Original question:
>Have a problem that cropped up last night. Compiling a binary here and
>running it
>yields the following error:
>
>libc internal error: _rmutex_unlock: rmutex not held
>
>This is on a Ultra5, Solaris 2.6, using Sun Workshop 4.2 C/C++ compilers.
>On a previous machine,
>a SparcStation10 with Solaris 2.5.1, this worked fine. The old machine is
>no longer available. :-(
>
>I've been poking around a bit for patches... 105210-25 and 105568-14 are
>installed on the
>system. Any of you run into this or similar before? This is a bit
>critical, so any help
>would be appreciated. I will continue to poke around for clues, and have a
>few other patches to check
>out that I've gleaned from past posts on this list. I've attached a truss
>output
>for you folks to review, error appears about 15 lines from bottom.
>
>Thanks, and regards,
#################################
Response from Casper Dik:
"This can occur when you've compiled a C++ library with -lthread but
didn't pass the "-mt" option."
This gave a big hint, still testing some things here. The program is
being compiled with -lpthread, but adding -mt to the flags seems to
have fixed the problem so far. Thanks, Casper.
This did bring up another question... was -mt a standard option on other
installations of C++ compilers?
#################################
Response from Birger A. Wathne:
"Peraps you need to set up some shared memory/semaphore resources in
/etc/system?"
I'll keep a watch on this as testing goes on, as there is also a database on
this system.
#################################
Response from Thomas Wardman:
"Have you installed the latest Y2K patches? There's an update from the end
of
November that fixes a SPARCcompiler issue w/ y2k.. that could be your
problem."
I've installed the cluster as it was in mid- to late-December. As testing
progresses,
if there are still problems, I'll pursue this further.
#################################
Thanks, all!
- Alan
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:14:01 CDT