GDB-Protocol: Formalize ``!'' extended remote

Andrew Cagney
Thu Jun 24 04:43:00 GMT 1999

Stan Shebs wrote:
>    From: Andrew Cagney <>
>    Date: Thu, 17 Jun 1999 11:01:42 +1000
>    At present there are two general forms of the remote protocol ``remote''
>    and ``extended-remote'' (there is also extended - extended - unofficial
>    - confused - remote but that is another issue ;-).
>    At present the user must issue ``extended-remote'' before GDB will try
>    to use that command.  I'd like to propose that, instead GDB always try
>    for extended-remote (sending the ``!'' packet and if nack'd revert to
>    the old protocol.
> I think that would be great.  I vaguely remember that someone didn't
> want to do that originally because of old stubs not complying with the
> requirement to ignore unrecognized packets, but if you have the
> override option that possibility is covered.

I suspect that the reason was actually that new stubs return "" when
sent "!".  That just happens to identical to response as old stub would
return :-(.  You can't tell the diference...

The second alternative is to just make ``target remote'' assume
extended-remote and control it by a ``set command''.

Now about that version packet ... :-)


More information about the Gdb mailing list