Title: | DIGITAL UNIX (FORMERLY KNOWN AS DEC OSF/1) |
Notice: | Welcome to the Digital UNIX Conference |
Moderator: | SMURF::DENHAM |
Created: | Thu Mar 16 1995 |
Last Modified: | Fri Jun 06 1997 |
Last Successful Update: | Fri Jun 06 1997 |
Number of topics: | 10068 |
Total number of notes: | 35879 |
There seems to be a problem when using the shutdown command from within a multithreaded program, as the system gets hung. The application behaved correctly under OSF 3.2g and before but doesn't work on 4.0b. The application issues a "shutdown -r now" by using the system() system call. The machine hangs with a blue screen. The syncing disks ...done message never appears. The >>> never appears, the system doesn't reboot. The firmware was updated to 3.8. A non-threaded sample program that only does the "shutdown -r now" via the system() system call will work. The program that fails has 14 other threads running, and 1 file open. If the application is changed so that all the threads are exited before issueing the shutdown command, things will work. Is this a problem with the UNIX or the threads library? lee arnold [Posted by WWW Notes gateway]
T.R | Title | User | Personal Name | Date | Lines |
---|---|---|---|---|---|
9454.1 | SMURF::DENHAM | Digital UNIX Kernel | Wed Apr 09 1997 17:12 | 6 | |
Hey, that's pretty cool! A crash dump of the hung system would be most informative. Got a program I can use, so I don't have code one up? Even an executable would be nice. Sounds like some kind of deadlock clearning the processes out of the system. Doubt that the threads library can do that.... | |||||
9454.2 | test executable | NETRIX::"[email protected]" | lee arnold | Wed Apr 09 1997 17:26 | 6 |
You could try the program that is causing the problem. There is a lot of baggage in the code, but it will run with only a few other files needed. How can I send you the files? lee [Posted by WWW Notes gateway] | |||||
9454.3 | SMURF::DENHAM | Digital UNIX Kernel | Thu Apr 10 1997 09:46 | 2 | |
I dunno, tar/compress/uuencode with a README and mail to [email protected] ought to do it. |