Reproducable Python core dump

[Guido van Rossum]

Bradley Baetz <bbaetz at ug.cs.usyd.edu.au> writes:

> My CS project for this semester involves modifying Grail to support
> navigation through a keypad, similarly to Lynx. Whilst testing my
> handing of nested tables, I came across this bug:
> 
> The following will cause a reproducable core dump in Python:
[...]
> This is reproducable on two different RedHat 5.2 Intel boxes, and on a
> SunOS SPARC machine. One of the Redhat boxes and the SPARC box were
> running 1.5.2, the other RedHat box 1.5.1.

Thos will probably be hard to debug.  I looked briefly at the core
dump and it seems to be deep in Tk code -- perhaps the problem is with 
Tcl/Tk?  (That would explain why it's independent of the Python
version you use.)

For whoever wants to look into this more, here's part of the stack
trace from gdb:

#0  0xeeb50094 in Tk_GetGC ()
#1  0xeeb37c10 in GetStyle ()
#2  0xeeb37db4 in LayoutDLine ()
#3  0xeeb38488 in UpdateDisplayInfo ()
#4  0xeeb3b6bc in TkTextPixelIndex ()
#5  0xeeb401d8 in TkTextPickCurrent ()
#6  0xeeb400b0 in TkTextBindProc ()
#7  0xeeb4b9e8 in Tk_HandleEvent ()
#8  0xeeb4bd88 in WindowEventProc ()
#9  0xeeb99d9c in Tcl_ServiceEvent ()
#10 0xeeb9a00c in Tcl_DoOneEvent ()

Since Tk_GetGC() is a huge function, I can't say more without
recompiling Tk with -g.  Perhaps you can try this yourself?

--Guido van Rossum (home page: http://www.python.org/~guido/)