[Xastir] Re: Crash on 64-bit FC3

[Jim King]

I wrote:

>I'm running FC3 test 2, Xastir source pulled from CVS yesterday.  I 
>suspect this is an FC3 bug, but I'm an X11 noob.  Xastir crashes with a 
>segmentation violation.  In gdb I get:
>
>(gdb) bt
>#0  0x0000003b07c4d061 in _XtCountVaList () from /usr/X11R6/lib64/libXt.so.6
>#1  0x0000003b07c4cabe in XtVaGetValues () from /usr/X11R6/lib64/libXt.so.6
>#2  0x0000000000473925 in da_expose (w=0x8b9960, client_data=0x3b00000000,
>    call_data=0x7fbfffe7f8) at main.c:9069
>#3  0x0000003b07c13fa5 in XtCallCallbackList ()
>   from /usr/X11R6/lib64/libXt.so.6
>[etc.]
>
>It looks to me like the call from da_expose to XtVaGetValues is causing 
>a segv.  Since this same code works on 32-bit Linux that led to me guess 
>that this is a bug in 64-bit FC3, unless maybe Xastir is using 
>XtVaGetValues in a not-quite-sanctioned manner that happens to work on 
>32-bit but not on 64-bit.
>
>If somebody can confirm that Xastir's use of XtVaGetValues is OK I'll 
>submit a Fedora bug.  Otherwise if somebody can suggest some variations 
>on the function call to try I'll do that.
>

A little bit of looking at the source code was inspirational, and I 
think I've found the problem.  It looks to me like the XtVaXXX functions 
require a NULL as the final parameter to signal the end of argument 
list.  In main.c, a little over half of the calls to XtVaGetValues use 
"0" as the final parameter, and the rest of the calls use "NULL" as the 
final parameter.  I think that all of the calls to other XtVaXXX 
functions use "NULL" as the final parameter.  On x86_64, int's are still 
32 bits, but pointers are 64 bits, so "0" (an int) and "NULL" (a 
pointer) are not the same thing.  I changed all the of the 0 final 
parameters to NULL in calls to XtVaGetValues and Xastir seems to run OK now.

Jim, KD5LUQ