Re: popen

Re: popen

[Bruce Korb]

Chris Hamilton wrote:
> 
> Bruce,
> I came across your September 24th posting about gdb throwing off a
> "Cannot find thread 2049" when it hits a popen
> (http://sources.redhat.com/ml/insight/2002-q3/msg00185.html).
> I'm currently experiencing the same problem in my application, and I was
> wondering if you had any advice on how to solve it.

printf debugging, I guess.  Essentially, the problem boils down
to there not being a good implementation of multithreaded
process management for GDB on i386.  If you are running UnixWare,
then Debug works great.  Otherwise, you might try a supported
multithread platform (Solaris on Sparc??).

Re: popen

[Daniel Jacobowitz]

On Mon, Jan 20, 2003 at 01:30:04PM -0800, Bruce Korb wrote:
> Chris Hamilton wrote:
> > 
> > Bruce,
> > I came across your September 24th posting about gdb throwing off a
> > "Cannot find thread 2049" when it hits a popen
> > (http://sources.redhat.com/ml/insight/2002-q3/msg00185.html).
> > I'm currently experiencing the same problem in my application, and I was
> > wondering if you had any advice on how to solve it.
> 
> printf debugging, I guess.  Essentially, the problem boils down
> to there not being a good implementation of multithreaded
> process management for GDB on i386.  If you are running UnixWare,
> then Debug works great.  Otherwise, you might try a supported
> multithread platform (Solaris on Sparc??).

Popen should not present this problem.  I don't know why it does, so we
need more information.

Meanwhile multi-process debugging is forthcoming.  The Linux kernel
patches are already in and I have more GDB patches queued.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: popen

[Bruce Korb]

Daniel Jacobowitz wrote:
> 
> On Mon, Jan 20, 2003 at 01:30:04PM -0800, Bruce Korb wrote:
> > Chris Hamilton wrote:
> > >
> > > Bruce,
> > > I came across your September 24th posting about gdb throwing off a
> > > "Cannot find thread 2049" when it hits a popen
> > > (http://sources.redhat.com/ml/insight/2002-q3/msg00185.html).
> > > I'm currently experiencing the same problem in my application, and I was
> > > wondering if you had any advice on how to solve it.

> Popen should not present this problem.  I don't know why it does, so we
> need more information.

It seems to occur when using popen, but not when I do all
the fork/exec stuff myself.  I'll see if a trivial popen
example duplicates the problem, but I would guess that many
would have complained by now if it were that easy.

> Meanwhile multi-process debugging is forthcoming.  The Linux kernel
> patches are already in and I have more GDB patches queued.

Excellent!!  Thank you very much!

Re: popen

[Chris Hamilton]

[-- Attachment #1: Type: text/plain, Size: 1253 bytes --]

The problem seems to come when I include the odbc libraries included 
with redhat 7.3.  I've included a test .c file that will cause the 
problem when run with:
cc popen_test.c -o a.out -g -lodbc


Bruce Korb wrote:

>Daniel Jacobowitz wrote:
>  
>
>>On Mon, Jan 20, 2003 at 01:30:04PM -0800, Bruce Korb wrote:
>>    
>>
>>>Chris Hamilton wrote:
>>>      
>>>
>>>>Bruce,
>>>>I came across your September 24th posting about gdb throwing off a
>>>>"Cannot find thread 2049" when it hits a popen
>>>>(http://sources.redhat.com/ml/insight/2002-q3/msg00185.html).
>>>>I'm currently experiencing the same problem in my application, and I was
>>>>wondering if you had any advice on how to solve it.
>>>>        
>>>>
>
>  
>
>>Popen should not present this problem.  I don't know why it does, so we
>>need more information.
>>    
>>
>
>It seems to occur when using popen, but not when I do all
>the fork/exec stuff myself.  I'll see if a trivial popen
>example duplicates the problem, but I would guess that many
>would have complained by now if it were that easy.
>
>  
>
>>Meanwhile multi-process debugging is forthcoming.  The Linux kernel
>>patches are already in and I have more GDB patches queued.
>>    
>>
>
>Excellent!!  Thank you very much!
>
>  
>

[-- Attachment #2: popen_test.c --]
[-- Type: text/plain, Size: 218 bytes --]

#include <stdio.h>

int main()
{

FILE *tpPipeFp;

if((tpPipeFp = popen("test", "w")) == NULL)
    {
    printf("Can't open pipe\n");
    exit();
    }
else
    printf("I've opened the pipe\n");

pclose(tpPipeFp);

}

Re: popen

[Chris Hamilton]

I didn't really think that odbc was causing the problem, so I linked in 
another threaded library and got the same result.  I think the general 
case is that gdb loses it's brains anytime a threaded library is linked 
in, a breakpoint is set and a popen is passed over.  I've tried this 
with a few safe libraries ( ie. libraries I didn't write ;) ), and this 
seems to hold up.


Chris Hamilton wrote:

> The problem seems to come when I include the odbc libraries included 
> with redhat 7.3.  I've included a test .c file that will cause the 
> problem when run with:
> cc popen_test.c -o a.out -g -lodbc
>
>
> Bruce Korb wrote:
>
>> Daniel Jacobowitz wrote:
>>  
>>
>>> On Mon, Jan 20, 2003 at 01:30:04PM -0800, Bruce Korb wrote:
>>>   
>>>
>>>> Chris Hamilton wrote:
>>>>     
>>>>
>>>>> Bruce,
>>>>> I came across your September 24th posting about gdb throwing off a
>>>>> "Cannot find thread 2049" when it hits a popen
>>>>> (http://sources.redhat.com/ml/insight/2002-q3/msg00185.html).
>>>>> I'm currently experiencing the same problem in my application, and 
>>>>> I was
>>>>> wondering if you had any advice on how to solve it.
>>>>>       
>>>>
>>
>>  
>>
>>> Popen should not present this problem.  I don't know why it does, so we
>>> need more information.
>>>   
>>
>>
>> It seems to occur when using popen, but not when I do all
>> the fork/exec stuff myself.  I'll see if a trivial popen
>> example duplicates the problem, but I would guess that many
>> would have complained by now if it were that easy.
>>
>>  
>>
>>> Meanwhile multi-process debugging is forthcoming.  The Linux kernel
>>> patches are already in and I have more GDB patches queued.
>>>   
>>
>>
>> Excellent!!  Thank you very much!
>>
>>  
>>
>------------------------------------------------------------------------
>
>#include <stdio.h>
>
>int main()
>{
>
>FILE *tpPipeFp;
>
>if((tpPipeFp = popen("test", "w")) == NULL)
>    {
>    printf("Can't open pipe\n");
>    exit();
>    }
>else
>    printf("I've opened the pipe\n");
>
>pclose(tpPipeFp);
>
>}
>
>  
>