Insight and breakpoints

Insight and breakpoints

[Owen Savill]

Hello,

I have been using Insight, with varying degrees of success, for a while
now. It has always troubled me that breakpoints can't be set while the
program is running. I now discover that Code Medic can set breakpoints on
the fly. Will we be seeing this feature in Insight ?

Thanks,
Owen

P.S. I've just submitted two bug reports (insight/63 and 64) as a 'net'
submitter (I thought that 'spam' might get ignored !). Is this okay, or
should I take out a user name and re-submit them.

Re: Insight and breakpoints

[Keith Seitz]

Owen Savill wrote:
> 
> Hello,
> 
> I have been using Insight, with varying degrees of success, for a while
> now. It has always troubled me that breakpoints can't be set while the
> program is running. I now discover that Code Medic can set breakpoints on
> the fly. Will we be seeing this feature in Insight ?

The problem is not that Insight cannot insert breakpoints while the
inferior runs. Rather, it is gdb (the "backend") that cannot do this...

In order to do this sort of thing, gdb would have to be taught how to do
this. Specifically, gdb would have to stop the inferior, insert the
breakpoint, and then continue the inferior. This may not be difficult,
but with the way gdb (and Insight) is architected, I'll bet this would
be more difficult than not.

Of course, the other problem is that not all targets can be interrupted
-- monitor and other remote/stub targets, in particular. We would need
to start tracking the abilities of every target interface in order to do
this. (This is pretty easy when you only have a handful of targets; not
quite so easy for all the targets -- and interfaces -- that gdb
supports.)

Of course, we in gui-land have always wanted gdb to maintain a list of
target/interface features so that we could enable/disable GUI features
at runtime... It would be one step closer to getting rid of Insight's
(ugly) "target database" in targetselection code. :-)

In short: ask about this on the gdb list and don't expect to see this
added anytime soon. :-(

Sorry,
Keith

Re: Insight and breakpoints

[Fernando Nasser]

Owen Savill wrote:
> 
> Hello,
> 
> I have been using Insight, with varying degrees of success, for a while
> now. It has always troubled me that breakpoints can't be set while the
> program is running. I now discover that Code Medic can set breakpoints on
> the fly. Will we be seeing this feature in Insight ?
> 
> Thanks,
> Owen
> 

This came up before.  We would need to stop the inferior program (the 
program being debugged), insert the breakpoint and continue it.

The problem is that this would require some selective activation of the 
GUI and "lots" of care to prevent GDB notifications (that are sent back
as a result of these actions) of causing havoc in the GUI.

It is a nice feature, just not trivial to implement at the moment.
Things may change in the future though.


> P.S. I've just submitted two bug reports (insight/63 and 64) as a 'net'
> submitter (I thought that 'spam' might get ignored !). Is this okay, or
> should I take out a user name and re-submit them.

The problem with that is that the messages that are sent regarding the
bug reports keep being sent back.  We have no way to get in touch with you
to get more details, to ask you to try something etc. and also get lots
of e-mail failures back.


-- 
Fernando Nasser
Red Hat Canada Ltd.                     E-Mail:  fnasser@redhat.com
2323 Yonge Street, Suite #300
Toronto, Ontario   M4P 2C9

Re: Insight and breakpoints

[Tom Tromey]

[ I think this bounced the first time ]

>> P.S. I've just submitted two bug reports (insight/63 and 64) as a
>> 'net' submitter (I thought that 'spam' might get ignored !). Is
>> this okay, or should I take out a user name and re-submit them.

Fernando> The problem with that is that the messages that are sent
Fernando> regarding the bug reports keep being sent back.  We have no
Fernando> way to get in touch with you to get more details, to ask you
Fernando> to try something etc. and also get lots of e-mail failures
Fernando> back.

He's just talking about the `submitter-id' field, which is really
pretty useless for a public Gnats database.  His PRs do have his email
address in them.  So it really is ok as-is.

Tom

Re: Insight and breakpoints

[Tom Tromey]

>>>>> "Owen" == Owen Savill <Owen_Savill@cursor-system.com> writes:

Owen> I have been using Insight, with varying degrees of success, for
Owen> a while now. It has always troubled me that breakpoints can't be
Owen> set while the program is running. I now discover that Code Medic
Owen> can set breakpoints on the fly. Will we be seeing this feature
Owen> in Insight ?

How does Code Medic do it?  I can think of two ways.

One way is to stop the inferior, insert the breakpoints, and then
continue.  The GUI could do all of this without the user noticing.

The other way would be to wait until the program next stops and only
then insert the breakpoints.

The first approach is probably better.  My guess is that in the
current Insight framework it would probably be hard to implement
invisibly.

Owen> P.S. I've just submitted two bug reports (insight/63 and 64) as
Owen> a 'net' submitter (I thought that 'spam' might get ignored
Owen> !). Is this okay, or should I take out a user name and re-submit
Owen> them.

That is fine.

Tom