Can't attach to child using gdb-5.1.1 on Linux

Can't attach to child using gdb-5.1.1 on Linux

[Jesse Marlin]

I have a process that forks a child.  If I try to attach to the
child it reads the symbols and shows an accurate stack trace, but when
I set a breakpoint and try to continue I get the following:

(gdb) c
Continuing.
lin-lwp.c:1003: gdb-internal-error: lin_lwp_wait: Assertion `WIFSTOPPED (status) && WSTOPSIG (status) == SIGSTOP' failed.
An internal GDB error was detected.  This may make further
debugging unreliable.  Continue this debugging session? (y or n) n

Create a core file containing the current state of GDB? (y or n) n

Inferior GDB exited abnormally with code 1


I answered yes to the above also, but it did not improve the situation.  Has
anybody seen this behvior before.  I am using gcc-2.95.3 for compilation.

Re: Can't attach to child using gdb-5.1.1 on Linux

[Andrew Cagney]

> I have a process that forks a child.  If I try to attach to the
> child it reads the symbols and shows an accurate stack trace, but when
> I set a breakpoint and try to continue I get the following:
> 
> (gdb) c
> Continuing.
> lin-lwp.c:1003: gdb-internal-error: lin_lwp_wait: Assertion `WIFSTOPPED (status) && WSTOPSIG (status) == SIGSTOP' failed.
> An internal GDB error was detected.  This may make further
> debugging unreliable.  Continue this debugging session? (y or n) n
> 
> Create a core file containing the current state of GDB? (y or n) n
> 
> Inferior GDB exited abnormally with code 1
> 
> 
> I answered yes to the above also, but it did not improve the situation.  Has
> anybody seen this behvior before.  I am using gcc-2.95.3 for compilation.

(not sure if someone has responded to this)

The first thing to do is get a more recent GDB from:

	http://sources.redhat.com/gdb/current/

many problems have been fixed and if this problem still occures things 
have been narrowed down a bit.

Andrew

Re: Can't attach to child using gdb-5.1.1 on Linux

[Jesse Marlin]

Andrew Cagney writes:
 > > I have a process that forks a child.  If I try to attach to the
 > > child it reads the symbols and shows an accurate stack trace, but when
 > > I set a breakpoint and try to continue I get the following:
 > > 
 > > (gdb) c
 > > Continuing.
 > > lin-lwp.c:1003: gdb-internal-error: lin_lwp_wait: Assertion `WIFSTOPPED (status) && WSTOPSIG (status) == SIGSTOP' failed.
 > > An internal GDB error was detected.  This may make further
 > > debugging unreliable.  Continue this debugging session? (y or n) n
 > > 
 > > Create a core file containing the current state of GDB? (y or n) n
 > > 
 > > Inferior GDB exited abnormally with code 1
 > > 
 > > 
 > > I answered yes to the above also, but it did not improve the situation.  Has
 > > anybody seen this behvior before.  I am using gcc-2.95.3 for compilation.
 > 
 > (not sure if someone has responded to this)

No this is first.

 > 
 > The first thing to do is get a more recent GDB from:

I have been using gdb-5.1.1.  More recent as in the current CVS?
Thanks for the response.

 > 
 > 	http://sources.redhat.com/gdb/current/
 > 
 > many problems have been fixed and if this problem still occures things 
 > have been narrowed down a bit.
 > 
 > Andrew
 > 

Re: Can't attach to child using gdb-5.1.1 on Linux

[Andrew Cagney]

>> The first thing to do is get a more recent GDB from:
> 
> I have been using gdb-5.1.1.  More recent as in the current CVS?
> Thanks for the response.
CVS or a snap taken from the 5.2 branch or the mainline.  You should be 
able to find all of them via that page - if you can't it is a bug.

Andrew

Re: Can't attach to child using gdb-5.1.1 on Linux

[Jesse Marlin]

Andrew Cagney writes:
 > > I have a process that forks a child.  If I try to attach to the
 > > child it reads the symbols and shows an accurate stack trace, but when
 > > I set a breakpoint and try to continue I get the following:
 > > 
 > > (gdb) c
 > > Continuing.
 > > lin-lwp.c:1003: gdb-internal-error: lin_lwp_wait: Assertion `WIFSTOPPED (status) && WSTOPSIG (status) == SIGSTOP' failed.
 > > An internal GDB error was detected.  This may make further
 > > debugging unreliable.  Continue this debugging session? (y or n) n
 > > 
 > > Create a core file containing the current state of GDB? (y or n) n
 > > 
 > > Inferior GDB exited abnormally with code 1
 > > 
 > > 
 > > I answered yes to the above also, but it did not improve the situation.  Has
 > > anybody seen this behvior before.  I am using gcc-2.95.3 for compilation.
 > 
 > (not sure if someone has responded to this)
 > 
 > The first thing to do is get a more recent GDB from:
 > 
 > 	http://sources.redhat.com/gdb/current/
 > 
 > many problems have been fixed and if this problem still occures things 
 > have been narrowed down a bit.
 > 
 > Andrew
 > 

