Re: [Update REQ 5129]: Bug in DEC ALPHA OSF version of Geomview 1.5.0

["Daeron Meyer" <daeron>]

> 
> Date: Thu, 3 Nov 1994 17:06:05 -0500
> From: redrod at avl.umd.edu (Randy Paffenroth <redrod at avl.umd.edu>)
> Message-Id: <199411032206.RAA06668 at jupiter.avl.umd.edu>
> To: software@geom
> Subject: Bug in DEC ALPHA OSF version of Geomview 1.5.0
> 
> Good afternoon,
> 
> 	I would like to report a bug in version 1.5.0 for the
> DEC ALPHA OSF.  We have several ALPHAs in our lab and Geomview 1.5.0
> only seems to work correctly on machines that have unaccelerated
> graphics.  On machines that have either a PXG card or a Kubota
> Denali Subsystem Geomview will try to start up, flash the
> main view window for a fraction of a second, and then exit.
> 
> 	I have tried this with both the procompiled and source
> versions and get the same behavior.  Now, running it through
> dbx reveals that it exits in line 96 of the file mgx11windows.c,
> inside the function  Xmg_setx11display.  Looking in that area we have:
> 
>   while (colfail && colorlevels > 1)
>   {
>     if (XAllocColorCells(dpy, cmap, False, planemasks, nplanes, mgx11colors,
> 		 (unsigned int)(colorlevels*colorlevels*colorlevels+1)))
>       colfail = 0;
>     else
>       colorlevels--;
>   }
> 
> if (colfail)
>   {
>     fprintf(stderr, "Couldn't allocate enough colors. Try closing other X apps first.\n");
> 
>     exit(0);   <======== line it exits in
>   }
> 
> Now, I guess the trouble lay in the fact that the accelerated displays
> are also TrueColor displays, while the unaccelerated ones are
> PseudoColor displays.
> 
> 	Thanks a lot,
> 
> 		Randy Paffenroth
> 		Institute for Physical Science and Technology
> 
> 
> 
> 


Thanks a lot for the bug report... It helped us pinpoint and hopefully fix
this nontrivial bug. The problem was, we had grown somewhat lazy in our use
of X windows visuals, since the SGI we use for testing of 24 bits supports
all sorts of visuals, 24bit TrueColor, 8bit PseudoColor etc... all at once.
Geomview took advantage of both these visual types, and so that is probably
why it failed on a system that only supported TrueColor visuals. Anyway,
it should correctly work now on a display that only supports TrueColor.
We haven't replaced the binary or source distribution yet because we would
like you test it to see if it works first. Please grab the following file
as a drop in replacement for your gvx binary:

/priv/daeron/bin/gvx.alpha

(as always, available via ftp from ftp.geom.umn.edu)

If it works, please mail us back as soon as possible so we can patch up the
distributions promptly.

Once again thanks,

Daeron
daeron at geom.umn.edu