[GDB/MI] missing token in "stopped" output after -exec-run?

[GDB/MI] missing token in "stopped" output after -exec-run?

[Joel Brobecker]

Hello,

I noticed what looks like a change of behavior in MI and I was wondering
if this was intended. From my limited understanding of MI, it does look
like a bug, but it's hard to be sure for someone who's not familiar with
this protocol.

Consider the following program:

    int
    main (void)
    {
    }

Compile it with:

    % gcc -g -o foo foo.c

Then break on main, and run to the breakpoint. Make sure the -exec-run
command has a token. Here is a transcript of my session:

    (gdb) 
    -break-insert -t main
    ^done,bkpt={number="1",type="breakpoint",disp="del",enabled="y",addr="0x08048332",func="main",file="foo.c",fullname="/home/no-backup/brobecke/head-merge/ex/foo.c",line="6",times="0",original-location="main"}
    (gdb) 
    232-exec-run
    232^running
    (gdb) 
    =thread-created,id="1"
    *stopped,reason="breakpoint-hit",disp="del",bkptno="1",thread-id="1",frame={addr="0x08048332",func="main",args=[],file="foo.c",fullname="/home/no-backup/brobecke/head-merge/ex/foo.c",line="6"}

A few weeks ago, the output looked like this:

    232-exec-run
    232^running
    (gdb) 
    232*stopped,reason="breakpoint-hit",disp="del",bkptno="1",thread-id="1",frame={addr="0x08048332",func="main",args=[],file="foo.c",fullname="/home/no-backup/brobecke/head-merge/ex/foo.c",line="6"}
    
The token was preserved.  Is the new behavior intended?

-- 
Joel

Re: [GDB/MI] missing token in "stopped" output after -exec-run?

[Vladimir Prus]

Joel Brobecker wrote:

....
>    
*stopped,reason="breakpoint-hit",disp="del",bkptno="1",thread-id="1",frame={addr="0x08048332",func="main",args=[],file="foo.c",fullname="/home/no-backup/brobecke/head-merge/ex/foo.c",line="6"}
> 
> A few weeks ago, the output looked like this:
> 
>     232-exec-run
>     232^running
>     (gdb)
>     232*stopped,reason="breakpoint-hit",disp="del",...
>     
> The token was preserved.  Is the new behavior intended?

Yes (and I've explicitly posted to the list about this change, quite some time before
actually changing things). The situation is:

1. In current gdb, the token actually provides no information (since only
on run command can be active at a time).
2. No frontend author stepped in to say they care.
3. In non-stop mode, keeping the token is almost impossible.
4. The implementation of printing of the token was a mess.

Does it cause any problems for you?

- Volodya

Re: [GDB/MI] missing token in "stopped" output after -exec-run?

[Joel Brobecker]

> Yes (and I've explicitly posted to the list about this change, quite
> some time before actually changing things). The situation is:

Oh sorry, I vaguely remember this discussion indeed.

> Does it cause any problems for you?

No problem whatsoever. I have a testcase in our testsuite that did check
for the token on the "*stopped" notification, but the intent was really
to check that a -var-update doesn't crash the debugger.  So I relaxed
a little our test to accept the absence of token and that was it for me.

The purpose of my message was to inform you of the change, in case
it was not intended and somehow slipped through testing. Glad to be
wrong :).

-- 
Joel

Re: [GDB/MI] missing token in "stopped" output after -exec-run?

[Daniel Jacobowitz]

On Wed, May 21, 2008 at 09:04:28AM +0400, Vladimir Prus wrote:
> Yes (and I've explicitly posted to the list about this change, quite some time before
> actually changing things).

Do you think we should list MI changes like this in NEWS, for any
frontend author who doesn't read the list?

-- 
Daniel Jacobowitz
CodeSourcery

Re: [GDB/MI] missing token in "stopped" output after -exec-run?

[Vladimir Prus]

Daniel Jacobowitz wrote:

> On Wed, May 21, 2008 at 09:04:28AM +0400, Vladimir Prus wrote:
>> Yes (and I've explicitly posted to the list about this change, quite some time before
>> actually changing things).
> 
> Do you think we should list MI changes like this in NEWS, for any
> frontend author who doesn't read the list?

We probably should. I'll review recent MI changes for NEWS-worthness.

- Volodya

Re: [GDB/MI] missing token in "stopped" output after -exec-run?

[Eli Zaretskii]

> From:  Vladimir Prus <vladimir@codesourcery.com>
> Date:  Wed, 21 May 2008 16:59:05 +0400
> 
> Daniel Jacobowitz wrote:
> 
> > On Wed, May 21, 2008 at 09:04:28AM +0400, Vladimir Prus wrote:
> >> Yes (and I've explicitly posted to the list about this change, quite some time before
> >> actually changing things).
> > 
> > Do you think we should list MI changes like this in NEWS, for any
> > frontend author who doesn't read the list?
> 
> We probably should. I'll review recent MI changes for NEWS-worthness.

Yes, please.