Re: Standard GDB Remote Protocol

[Andrew Cagney <ac131313 at cygnus dot com>]

"J.T. Conklin" wrote:
> 
> >>>>> "Andrew" == Andrew Cagney <ac131313@cygnus.com> writes:
> Andrew> How GDB behaves should be clarified.  However, any definition should
> Andrew> avoid refering to absolute times and instead refer the user back to a
> Andrew> number of knobs that can be tweeked from the GDB command line.
> 
> Steven is/was examining the protocol specification without the GDB
> and stub sources to avoid becoming biased by exposure.  In addition,
> I recommend that, at least at the present time, this also be true of
> GDB's variables.
> 
> If Steven is willing to carry this a bit farther, I'd like to see the
> protocol defined in terms of 'variables' which can be later mapped to
> GDB's variables.  Or in the case where GDB does not have such a user
> configurable varible, a concious decision can be made whether to add
> one or continue using whatever constant is in the source.

That would be bliss.

> For example, this thread has me thinking that there are actually two
> separate timeouts in the protocol.  The timeout from when a packet is
> sent and it is ACK/NAKed.  And a timeout from when a command is sent
> and it's response is received.  At the present time, 'remotetimeout'
> is used for both purposes.  But it's reasonable to set the first to a
> short value and the latter to something much larger.

There is also (as noted earlier) the time between each character that
arrives.  GDB uses the same timeout for that.

	Andrew