Segfault running tapioca dsl

[natematykiewicz]

It segfaults processing the pg gem.

$ bundle exec tapioca dsl

Output file attached because it's 1.9MB. output.txt

[Morriar]

👋 Hey @natematykiewicz,

It looks like you're encountering a segfault in Ruby 3.3.5 itself. I'd suggest sending your crash report to https://bugs.ruby-lang.org/ instead.

Thanks.

[natematykiewicz]

@Morriar, I tried skipping the PG::Connection constant with bundle exec tapioca dsl --skip-constant=PG::Connection, and that didn't prevent the crash. I don't use the PG gem directly, ActiveRecord uses it, so I don't need any type definitions for PG at all. Is there a way to just have Tapioca skip this file?

[natematykiewicz]

I looked at the .ips file in ~/Library/Logs/DiagnosticReports and saw this:

  "vmRegionInfo" : "0x110c3cb06 is not in any region.  Bytes after previous region: 2823  Bytes before following region: 13562\n      REGION TYPE                    START - END         [ VSIZE] PRT\/MAX SHRMOD  REGION DETAIL\n      __LINKEDIT                  110c2c000-110c3c000    [   64K] r--\/rwx SM=COW  ...socket.bundle\n--->  GAP OF 0x4000 BYTES\n      VM_ALLOCATE                 110c40000-110c50000    [   64K] rw-\/rwx SM=COW  ",
  "exception" : {"codes":"0x0000000000000001, 0x0000000110c3cb06","rawCodes":[1,4576234246],"type":"EXC_BAD_ACCESS","signal":"SIGABRT","subtype":"KERN_INVALID_ADDRESS at 0x0000000110c3cb06"},
  "termination" : {"flags":0,"code":6,"namespace":"SIGNAL","indicator":"Abort trap: 6","byProc":"ruby","byPid":33576},
  "vmregioninfo" : "0x110c3cb06 is not in any region.  Bytes after previous region: 2823  Bytes before following region: 13562\n      REGION TYPE                    START - END         [ VSIZE] PRT\/MAX SHRMOD  REGION DETAIL\n      __LINKEDIT                  110c2c000-110c3c000    [   64K] r--\/rwx SM=COW  ...socket.bundle\n--->  GAP OF 0x4000 BYTES\n      VM_ALLOCATE                 110c40000-110c50000    [   64K] rw-\/rwx SM=COW  ",
  "asi" : {"CoreFoundation":["*** multi-threaded process forked ***"],"libsystem_c.dylib":["crashed on child side of fork pre-exec"]},
  "extMods" : {"caller":{"thread_create":0,"thread_set_state":0,"task_for_pid":0},"system":{"thread_create":0,"thread_set_state":0,"task_for_pid":0},"targeted":{"thread_create":0,"thread_set_state":0,"task_for_pid":0},"warnings":0},
  "faultingThread" : 0,

"crashed on child side of fork pre-exec" and "*** multi-threaded process forked ***" stuck out to me. In my sorbet/tapioca/config.yml I set:

dsl:
  workers: 1

It no longer crashes. I notice that this gem uses Parallel for the process forking:

tapioca/lib/tapioca/executor.rb

Lines 28 to 32 in 316688e

	def run_in_parallel(&block)
	# To have the parallel gem run jobs in the parent process, you must pass 0 as the number of processes
	number_of_processes = @number_of_workers == 1 ? 0 : @number_of_workers
	Parallel.map(@queue, { in_processes: number_of_processes }, &block)
	end

And Parallel has had an open issue about segfaults for a year now.
grosser/parallel#337

I've tried it on Ruby 3.2.4, 3.2.5, and 3.3.5. All segfault with more than 1 process. That linked issue lists Ruby 3.2.2.

Do you still think it's a Ruby issue, or is it a Parallel issue?

[natematykiewicz]

I had asked around about this as well, and someone showed me this: ged/ruby-pg#538 (comment)

Adding ENV['PGGSSENCMODE'] = 'disable' to my bin/tapioca fixes the issue and allows me to use the default worker count.