[remote protocol] support for disabling packet acknowledgement

Pedro Alves pedro@codesourcery.com
Fri Jul 11 15:11:00 GMT 2008


[so, comming back to the original functionality]

On Thursday 10 July 2008 21:13:20, Paul Koning wrote:

> By the way, note that "target remote host:port" does NOT mean that you
> have TCP end to end.  You may have TCP to a terminal server and raw
> UART the rest of the way, so you still need ACKs (of one kind or
> another).

Right, that's why we made it so that only if the stub reports
qSupported:QStartNoAck+, GDB will turn it on by default.  If you
notice the gdbserver patch, it will only report that, if the
connection is coming from TCP.  It is my understanding that
situations like these are not usually seen:

GDB UART <-> something <-> TCP/IP stub

Would it make people happier if GDB would only enable noack mode
*automatically* if:

  - The "set remote noack-packet" setting is set to auto

  - The stub reported support and *preference* to it with qSupported
    (it knows it's using a reliable connection on its end)

  - GDB also knows it is using a reliable connection (tcp/pipe)

That last bit was there until the last version of the patch
before submission, but I ended up removing it from the final
patch.  We can always go with or without that bit for now, and
change it later, as the difference is all on GDB's side.

-- 
Pedro Alves



More information about the Gdb mailing list