Re: [PATCH] Postpone connecting from gdb_start to gdb_load in native-extended-gdbserver

[Yao Qi <yao at codesourcery dot com>]

On 09/13/2012 12:54 AM, Pedro Alves wrote:
> The downside is that this disconnects between runs, which causes
> gdbserver to exit (due to --once).  Consider multi-process (+ non-stop)
> tests, that will want to "run" more than once in the same session.
> This patch would make gdbserver exit between those "run"s.
>
> So, I think it'd be better to solve this some other way.

We either teach GDB/GDBserver to flip 'non-stop' and 'async' after 
'target extended-remote' (as you described in LocalRemoteFeatureParity 
wiki page) or tweak testsuite to start gdbserver without --once and 
terminate it in proc gdb_exit by sending 'monitor exit'.  Which one do 
you prefer?

Considering that 'extended-remote' is to act as a native target on 
gdbserver, the former might be better.

--
Yao