vCont packets for next and step

vCont packets for next and step

[Kishore Paul]

Hi All,

How does the gdb next and step work? I am seeing a problem where next 
only sends one vCont packet, similar to stepi though I was expecting a 
set of vCont packets to be sent similar to a step.

Thanks,
Kishore

next:
-----
Server received packet: [vCont;s:1]
Server Sending packet:  [S05]

stepi:
-----
Server received packet: [vCont;s:1;c]
Server Sending packet:  [S05]

step: 0x100010bc-0x100010cc
----------------------------
Server received packet: [vCont;s:1]
Server Sending packet:  [S05]
Server received packet: [vCont;s:1;c]
Server Sending packet:  [S05]
Server received packet: [vCont;s:1;c]
Server Sending packet:  [S05]
Server received packet: [vCont;s:1;c]
Server Sending packet:  [S05]
Server received packet: [vCont;s:1;c]
Server Sending packet:  [S05]

Re: vCont packets for next and step

[Daniel Jacobowitz]

On Mon, Feb 12, 2007 at 03:36:26PM -0600, Kishore Paul wrote:
> Hi All,
> 
> How does the gdb next and step work? I am seeing a problem where next 
> only sends one vCont packet, similar to stepi though I was expecting a 
> set of vCont packets to be sent similar to a step.

That means that, after the single stepi, the "next" thinks it has
reached a different source line.  So it stops.

-- 
Daniel Jacobowitz
CodeSourcery

Re: vCont packets for next and step

[Kishore Paul]

Should the gdb host not be sending multiple vCont step packets for next, 
just like it does for step?

Thanks,
Kishore

Daniel Jacobowitz wrote:

>On Mon, Feb 12, 2007 at 03:36:26PM -0600, Kishore Paul wrote:
>  
>
>>Hi All,
>>
>>How does the gdb next and step work? I am seeing a problem where next 
>>only sends one vCont packet, similar to stepi though I was expecting a 
>>set of vCont packets to be sent similar to a step.
>>    
>>
>
>That means that, after the single stepi, the "next" thinks it has
>reached a different source line.  So it stops.
>
>  
>

Re: vCont packets for next and step

[Daniel Jacobowitz]

On Mon, Feb 12, 2007 at 03:58:54PM -0600, Kishore Paul wrote:
> Should the gdb host not be sending multiple vCont step packets for next, 
> just like it does for step?

Yes.  Are you sure those two were from the same location, i.e. in both
cases more than one step was necessary?

Failing that something may be wrong with your backtracing.

-- 
Daniel Jacobowitz
CodeSourcery

Re: vCont packets for next and step

[Joel Brobecker]

> Should the gdb host not be sending multiple vCont step packets for next, 
> just like it does for step?

I don't know about the "vCont" packet because I'm unfamiliar with
the remote protocol. However, the answer to your question about steps
is that the debugger will stop stepping as soon as it has reached
a different line. The only exception to that is when the stepi makes
us land inside a function call, in which case the debugger sets a
breakpoint at the return address and does a continue until we reach
it. You might find the output of "set debug infrun 1" very useful.

-- 
Joel