In message <199408111627.JAA18169@large.cisco.com>, David Carrel writes:
>Jason,
>
>obviously you've found the same bug that I am chasing.  Unfortunately, I
>made the same patch that you posted and my server still crashes.  I found
>that checking that pGC->tile.pixmap is not NULL helps somewhat, but
>occasionally I get to this same point and pGC->tile.pixmap = 1.  I
>definitely can't dereference it, and it passes the NULL pointer check.  I
>was backtracking to figure out how this gets initialized, but I'm not done
>yet.
>
>Thoughts???

my observation is that whenever i allow the server to dump core, the
pointer will be referencing some odd low memory address (7, for example),
but by debugging the server while it's running, the pointer 'always'
appears to be NULL.

the check for NULL seemed correct, since my test case was triggering it
when it was doing some particularly nasty vodo with windows that appeared
to only be partially created (and weren't really 'onScreen').

obviously if you've seen it pass the NULL check, something else is wrong.
i think pGC->tile is a union, so perhaps something is setting the other
field when it shouldn't be.

--
                    ----------------------------------------
-------------------// jason downs // downsj@CSOS.ORST.EDU //------------------
                   ----------------------------------------      JD105
		  http://www.CSOS.ORST.EDU/downsj/index.html
		       have you fed your sysadmin today?

------------------------------------------------------------------------------