grossman@BBN.COM writes:
>
> On an alpha (PC164 board from DEC) running netbsd V1.2G, when the system
> panics
> and I get a good dump (takes along time since 512MB of memory), I have a gdb
> problem!
> 
> gdb -k netbsd.X.gdb netbsd.X.core
> 
> The stack backtrace is from the dump routines back to the fault condition, and
> what I want is the stack from the fault condition back to where the kernel was
> entered!
> 
> Has this been fixed, if so, where, and can it be fixed in 1.2 for the alpha?
> 
> PS I know it works correctly in the i386 with netbsd V1.3 running!
>

This was partially fixed for 1.3.  Don't spend time installing 1.3 now,
though.  1.3.1 binaries should be available next week, or switch to 1.3E
the next time a snapshot is posted.

The main problem is that gdb doesn't understand about exception frames.
In 1.3, the assembly source in locore.s is augmented by hand with
debugger symbols that make the exception frames look just enough like
procedure calls so that the gdb traceback usually works. Usually.

One of the things on my list to do is bang on gdb to properly grok
alpha exception frames.

Ross Harvey
ross@teraflop.com