Open LouisJenkinsCS opened 4 years ago
Should SIGALRM and have asynchronous signal handler just set the running
flag to false
.
When running it, I found that with super-small granularities --scheduling-quantum={1,2,3,4,5,6,7,8}
etc. it works, but perhaps it has to do with RNG, sometimes it never finds it no matter how long it runs for. I believe I need to make some revisions w.r.t RNG, such as replacing the seed every now and then.
Since prior to the enhancement, it was possible to create scenarios where the false branch was never taken due to the atomic load falling inside the same superblock as the CAS that depends on it, I should have a simple unit test that asserts that after X minutes it ends up failing.
The above should not fail with the old translation but should fail with the new.