follow-fork-mode child on SuSE 9.2 (i586 and x86_64) not working

follow-fork-mode child on SuSE 9.2 (i586 and x86_64) not working

[Wimmer Andreas]

Hi,
I am using SuSE 9.2 both on i586 and x86_64 platforms and have some
troubles to get the follow-fork-mode=child feature working
The system is using the kernels provided by the distribution 
2.6.8-11.24 (based on 2.6.8 ?)
I tried with the distribution provided gdb (6.2.1) and also with 6.3
all with the same effect.
*) A debugged program starts up normally untill it reaches a fork
	statement
*) The child process is created the debugger detects the child and
	gives the message "Attaching after fork to child process .."
*) The child runs a while (it executed some of the initial code after
	the fork) and then blocks
*) After that GDB also blocks without any chance to do something

NOTE: the parent process is left in "S+" state and the child remains in
"T+" state.

Regards,
Andreas

Re: follow-fork-mode child on SuSE 9.2 (i586 and x86_64) not working

[Daniel Jacobowitz]

On Thu, May 19, 2005 at 03:54:40PM +0200, Wimmer Andreas wrote:
> Hi,
> I am using SuSE 9.2 both on i586 and x86_64 platforms and have some
> troubles to get the follow-fork-mode=child feature working
> The system is using the kernels provided by the distribution 
> 2.6.8-11.24 (based on 2.6.8 ?)
> I tried with the distribution provided gdb (6.2.1) and also with 6.3
> all with the same effect.
> *) A debugged program starts up normally untill it reaches a fork
> 	statement
> *) The child process is created the debugger detects the child and
> 	gives the message "Attaching after fork to child process .."
> *) The child runs a while (it executed some of the initial code after
> 	the fork) and then blocks
> *) After that GDB also blocks without any chance to do something
> 
> NOTE: the parent process is left in "S+" state and the child remains in
> "T+" state.

Can you reproduce this without using a SuSE kernel?  If not, you may
want to report this to SuSE.

-- 
Daniel Jacobowitz
CodeSourcery, LLC

Re: follow-fork-mode child on SuSE 9.2 (i586 and x86_64) not working

[Wimmer Andreas]

On Thu, 2005-05-19 at 10:22 -0400, Daniel Jacobowitz wrote:
> On Thu, May 19, 2005 at 03:54:40PM +0200, Wimmer Andreas wrote:
> > Hi,
> > I am using SuSE 9.2 both on i586 and x86_64 platforms and have some
> > troubles to get the follow-fork-mode=child feature working
> > The system is using the kernels provided by the distribution 
> > 2.6.8-11.24 (based on 2.6.8 ?)
> > I tried with the distribution provided gdb (6.2.1) and also with 6.3
> > all with the same effect.
> > *) A debugged program starts up normally untill it reaches a fork
> > 	statement
> > *) The child process is created the debugger detects the child and
> > 	gives the message "Attaching after fork to child process .."
> > *) The child runs a while (it executed some of the initial code after
> > 	the fork) and then blocks
> > *) After that GDB also blocks without any chance to do something
> > 
> > NOTE: the parent process is left in "S+" state and the child remains in
> > "T+" state.
> 
> Can you reproduce this without using a SuSE kernel?  If not, you may
> want to report this to SuSE.

Yes I can reproduce this on a a vanilla 2.6.11.10 kernel which I just
compiled for testing.
Same situation, gdb and child process are blocked, without any further
debugging possible.

---------------------
Andreas Wimmer
Joanneum Research

Re: follow-fork-mode child on SuSE 9.2 (i586 and x86_64) not working

[Daniel Jacobowitz]

On Thu, May 19, 2005 at 05:57:24PM +0200, Wimmer Andreas wrote:
> On Thu, 2005-05-19 at 10:22 -0400, Daniel Jacobowitz wrote:
> > On Thu, May 19, 2005 at 03:54:40PM +0200, Wimmer Andreas wrote:
> > > Hi,
> > > I am using SuSE 9.2 both on i586 and x86_64 platforms and have some
> > > troubles to get the follow-fork-mode=child feature working
> > > The system is using the kernels provided by the distribution 
> > > 2.6.8-11.24 (based on 2.6.8 ?)
> > > I tried with the distribution provided gdb (6.2.1) and also with 6.3
> > > all with the same effect.
> > > *) A debugged program starts up normally untill it reaches a fork
> > > 	statement
> > > *) The child process is created the debugger detects the child and
> > > 	gives the message "Attaching after fork to child process .."
> > > *) The child runs a while (it executed some of the initial code after
> > > 	the fork) and then blocks
> > > *) After that GDB also blocks without any chance to do something
> > > 
> > > NOTE: the parent process is left in "S+" state and the child remains in
> > > "T+" state.
> > 
> > Can you reproduce this without using a SuSE kernel?  If not, you may
> > want to report this to SuSE.
> 
> Yes I can reproduce this on a a vanilla 2.6.11.10 kernel which I just
> compiled for testing.
> Same situation, gdb and child process are blocked, without any further
> debugging possible.

I'm afraid that I haven't seen this problem, using kernels up to
2.6.11-rc3.  Can you give me a simple testcase and a transcript of the
session?

-- 
Daniel Jacobowitz
CodeSourcery, LLC