Multiple breakpoints - few problems [MI commands, gdb6.7.50.20080221]

Multiple breakpoints - few problems [MI commands, gdb6.7.50.20080221]

[Bogdan Slusarczyk]

Hi, I have few problems using <MUTLIPLE> breakpoints with MI commands, 
maybe you can help me.
1) to be sure: I can't delete sub-breakpoint, only disable it?
2) sometimes gdb complains that can't set sub-breakpoint (Cannot insert 
breakpoint <whole number>. Error accessing memory address <address>.) I 
want to disable such sub-breakpoint, but I can't because:
a) <whole number> it's not sub-breakpoint id
b) I could use <address> to find sub-breakpoint, but it's impossible 
because -break-list (or -break-info <whole number>) doesn't return any 
information about sub-breakpoints.
Is it a bug or I missed something? How can I list sub-breakpoints?

Best regards,
Bogdan

Re: Multiple breakpoints - few problems [MI commands, gdb6.7.50.20080221]

[Vladimir Prus]

Bogdan Slusarczyk wrote:

> Hi, I have few problems using <MUTLIPLE> breakpoints with MI commands,

Well, MI does not currently explicitly support multiple locations
breakpoint -- that is you cannot see the individual locations or
operate on them.

> maybe you can help me.
> 1) to be sure: I can't delete sub-breakpoint, only disable it?

Correct.

> 2) sometimes gdb complains that can't set sub-breakpoint (Cannot insert
> breakpoint <whole number>. Error accessing memory address <address>.) 

This is not supposed to happen in the first place. Any ideas what
causes this error?

> I 
> want to disable such sub-breakpoint, but I can't because:
> a) <whole number> it's not sub-breakpoint id
> b) I could use <address> to find sub-breakpoint, but it's impossible
> because -break-list (or -break-info <whole number>) doesn't return any
> information about sub-breakpoints.
> Is it a bug or I missed something? How can I list sub-breakpoints?

You cannot, in MI. Can you maybe provide a minimal self-contained example
so that we can figure the cause of your original error (and maybe use it
as testcase for making the error message more precise).

- Volodya


> 
> Best regards,
> Bogdan

Re: Multiple breakpoints - few problems [MI commands, gdb6.7.50.20080221]

[Bogdan Slusarczyk]

>> 2) sometimes gdb complains that can't set sub-breakpoint (Cannot insert
>> breakpoint <whole number>. Error accessing memory address <address>.) 
> 
> This is not supposed to happen in the first place. Any ideas what
> causes this error?
> Can you maybe provide a minimal self-contained example
> so that we can figure the cause of your original error (and maybe use it
> as testcase for making the error message more precise).

It will be difficult. It's SystemC project, to run it you have to have 
proper SystemC/HDL simulator :( I don't know if you know SystemC (I 
don't ;), see for example http://en.wikipedia.org/wiki/SystemC) It's 
some kind of strongly templated library. My SystemC project is made from 
my library (with debug) which is statically linked with SystemC library 
(without debug) - it's important, because even I have instantiated some 
of SystemC template in my library, gcc often gets symbols from non-debug 
SystemC library and I can't debug such variables (gdb complains 
something about incomplete types).
So I suppose two reasons:
1) I have instantiated template for two specializations (probably - 
SystemC can do it itself) and I try to set breakpoint inside it's 
method. I have problem to get debug information from SystemC library and 
maybe it's the same problem
2) probably simulator loads my library twice using dlopen (or similar), 
and gdb can't access memory for unknown reason

Regardless of this, don't you think it's good idea to adapt MI commands 
for <MULTIPLE> breakpoints? Do you consider it?

Regards,
Bogdan

Re: Multiple breakpoints - few problems [MI commands, gdb6.7.50.20080221]

[Vladimir Prus]

Bogdan Slusarczyk wrote:

>>> 2) sometimes gdb complains that can't set sub-breakpoint (Cannot insert
>>> breakpoint <whole number>. Error accessing memory address <address>.)
>> 
>> This is not supposed to happen in the first place. Any ideas what
>> causes this error?
>> Can you maybe provide a minimal self-contained example
>> so that we can figure the cause of your original error (and maybe use it
>> as testcase for making the error message more precise).
> 
> It will be difficult. It's SystemC project, to run it you have to have
> proper SystemC/HDL simulator :( I don't know if you know SystemC (I
> don't ;), see for example http://en.wikipedia.org/wiki/SystemC) It's
> some kind of strongly templated library. My SystemC project is made from
> my library (with debug) which is statically linked with SystemC library
> (without debug) - it's important, because even I have instantiated some
> of SystemC template in my library, gcc often gets symbols from non-debug
> SystemC library and I can't debug such variables (gdb complains
> something about incomplete types).
> So I suppose two reasons:
> 1) I have instantiated template for two specializations (probably -
> SystemC can do it itself) and I try to set breakpoint inside it's
> method. I have problem to get debug information from SystemC library and
> maybe it's the same problem
> 2) probably simulator loads my library twice using dlopen (or similar),
> and gdb can't access memory for unknown reason

So essentially, we don't know where there error comes from :-(

> 
> Regardless of this, don't you think it's good idea to adapt MI commands
> for <MULTIPLE> breakpoints? Do you consider it?

Of, course, it will be good. Unfortunately, at this point it's not the
biggest problem with MI, in my opinion, so I don't have an estimate when
I'll have time to address this.

- Volodya

Re: Multiple breakpoints - few problems [MI commands, gdb6.7.50.20080221]

[Bogdan Slusarczyk]

>> Regardless of this, don't you think it's good idea to adapt MI commands
>> for <MULTIPLE> breakpoints? Do you consider it?
> 
> Of, course, it will be good. Unfortunately, at this point it's not the
> biggest problem with MI, in my opinion, so I don't have an estimate when
> I'll have time to address this.

I hope it will be sooner than later :) I will be very happy with it ;) 
Thanks for help.

Best regards,
Bogdan