Okay I downloaded the 20020404 daily snapshot, compiled, installed, and
used it.  Here is what I am getting now.  The child is sleeping for 2
minutes.  Here is what I see when I initially attach to the child.

(gdb) where
#0  0x40a35621 in nanosleep () from /lib/libc.so.6
#1  0x40a355b8 in sleep () from /lib/libc.so.6
#2  0x0804d3fa in start_archive_client (dev_ptr=0x809e580)
    at dcs_archive_mgr.c:2258
#3  0x0804ab8c in ping_child (timer_id=0x809e410, context=0x0, now=1017936110)
    at dcs_archive_mgr.c:798
#4  0x4051de96 in evn_dispatch_event (event=0xbfffcd24, 
    current_time=0x40524af0) at event_manager.c:803
#5  0x4051dfef in evn_process_event (wait=1) at event_manager.c:879
#6  0x4051e04b in evn_process_all_events () at event_manager.c:932
#7  0x0804ae3c in proc () at dcs_archive_mgr.c:892
#8  0x0804a510 in main (argc=7, argv=0xbfffce64) at dcs_archive_mgr.c:606
#9  0x409a8c5f in __libc_start_main () from /lib/libc.so.6
(gdb) br child_init
Breakpoint 1 at 0x804de64: file dcs_archive_mgr.c, line 2810.
(gdb) break dcs_archive_mgr.c:2262
Breakpoint 2 at 0x804d3fd: file dcs_archive_mgr.c, line 2262.



After setting a few breakpoints I continue and see this:

(gdb) c
Continuing.
lin-lwp.c:1105: gdb-internal-error: lin_lwp_wait: Assertion `WIFSTOPPED (status) && WSTOPSIG (status) == SIGSTOP' failed.
An internal GDB error was detected.  This may make further
debugging unreliable.  Quit this debugging session? (y or n) n

Create a core file containing the current state of GDB? (y or n) y
(gdb) where
#0  0x40a35621 in nanosleep () from /lib/libc.so.6
#1  0x40a355b8 in sleep () from /lib/libc.so.6
#2  0x0804d3fa in start_archive_client (dev_ptr=0x809e580)
    at dcs_archive_mgr.c:2258
#3  0x0804ab8c in ping_child (timer_id=0x809e410, context=0x0, now=1017936110)
    at dcs_archive_mgr.c:798
#4  0x4051de96 in evn_dispatch_event (event=0xbfffcd24, 
    current_time=0x40524af0) at event_manager.c:803
#5  0x4051dfef in evn_process_event (wait=1) at event_manager.c:879
#6  0x4051e04b in evn_process_all_events () at event_manager.c:932
#7  0x0804ae3c in proc () at dcs_archive_mgr.c:892
#8  0x0804a510 in main (argc=7, argv=0xbfffce64) at dcs_archive_mgr.c:606
#9  0x409a8c5f in __libc_start_main () from /lib/libc.so.6


The stack traces look good before and after, but I can't debug anything.
I did create a core file of gdb this time, and will submit if you want me
to.

Re: Can't attach to child using gdb-5.1.1 on Linux

[Andrew Cagney]

> The stack traces look good before and after, but I can't debug anything.
> I did create a core file of gdb this time, and will submit if you want me
> to.

I don't know that the core fill will be of much use. However, a test 
case and the backtrace from the GDB core file would be.

enjoy,
Andrew

Re: Can't attach to child using gdb-5.1.1 on Linux

[Jesse Marlin]

Andrew Cagney writes:
 > > The stack traces look good before and after, but I can't debug anything.
 > > I did create a core file of gdb this time, and will submit if you want me
 > > to.
 > 
 > I don't know that the core fill will be of much use. However, a test 
 > case and the backtrace from the GDB core file would be.

In the general case this seems to work (ie. fork and try to attach to a
sleeping child).  My program seems to be doing something that causes gdb
to behave differently.  I am still trying to generate a test case for it.