tribe@groucho (08/18/90)
I've been having some trouble with HP's "Starbase-on-X11" (sox11) device
driver. Local people have not been able to help, so I would appreciate any
help from starbase gurus (!) on the net.
The program draws to a window created with xwcreate(1) and reads mouse choice
and locator input from the window.
1) The 31st and all succeeding executions of the program on a machine will
dump the core on the track call. Other diskless nodes in a cluster are
unaffected. Powering down the node will reset things. The message given
by starbase is
Starbase error 2053: Shared memory call failed.
Procedure name: track
File descriptor: 3
Device file name: /dev/screen/Starbase
Library location: operating system utilities (0)
System errno: No space left on device
Bus error - core dumped
Cdb(1) traces the offense to
0 __hp_set_track_mode + 0x10 (0x3, 0x1, 0x3, 0xffeffbd8, 0x197a)
1 _track + 0x86 (0x3, 0x3, 0x1, 0x683b4, 0xffeffbe0)
Track is called via
display = gopen("/dev/screen/Starbase", OUTDEV, "sox11", INIT | RESET_DEVICE);
locator = display;
track(locator, display, 1);
It seems that a 30-item table somewhere is being filled up by the
executions of my program. Setting "locator = display" seemed odd to me
but was the only way to get the mouse read.
2) gclose fails, on any execution. Starbase says
Starbase error 100: X library call failed
Procedure name: gclose
File descriptor: 3
Device file name: /dev/screen/Starbase
Library location: sox11 driver (10)
So I commented it out. Seems reasonable that glose not deallocating
resources is the cause of my tables filling up. A call to track_off
is successful, though.
3) After a successful run of the program stops, the mouse disappears if it
is moved into the graphics window.
4) Two hp X routines are not linked, so I made a file that defines them
as null. The routines are XHPGetWindowCursor and XHPKeysymToRoman8.
After my routines, I link in "/usr/lib/end.o -ldd300h -lddsox11 -lsb1
-lsb2 -lXr11 -lX11 -ldvio -lm".
I am using a hp9000 series 300 (370 & 330) running HP-UX 6.5A.
Please E-Mail to me (preferably to doc.ee.uidaho.edu) as I will be away from
the net next week and your reply may expire before I return. If there are
sufficient "me-too"s and a working solution, I will summarize. Thanks for
any help you can provide.
--
Duane Tribe, University of Idaho
tribe@doc.ee.uidaho.edu, tribe@ted.cs.uidaho.edu