This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [RFA] fix *stopped for CLI commands
On Tuesday 10 February 2009 23:44:28 Marc Khouzam wrote:
> > Vladimir Prus
> > Sent: Friday, February 06, 2009 2:45 AM
> > To: gdb-patches@sources.redhat.com; Nick Roberts; Marc Khouzam
> > Subject: [RFA] fix *stopped for CLI commands
> >
> > This patch fixes this by making MI observer print frame
> > again, into MI uiout,
> > if necessary.
>
> I've applied the patch and tried it out. It works as expected
> but it 'allowed' me to run into another problem.
>
> When issuing a CLI command that triggers a ^running event,
> the token id is not part of the ^running event anymore.
> You can see this in the summarized session below.
>
> 26-exec-run
> [...]
> 26^running <-- token is there for MI command
> *running,thread-id="all"
> (gdb)
> [...]
> (gdb)
> 27run
> &"run\n"
> [...]
> ^running <-- token is not there for CLI command
> *running,thread-id="all"
> (gdb)
>
> I never noticed this before because DSF-GDB never uses CLI commands
> that would trigger a ^running. So, to DSF-GDB it looks like the
> CLI command never completed.
Why does DSF use tokens at all? I probably can fix this up, but I am
not sure the tokens are really necessary. Does anybody envision a case
where GDB would report command results out-of-order?
- Volodya