rtems gdb stub with virtex issue

[Till Straumann]

Tim Cussins wrote:
> On Tue, 05 Feb 2008 11:12:27 -0800, "Till Straumann"
> <strauman at slac.stanford.edu> said:
>   
>>> Question: I've set a breakpoint, my thread has hit it and stopped.
>>> At this point the breakpoint is 'disabled' by the stub.
>>>       
>> What makes you believe that? The stub only sets/removes
>> breakpoints on behalf of gdb.
>>     
>
> You're dead right - I should've turned on the DEBUG_COMM switch...
>
>   
>> The only exception being when
>> it is asked to single-step over an existing breakpoint.
>> In that case, the breakpoint is repaced by the original
>> instruction which is subsequently executed and again
>> replaced by the breakpoint.
>>
>> gdb usually lets the stub insert breakpoints before
>> issuing 'cont' or 'step' and removes them from the target
>> as soon as it gains control (but would re-insert them
>> before issuing 'cont' again).
>>     
>
> This is true. gdb is setting the breakpoint before continuing,
> the trap is being patched in by the stub, but the thread fails to
> stick on it the next time around...
>
>   
Hmm - it certainly works fine here.
 - did you change the 'do_patch()' assembly code ?
 - are you sure the breakpoint should be hit ?
 - can you cook your code down to a simple example
   still producing the problem? If so, please attach...

T.
>> What version of gdb are you using?
>>
>>     
>
> GNU gdb 6.6
> This GDB was configured as "--host=i486-linux-gnu
> --target=powerpc-ssrl-rtems".
>
> Cheers,
> Tim
>