MI, remote protocol and error codes

MI, remote protocol and error codes

[Sascha]

Hi,

I'm currently using GDB (CS 6.5) in a situation like this:

C/C++ IDE ---[MI]---> GDB ---[Remote Protocol]---> Remote stub

Am I right and GDB does not care at all about remote error codes (ENN reply)
?
The remote stub supports breakpoints, but there are situations when a
breakpoint has to be rejected (depending on internal state). When the stub
replies with an error code, GDB won't resume - which is good - but it just
reports "Protocol error: Z0 (software-breakpoint) conflicting enabled
responses". The MI protocol does not report the error code, so the IDE does
not get to know why the breakpoint got rejected. 

Is there any way (or any plans) to report error codes over MI ? Or is there
a different solution ?

Thanks

Sascha

Re: MI, remote protocol and error codes

[Daniel Jacobowitz]

On Thu, Feb 22, 2007 at 08:26:32PM +0100, Sascha wrote:
> Hi,
> 
> I'm currently using GDB (CS 6.5) in a situation like this:
> 
> C/C++ IDE ---[MI]---> GDB ---[Remote Protocol]---> Remote stub
> 
> Am I right and GDB does not care at all about remote error codes (ENN reply)
> ?

I believe that's true.

> The remote stub supports breakpoints, but there are situations when a
> breakpoint has to be rejected (depending on internal state). When the stub
> replies with an error code, GDB won't resume - which is good - but it just
> reports "Protocol error: Z0 (software-breakpoint) conflicting enabled
> responses". The MI protocol does not report the error code, so the IDE does
> not get to know why the breakpoint got rejected. 

Are you sure you're replying with an error code, not with an empty
packet?  There should be no other way to get this message.

> Is there any way (or any plans) to report error codes over MI ? Or is there
> a different solution ?

There's no plan for this, but it's come up before, so something may be
necessary in the future.

-- 
Daniel Jacobowitz
CodeSourcery

AW: MI, remote protocol and error codes

[Sascha]

Hi,

>Are you sure you're replying with an error code, not with an empty
packet?  There should be no other way to get this message.

You are right. There was a bug in the stub code. The stub does now reply
with ENN reply and the message I get is: "Error accessing memory address
0x2001a4: (undocumented errno -1)".

It does mention an errno, but it's not the one I put in the ENN reply.

Well, I guess I will have to create hack for GDB to make it reply with the
information I need.

Thanks

Sascha