT_GC_EXIT tracepoint
https://bugs.ruby-lang.org/issues/16682#change-84576

* Author: byroot (Jean Boussier)
* Status: Open
* Priority: Normal
* ruby -v: ruby 2.7.0p0 (2019-12-25 revision 647ee6f091) [x86_64-darwin19]
* Backport: 2.5: UNKNOWN, 2.6: UNKNOWN, 2.7: UNKNOWN
----------------------------------------
```
[BUG] object allocation during garbage collection phase
ruby 2.7.0p0 (2019-12-25 revision 647ee6f091) [x86_64-darwin19]

-- Crash Report log information --------------------------------------------
   See Crash Report log file under the one of following:                    
     * ~/Library/Logs/DiagnosticReports                                     
     * /Library/Logs/DiagnosticReports                                      
   for more details.                                                        
Don't forget to include the above Crash Report log file in bug reports.     

-- Control frame information -----------------------------------------------
c:0001 p:0001 s:0003 E:002690 (none) [FINISH]


-- C level backtrace information -------------------------------------------
/Users/byroot/.rubies/ruby-2.7.0/bin/ruby(rb_vm_bugreport+0x96) [0x10fa9f266]
/Users/byroot/.rubies/ruby-2.7.0/bin/ruby(rb_bug+0xcc) [0x10faabb86]
/Users/byroot/.rubies/ruby-2.7.0/bin/ruby(newobj_slowpath+0x99c) [0x10f8f939c]
/Users/byroot/.rubies/ruby-2.7.0/bin/ruby(newobj_slowpath_wb_protected+0x14) [0x10f8f89d4]
/Users/byroot/.rubies/ruby-2.7.0/bin/ruby(rb_str_buf_new+0x1e) [0x10fa151be]
/Users/byroot/.rubies/ruby-2.7.0/bin/ruby(rb_enc_vsprintf+0x48) [0x10fa03178]
/Users/byroot/.rubies/ruby-2.7.0/bin/ruby(rb_vraise+0x14) [0x10f8d4d84]
/Users/byroot/.rubies/ruby-2.7.0/bin/ruby(rb_raise+0x7b) [0x10f8d052b]
/Users/byroot/.rubies/ruby-2.7.0/bin/ruby(rb_check_typeddata+0xf3) [0x10f8d19a3]
/Users/byroot/.rubies/ruby-2.7.0/bin/ruby(tp_call_trace+0x2a) [0x10faa1aaa]
/Users/byroot/.rubies/ruby-2.7.0/bin/ruby(rb_exec_event_hooks+0x163) [0x10faa0ab3]
/Users/byroot/.rubies/ruby-2.7.0/bin/ruby(rb_objspace_call_finalizer+0x8b7) [0x10f8eded7]
/Users/byroot/.rubies/ruby-2.7.0/bin/ruby(rb_ec_cleanup+0x2de) [0x10f8dc35e]
/Users/byroot/.rubies/ruby-2.7.0/bin/ruby(ruby_run_node+0x5f) [0x10f8dc4ff]
/Users/byroot/.rubies/ruby-2.7.0/bin/ruby(main+0x5d) [0x10f833d0d]
```

It also crash in other circumstances, but I'm not able to reproduce them as easily.

Older versions are not affected.

I created a repository to easily reproduce the issue: https://github.com/casperisfine/ruby-tracepoint-crash



-- 
https://bugs.ruby-lang.org/

Unsubscribe: <mailto:[e-mail]?subject=unsubscribe>
<http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>