in mjit_pause() fired by before_fork_ruby(). When calling RUBY_VM_CHECK_INTS() in mjit_pause()'s loop, other threads may call start_worker() which sets stop_worker_p = FALSE and it could wait for MJIT worker stop forever even while stop_worker_p is FALSE.
Added by k0kubun (Takashi Kokubun) almost 7 years ago
mjit.c: fix deadlock introduced by r65785
in mjit_pause() fired by before_fork_ruby(). When calling
RUBY_VM_CHECK_INTS() in mjit_pause()'s loop, other threads may call
start_worker() which sets
stop_worker_p = FALSEand it could wait forMJIT worker stop forever even while
stop_worker_pis FALSE.http://ci.rvm.jp/results/trunk-mjit@silicon-docker/1466333
https://gist.github.com/ko1/a57ef1a03e1c7cfd22f1495e0864b63d
http://ci.rvm.jp/results/trunk-mjit-wait@silicon-docker/1466335
https://gist.github.com/ko1/96e867e36e6b75559b3d926f8c0bdaeb
https://app.wercker.com/ruby/ruby/runs/mjit-test1/5bf1221c183106002855989c?step=5bf1225987436a0006192c31
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@65790 b2dd03c8-39d4-4d8f-98ff-823fe69b080e