pthread_rwlock_unlock change the signal mask!

[Changho]

I have a problem with using pthread_rwlock API.

I think it changes the signal mask abnormally.

Test program source file attached. (Just change the extension to c.)

Shortly, It just uses pthread_rwlock_rdlock -> pthread_rwlock_rdlock -> pthread_rwlock_unlock.

The last pthread_rwlock_unlock called, SIGUSR2 signal is blocked unexpectedly.

P.S. This thread has been moevd from HP-UX > General to HP-UX > languages - Hp forum moderator

[Changho]

pthread library version is like this.

$ /usr/sbin/swlist -l patch | grep pthread

# PHCO_39298                            1.0            pthread library Japanese manpage cumulative patch

# PHCO_41407                            1.0            pthread library cumulative patch

[Dennis Handly]

Have you tried using tusc to see if it changes the signal mask?

[Changho]

I don't know the location of signal masking info in memory.
I just can not help to check the result of pthread_sigmask() call.
(Please refer to my source code.)

Could you tell me more specifically?

[blade_runner]

$ cc pthread_rwlock.c -o pthread_rwlock -lpthread

$ ./pthread_rwlock

sigmask error - -2 (line:57)

[Changho]

Yes, it's just what I'm asking.
It seems like pthread library bug.

[Dennis Handly]

>I don't know the location of signal masking info in memory.

You run tusc to trace the system calls.  I only see sigprocmask.  In fact if I remove pthread_sigmask and replace by sigprocmask I get the same bad thing.  But the only syscall is sigprocmask with SIG_BLOCK and NULL for the set parm.

I have no idea how the mask can change without a syscall?  Unless it is stored in user space?

It gets changed to: 0xff7ff843fffffff3

It looks like you need to contact Support and file a bug.

The recommended libpthread patch is PHCO_42326 but I don't see any fixes for this.