[RFA] Make QUIT work

[RFA] Make QUIT work

[Michael Snyder]

[-- Attachment #1: Type: text/plain, Size: 269 bytes --]

Elena, here's something that must have fallen thru the cracks
at some point.  handle_sigint needs to set quit_flag -- or else
QUIT won't quit.  As is currently, QUIT won't work unles
immediate_quit is set -- and I'm pretty sure that's not how
it was intended to work.


[-- Attachment #2: quit --]
[-- Type: text/plain, Size: 704 bytes --]

2004-05-08  Michael Snyder  <msnyder@redhat.com>

	* event-top.c (handle_sigint): Set quit_flag, so that QUIT
	will work the next time it's invoked.

Index: event-top.c
===================================================================
RCS file: /cvs/src/src/gdb/event-top.c,v
retrieving revision 1.33
diff -p -r1.33 event-top.c
*** event-top.c	28 Feb 2004 18:04:36 -0000	1.33
--- event-top.c	8 May 2004 03:22:55 -0000
*************** void
*** 950,955 ****
--- 950,956 ----
  handle_sigint (int sig)
  {
    signal (sig, handle_sigint);
+   quit_flag=1;
  
    /* If immediate_quit is set, we go ahead and process the SIGINT right
       away, even if we usually would defer this to the event loop. The

Re: [RFA] Make QUIT work

[Daniel Jacobowitz]

On Sat, May 08, 2004 at 03:27:16AM +0000, Michael Snyder wrote:
> Elena, here's something that must have fallen thru the cracks
> at some point.  handle_sigint needs to set quit_flag -- or else
> QUIT won't quit.  As is currently, QUIT won't work unles
> immediate_quit is set -- and I'm pretty sure that's not how
> it was intended to work.
> 

See the patch Joel submitted for this last August.  Elena deferred it
to after 6.1 because just setting quit_flag invokes both sets of event
handling machinery, potentially quitting twice.

> 2004-05-08  Michael Snyder  <msnyder@redhat.com>
> 
> 	* event-top.c (handle_sigint): Set quit_flag, so that QUIT
> 	will work the next time it's invoked.
> 
> Index: event-top.c
> ===================================================================
> RCS file: /cvs/src/src/gdb/event-top.c,v
> retrieving revision 1.33
> diff -p -r1.33 event-top.c
> *** event-top.c	28 Feb 2004 18:04:36 -0000	1.33
> --- event-top.c	8 May 2004 03:22:55 -0000
> *************** void
> *** 950,955 ****
> --- 950,956 ----
>   handle_sigint (int sig)
>   {
>     signal (sig, handle_sigint);
> +   quit_flag=1;
>   
>     /* If immediate_quit is set, we go ahead and process the SIGINT right
>        away, even if we usually would defer this to the event loop. The


-- 
Daniel Jacobowitz

Re: [RFA] Make QUIT work

[Joel Brobecker]

> See the patch Joel submitted for this last August.  Elena deferred it
> to after 6.1 because just setting quit_flag invokes both sets of event
> handling machinery, potentially quitting twice.

Yes, I look at this a little bit more after you reviewed my patch,
and omigod yes, it quits twice. And unfortunately, I wasn't too sure
by reading the code how this should be made to work (wasn't able to
deduce the design, CRC too weak :-).

We're after 6.1 BTW :-) but poor Elena seems to be drowned with other
stuff, we've hardly seen any message from her for a while :-/.

-- 
Joel

Re: [RFA] Make QUIT work

[Michael Snyder]

Joel Brobecker wrote:
See the patch Joel submitted for this last August.  Elena deferred it
to after 6.1 because just setting quit_flag invokes both sets of event
handling machinery, potentially quitting twice.


Yes, I look at this a little bit more after you reviewed my patch,
and omigod yes, it quits twice. And unfortunately, I wasn't too sure
by reading the code how this should be made to work (wasn't able to
deduce the design, CRC too weak :-).
We're after 6.1 BTW :-) but poor Elena seems to be drowned with other
stuff, we've hardly seen any message from her for a while :-/.
OK, I see the "quit twice" problem too, but it leads one to ask:
which is worse?  Quitting once too often, or not quitting at all?
Right now, in some situations at least, you can't quit at all.
That's bad.

Re: [RFA] Make QUIT work

[Joel Brobecker]

> OK, I see the "quit twice" problem too, but it leads one to ask:
> which is worse?  Quitting once too often, or not quitting at all?
> 
> Right now, in some situations at least, you can't quit at all.
> That's bad.

I agree.

-- 
Joel

Re: [RFA] Make QUIT work

[Andrew Cagney]

On Sat, May 08, 2004 at 03:27:16AM +0000, Michael Snyder wrote:

Elena, here's something that must have fallen thru the cracks
at some point.  handle_sigint needs to set quit_flag -- or else
QUIT won't quit.  As is currently, QUIT won't work unles
immediate_quit is set -- and I'm pretty sure that's not how
it was intended to work.


