[Please cc, I don't subscribe to this list]
First, this is not a bug in fvwm: no client should be able to crash an
Xserver. I send this to you because I suspect it is highly unlikely
that SGI will acknowledge and fix this problem in a timely manner
(IRIX 6 will be out for everything at some point and they'll just say
'upgrade'), and I hope someone can provide a workaround. I don't mean
to pick on SGI, I just need a quick fix.
OS: SGI IRIX 5.x (only) on all graphics hardware I could test (VGX,
XL, XT, RE^2). It does not affect 4.x or 6.x on any hardware I
could detect.
Software: Xsgi (X11R5 unknown patchlevel provided by SGI), and
fvwm-1.2x (built with gcc, IRIX cc, linked against SGI Xlibs, MIT
Xlibs at R5pl26 and R6pl13) or fvwm-2.0.40 (any build and link
combination as above), and ImageMagick-3.x (as above). [NOTE: I
have not gone through every single combination of the 3 since at
some point the number of combinations approaches ridiculousness,
>1M I'm sure if I recall anything from my statistics class] Oh,
and xpm-3.4f as well if that matters.
Descripton: It's the combination of software that causes Xsgi to
crash. The same ImageMagick binary running under SGI's 4Dwm does not
cause the crash (in fact, the ImageMagick author tells me that she
develops on an SGI not unlike the one on my desk and doesn't see the
problem). Displaying images with ImageMagick works fine (in general),
but as soon as any image manipulation is done (say to double the
viewing size) the server dumps core and xdm restarts.
At one point someone here was working on a solution to the problem and
mentioned tracking it down to a backing store request being made by
ImageMagick. I've fiddled a little with the code of ImageMagick to
get it to set backing store requests to 'Never' but it has not helped
the problem, so that seems like a dead end. I'm not at a loss.
Any intrepid soul out there willing to offer some assistance? I'm
currently running R5 Xsgi (from vendor), and fvwm-1.24r linked against
R6pl13 with gcc. I can change to fvwm-2.0.40 and use R5 libs (or
vendor libs) if needed to find a cure.
Travis
--
To unsubscribe from the list, send "unsubscribe fvwm" in the body of a
message to majordomo_at_hpc.uh.edu.
To report problems, send mail to fvwm-owner_at_hpc.uh.edu.
Received on Wed Jan 17 1996 - 12:15:40 GMT