Debugger MI port

Alexander Neundorf neundorf at kde.org
Mon Jan 30 18:23:05 UTC 2006


On Monday 30 January 2006 13:37, Vladimir Prus wrote:
> Hi,
> I've just created a branch for the work to make KDevelop's debugger
> frontend use gdb's MI protocol, as opposed to command line protocol.

Cool :-)
Your debugger work is really great :-)

...
> So, it's too early to try the branch for real work, but if anybody wants to
> help with coding, or just watch what's going on, he's welcome.

Currently I'm occupied with other things...

Bye
Alex
-- 
Work: alexander.neundorf AT jenoptik.com - http://www.jenoptik-los.de
Home: neundorf AT kde.org                - http://www.kde.org
      alex AT neundorf.net               - http://www.neundorf.net


>From "Matt Rogers <mattr"@kde.org Mon Jan 30 19:48:21 2006
Received: from mout1.mail.vrmd.de ([81.28.224.19])
	by barney.cs.uni-potsdam.de with esmtp (Exim 3.36 #2)
	id 1F3e4T-0000lh-00
	for kdevelop-devel at barney.cs.uni-potsdam.de; Mon, 30 Jan 2006 19:48:13 +0100
Received: from [81.28.224.26] (helo=mail.variomedia.de)
	by mout1.mail.vrmd.de with esmtpa (Exim 4.43)
	id 1F3e4U-0000kh-0z
	for kdevelop-devel at barney.cs.uni-potsdam.de; Mon, 30 Jan 2006 19:48:14 +0100
Received: from smtp106.sbc.mail.mud.yahoo.com ([68.142.198.205])
	by mail.variomedia.de with smtp (Exim 4.43)
	id 1F3e4R-0007IZ-1V
	for kdevelop-devel at kdevelop.org; Mon, 30 Jan 2006 19:48:11 +0100
Received: (qmail 32229 invoked from network); 30 Jan 2006 18:42:01 -0000
Received: from unknown (HELO localhost) (mgrogers at sbcglobal.net@70.255.232.184 with login)
  by smtp106.sbc.mail.mud.yahoo.com with SMTP; 30 Jan 2006 18:42:01 -0000
From: Matt Rogers <mattr at kde.org>
To: kdevelop-devel at kdevelop.org
Subject: Re: Debugger MI port
Message-ID: <20060130184611.GA11180 at shocker.internal>
References: <200601301537.00507.ghost at cs.msu.su>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <200601301537.00507.ghost at cs.msu.su>
User-Agent: Mutt/1.4.2.1i
Sender: kdevelop-devel-admin at barney.cs.uni-potsdam.de
Errors-To: kdevelop-devel-admin at barney.cs.uni-potsdam.de
X-BeenThere: kdevelop-devel at barney.cs.uni-potsdam.de
X-Mailman-Version: 2.0.13
Precedence: bulk
Reply-To: kdevelop-devel at kdevelop.org
List-Help: <mailto:kdevelop-devel-request at barney.cs.uni-potsdam.de?subject=help>
List-Post: <mailto:kdevelop-devel at barney.cs.uni-potsdam.de>
List-Subscribe: <http://barney.cs.uni-potsdam.de/mailman/listinfo/kdevelop-devel>,
	<mailto:kdevelop-devel-request at barney.cs.uni-potsdam.de?subject=subscribe>
List-Id: KDevelop Developer Mailinglist <kdevelop-devel.barney.cs.uni-potsdam.de>
List-Unsubscribe: <http://barney.cs.uni-potsdam.de/mailman/listinfo/kdevelop-devel>,
	<mailto:kdevelop-devel-request at barney.cs.uni-potsdam.de?subject=unsubscribe>
List-Archive: <http://barney.cs.uni-potsdam.de/mailman/private/kdevelop-devel/>
Date: Mon Jan 30 19:49:06 2006
X-Original-Date: Mon, 30 Jan 2006 12:46:11 -0600

On Mon, Jan 30, 2006 at 03:37:00PM +0300, Vladimir Prus wrote:
> 
> Hi,
> I've just created a branch for the work to make KDevelop's debugger frontend 
> use gdb's MI protocol, as opposed to command line protocol.
> 
> The MI protcol is designed to be machine readable, and it's much easier to 
> extract information from MI responses, which will simplify implementation of 
> new features. For example, I already have a code to report, for watchpoints, 
> the old and the new values of the watched expression. Extracting the return 
> value of a function, on "finish" command, is also trivial and so on.
> 
> MI will also increase robustness. Right now, all error messages from gdb are 
> just ignored, without notification with the user. In addition, sometimes 
> error messages mix with normal output and blow up the parser. Both problems 
> will be gone -- I intend for all error messages from gdb to result in error 
> message box.
> 
> Of course, there are downsides. MI at the moment is immature protocol and has 
> certain omissions. The documentation is pretty weak as well. However I hope 
> I'll be able to workaround this.
> 
> The branch is:
> 
>     http://websvn.kde.org/branches/work/kdevelop-debugger-mi/
> 
> At the current state you can set breakpoint on main, run the program, step 
> though it, and watch variables. Breakpoints are half-broken, everything else 
> was not tested, and the only testcase I've tried with is:
> 
>     http://websvn.kde.org/branches/work/kdevelop-debugger-mi/tests/two_module/
> 
> So, it's too early to try the branch for real work, but if anybody wants to 
> help with coding, or just watch what's going on, he's welcome.
> 
> - Volodya
> 

Nice work. Out of curiousity, are you working with upstream at all on
getting omissions in MI added?

Thanks
--
Matt





More information about the KDevelop-devel mailing list