See the patch Joel submitted for this last August.  Elena deferred it
to after 6.1 because just setting quit_flag invokes both sets of event
handling machinery, potentially quitting twice.
http://sources.redhat.com/ml/gdb-patches/2004-02/msg00530.html

given quit_flag hasn't worked for 4+ years, we might as well delete 
that.  At least that way there won't be confusion over what needs to be 
fixed.

Andrew

Re: [RFA] Make QUIT work

[Michael Snyder]

Andrew Cagney wrote:
On Sat, May 08, 2004 at 03:27:16AM +0000, Michael Snyder wrote:

Elena, here's something that must have fallen thru the cracks
at some point.  handle_sigint needs to set quit_flag -- or else
QUIT won't quit.  As is currently, QUIT won't work unles
immediate_quit is set -- and I'm pretty sure that's not how
it was intended to work.


See the patch Joel submitted for this last August.  Elena deferred it
to after 6.1 because just setting quit_flag invokes both sets of event
handling machinery, potentially quitting twice.


http://sources.redhat.com/ml/gdb-patches/2004-02/msg00530.html

given quit_flag hasn't worked for 4+ years, we might as well delete 
that.  At least that way there won't be confusion over what needs to be 
fixed.
I don't agree.  If gdb hangs and you can't quit by typing ^C,
that's a bug.

Re: [RFA] Make QUIT work

[Daniel Jacobowitz]

On Tue, May 11, 2004 at 01:38:24AM +0000, Michael Snyder wrote:
> Andrew Cagney wrote:
> >>On Sat, May 08, 2004 at 03:27:16AM +0000, Michael Snyder wrote:
> >>
> >>>>Elena, here's something that must have fallen thru the cracks
> >>>>at some point.  handle_sigint needs to set quit_flag -- or else
> >>>>QUIT won't quit.  As is currently, QUIT won't work unles
> >>>>immediate_quit is set -- and I'm pretty sure that's not how
> >>>>it was intended to work.
> >>>>
> >>
> >>
> >>See the patch Joel submitted for this last August.  Elena deferred it
> >>to after 6.1 because just setting quit_flag invokes both sets of event
> >>handling machinery, potentially quitting twice.
> >
> >
> > http://sources.redhat.com/ml/gdb-patches/2004-02/msg00530.html
> >
> >given quit_flag hasn't worked for 4+ years, we might as well delete 
> >that.  At least that way there won't be confusion over what needs to be 
> >fixed.
> 
> I don't agree.  If gdb hangs and you can't quit by typing ^C,
> that's a bug.

In fact, I've gotten multiple bug reports about it.

-- 
Daniel Jacobowitz

Re: [RFA] Make QUIT work

[Andrew Cagney]

Andrew Cagney wrote:

On Sat, May 08, 2004 at 03:27:16AM +0000, Michael Snyder wrote:

Elena, here's something that must have fallen thru the cracks
at some point.  handle_sigint needs to set quit_flag -- or else
QUIT won't quit.  As is currently, QUIT won't work unles
immediate_quit is set -- and I'm pretty sure that's not how
it was intended to work.


See the patch Joel submitted for this last August.  Elena deferred it
to after 6.1 because just setting quit_flag invokes both sets of event
handling machinery, potentially quitting twice.


http://sources.redhat.com/ml/gdb-patches/2004-02/msg00530.html

given quit_flag hasn't worked for 4+ years, we might as well delete that.  At least that way there won't be confusion over what needs to be fixed.


I don't agree.  If gdb hangs and you can't quit by typing ^C,
that's a bug. 
(Hmm, no one is saying that this isn't a bug.)

Anyway, lets not drop this again.  Given that you appear to have been 
tasked with the problem, can I leave it to you to to submit the 
correct-fix (as sketched out in the above discussion thread)?

Andrew

Re: [RFA] Make QUIT work

[Michael Snyder]

Andrew Cagney wrote:
Andrew Cagney wrote:

On Sat, May 08, 2004 at 03:27:16AM +0000, Michael Snyder wrote:

Elena, here's something that must have fallen thru the cracks
at some point.  handle_sigint needs to set quit_flag -- or else
QUIT won't quit.  As is currently, QUIT won't work unles
immediate_quit is set -- and I'm pretty sure that's not how
it was intended to work.


See the patch Joel submitted for this last August.  Elena deferred it
to after 6.1 because just setting quit_flag invokes both sets of event
handling machinery, potentially quitting twice.




http://sources.redhat.com/ml/gdb-patches/2004-02/msg00530.html

given quit_flag hasn't worked for 4+ years, we might as well delete 
that.  At least that way there won't be confusion over what needs to 
be fixed.


I don't agree.  If gdb hangs and you can't quit by typing ^C,
that's a bug. 


(Hmm, no one is saying that this isn't a bug.)

Anyway, lets not drop this again.  Given that you appear to have been 
tasked with the problem,
???

can I leave it to you to to submit the 
correct-fix (as sketched out in the above discussion thread)?
No, sorry, I had nothing to do with this.  I'm just the latest
to notice that it is broken.