Using '--enable-logging' leads to hang in a simplest case about dmtcp

[karya0]

Thanks for the report. I'll generate a PR for the fix

[karya0]

I am unable to reproduce this issue with the master branch. Can you confirm the commit SHA/version you are using?

[Koishi-love]

I am unable to reproduce this issue with the master branch. Can you confirm the commit SHA/version you are using?

Yes. I downloaded "DMTCP 3.0 released" codes on the github manual page. https://github.com/dmtcp/dmtcp/releases/tag/3.0.0

And further, I try dmtcp 2.6.0 release version, It's ok, no deadlock issue.

[Koishi-love]

I am unable to reproduce this issue with the master branch. Can you confirm the commit SHA/version you are using?

I use the master branch codes for this case, then I also meet the issue.

There are JTRACE logs:

[2024-03-19T09:59:46.794, 40000, 40002, Trace] at dmtcpworker.cpp:335 in waitForPreSuspendMessage; REASON='waiting for CHECKPOINT message'
1 2 [2024-03-19T09:59:47.671, 40000, 40002, Trace] at threadsync.cpp:296 in presuspendEventHookLockLock; REASON='Acquiring event-hook lock'
[2024-03-19T09:59:47.671, 40000, 40002, Trace] at dmtcpworker.cpp:380 in waitForCheckpointRequest; REASON='Procesing pre-suspend barriers'
[2024-03-19T09:59:47.671, 40000, 40002, Trace] at threadsync.cpp:303 in presuspendEventHookLockUnlock; REASON='Releasing event-hook lock'
[2024-03-19T09:59:47.671, 40000, 40002, Trace] at dmtcpworker.cpp:384 in waitForCheckpointRequest; REASON='Preparing to acquire locks before DMT:SUSPEND barrier'
[2024-03-19T09:59:47.671, 40000, 40002, Trace] at threadsync.cpp:103 in acquireLocks; REASON='Waiting for libdlLock'
[2024-03-19T09:59:47.671, 40000, 40002, Trace] at threadsync.cpp:106 in acquireLocks; REASON='Waiting for other threads to exit DMTCP-Wrappers'
[2024-03-19T09:59:47.671, 40000, 40002, Trace] at threadsync.cpp:109 in acquireLocks; REASON='Done acquiring all locks'
[2024-03-19T09:59:47.671, 40000, 40002, Trace] at dmtcpworker.cpp:387 in waitForCheckpointRequest; REASON='Waiting for DMT:SUSPEND barrier'
[2024-03-19T09:59:47.671, 40000, 40002, Trace] at coordinatorapi.cpp:439 in waitForBarrier; REASON='waiting for DMT_BARRIER_RELEASED message'
barrier = DMT:SUSPEND
[2024-03-19T09:59:47.712, 40000, 40002, Trace] at dmtcpworker.cpp:393 in waitForCheckpointRequest; REASON='Starting checkpoint, suspending threads...'

[Koishi-love]

I am unable to reproduce this issue with the master branch. Can you confirm the commit SHA/version you are using?

Now I'm sure that JTRACEs in function 'void stopthisthread(int signum)' make the issue. After I delete these JTRACEs, the program don't hang at all. It seems like that JTRACE blocks 'sem_post', which blocks 'sem_wait' in 'suspendThreads()'. This leads the issue, but I don't know the root cause.

I hope that it can make dmtcp better.