threads/1609: gdb 6.0 detach after fork

[thomas dot lavergne at jrc dot it]

>Number:         1609
>Category:       threads
>Synopsis:       gdb 6.0 detach after fork
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          change-request
>Submitter-Id:   net
>Arrival-Date:   Thu Apr 08 16:18:01 UTC 2004
>Closed-Date:
>Last-Modified:
>Originator:     thomas.lavergne@jrc.it
>Release:        unknown-1.0
>Organization:
>Environment:
Linux RedHat9 kernel-2.4.20-18.9smp
gdb 6.0 (compiled with gcc3.3.3)
executable to be debugged is compiled with gcc 3.2.2-5
>Description:
My sysadmin just upgraded to gdb 6.0 and I run into the following problem.

IMPORTANT NOTE: it was allright with the gdb-5.3post-0.20021129.18

My C program uses the popen command to fork the gnu cpp while opening an input file. The command is:
yyin = popen(cmd, "r");
with cmd being "/usr/bin/cpp -D<macro_defs> filename"

1) the program works fine when I execute it stand-alone (without trying to debug it with gdb). The input file is perfectly parsed and the behaviours are ok.

2) under gdb (gdb --args myprogram), when I try to 'run', the program executes ok but gdb reports an error related to fork:
*) it outputs at the beggining of execution:
   "Detaching after fork from child process 24945."
*) after my program finishes it writes:
   warning: Unexpected waitpid result 000000 when waiting for vfork-done
   ptrace: No such process.
   Cannot remove breakpoints because program is no longer writable.
  It might be running in another process.
  Further execution is probably impossible.
  0x420acda8 in vfork () from /lib/tls/libc.so.6

I tried to locate the process 24945, but it is dead after the execution.
It is important to note that the forked cpp *does* its job and that the program works fine.

3) it means that I can use breakpoints till the call to popen. I can break just before instruction popen(cmd,"r") and then, by asking 'n', gdb reports it detach and the same "warning: Unexpected waitpid result..." stuff appears at the end of the execution.
  
>How-To-Repeat:
It is too big a code for me to create a simple test-case. Moreover, I did not wrote this part of the code and know very little about threads.
It is perhaps related to 
http://www.redhat.com/archives/fedora-test-list/2004-January/msg00179.html 

If you need me to test a test_case on my configuration, please ask.

Thomas
>Fix:

>Release-Note:
>Audit-Trail:
>Unformatted: