Re: [RFC] [PATCH] [PR gdb/29796] Fix gdbserver delay before quitting in extended-remote

[Kevin Buettner <kevinb@redhat.com>]

On Sun, 15 Oct 2023 16:33:16 +0000
Shivam Saxena <saxens12@mcmaster.ca> wrote:

> >> Investigating PR29796, where it was noticed that there was a delay when trying to quit the gdb session if connected to gdbserver via stdio. It was also mentioned that using --once​ when launching gdbserver​ would fix this problem and a recommendation was made to always apply --once​ when launching gdbserver​ with stdio​.
> >>
> >> I've explicitly set the run_once​ flag when stdio​ is detected at launch of gdbserver. This does fix the problem as suggested in the bug since the conditional at gdbserver/server.cc:3958​ is always triggered.
> >> ```
> >>          if (run_once || (!extended_protocol && !target_running ()))
> >>             throw_quit ("Quit");
> >> ```
> >> However, would it be better to create a new flag for this to 
> >> facilitate quitting despite extended_protocol​ instead of piggybacking on --once​ logic?  
> 
> > I don't see a problem with using 'run_once' for this purpose.  But if there's something I'm missing, please explain...  
> 
> By the conditional it seems the intent was to not allow quitting immediately if in extended protocol mode. We are adding another condition to quit - if launched in stdio. But nothing in this logic alone would make that apparent. (no comment, no flag). I thought this reduced traceability. But I might be overthinking things. So if the fix seems fine for this purpose I'll fix the patch and reply. It seems I might have to use a different email account to submit the patch via git send-email. 

I think it's fine.

(But don't forget to fix the nit that I found regarding the comment.)

Kevin