raveling@isi.edu (Paul Raveling) (11/07/89)
HELP!!!!!! We're seeing a lot of X11 display server crashes on HP 9000/300 series machines running MIT X11R3. If anyone else (HP people perhaps?) have seen this and tracked it down, we'd appreciate hearing about it. Symptoms are inconsistent, but often involve bus errors, segmentation faults, and "softer" failures in XRealloc. They're generally consistent with memory clobber. Some core files show stack traces that also strongly hint at memory clobber. What's common among these failures is that it appears they all involve either: -- Tiling the root window, especially with a pixmap that's larger than the screen. For example, I can reproduce it at will by running xbigphoon on a Topcat (1024x768) display. It's harder to produce failures when tiling with smaller pixmaps, but it happens. When it happens, it's clear that the pixmap needs to be clipped at the edge of the screen. -- An exposure on a window with at least one dimension larger than the corresponding screen dimension. A couple stack traces from dumps suggest one of the ValidateGC routines may be involved (perhaps cfbValidteGC or topcatValidateGC). Does anyone recognize these symptoms? ---------------- Paul Raveling Raveling@isi.edu
dan@hpurvmc.HP.COM ( Dan Felman (SE) ) (11/28/89)
> HELP!!!!!! > We're seeing a lot of X11 display server crashes on > HP 9000/300 series machines running MIT X11R3. If > anyone else (HP people perhaps?) have seen this and > tracked it down, we'd appreciate hearing about it. > Symptoms are inconsistent, but often involve bus errors, > segmentation faults, and "softer" failures in XRealloc. > They're generally consistent with memory clobber. Some > core files show stack traces that also strongly hint at > memory clobber. > What's common among these failures is that it appears > they all involve either: > ...... Since the S300 6.5 server is R2 based, I assume you have the MIT release instead. Instead of trying to tear your hair out with 'untested' s/w, I would wait for HP's 7.0 release where X11 will ship with R3 server and OSF MOTIF window manager. Hope this helps!!! regards Dan Dan Felman ; Systems Engineer ; Hewlett Packard ; 2 Choke Cherry Rd dan@hpurvmc (inside HP) ; (301) 258 - 5910 ; Rockville MD 20850 dan%hpurvmc@hplabs (outside HP)
tml@hemuli.atk.vtt.fi (Tor Lillqvist) (11/29/89)
In article <80013@hpurvmc.HP.COM> dan@hpurvmc.HP.COM ( Dan Felman (SE) ) writes: >Since the S300 6.5 server is R2 based, I assume you have the MIT release >instead. Instead of trying to tear your hair out with 'untested' s/w, I >would wait for HP's 7.0 release where X11 will ship with R3 server and >OSF MOTIF window manager. But... at about the same time 7.0 with the X11R3 style server (does this (I mean the server, not the MOTIF stuff) affect much else than font naming?) comes from HP, MIT will release X11R4, containing among other goodies a non-rectangular window extension... I will probably run the MIT server, and a mixture of MIT and HP clients. I haven't seen mwm, so I can't say whether I will use it or continue using twm. (I have an old version of hpwm, which is clearly inferior to twm.) -- Tor Lillqvist, VTT/ATK
harry@hpcvlx.cv.hp.com (Harry Phinney) (11/30/89)
I've cut your questions up a bit, but I think I preserved the content. > (does this (I mean the server, not the MOTIF stuff) affect much else than > font naming?) The "R3" changes for 7.0 are really minor, and you are correct that the only really visible change is the font naming. The 7.0 server also has better performance that the 6.5 server, and it generally uses less memory. > But... at about the same time 7.0 with the X11R3 style server > comes from HP, MIT will release X11R4, containing among > other goodies a non-rectangular window extension... > I will probably > run the MIT server, and a mixture of MIT and HP clients. > Tor Lillqvist, VTT/ATK There's certainly nothing wrong with using the server distributed on the MIT tape, but you should realize that we are not in a position to provide support for it. We are supplying a server for R4 specifically for those customers who wish to take advantage of such features as the shape extension, and are willing to suffer all the disadvantages of using unsupported software. For various reasons (primarily Starbase support), our production servers are significantly different than those distributed on the MIT tape. We simply can't devote too much time to maintaining the "free" MIT server. I view the MIT distribution as a developer's kit, with the customer-quality product delivered with the HP-UX releases. Harry Phinney