SIGSEGV in Insight cleanup

SIGSEGV in Insight cleanup

[Nicholas_Karagas]

I've noticed, while debugging Insight with gdb, that Insight always (to the
best of my knowledge, at least with remote targets) exits with a SIGSEGV.
Specifically,

Program received signal SIGSEGV, Segmentation fault.
0x47354e in serial_write (scb=0x0, str=0x291d1f0 "$k#6b", len=5)
     at serial.c:387
387       return (scb->ops->write (scb, str, len));

This is in remote_kill, when gdb sends the kill packet to the remote
target.  I've traced the reason to the following series of function calls
(the first quit_force originates from exiting Insight and answering yes to
the "A debugging session is active.  Do you still want to close the
debugger?" dialog box):

quit_force -> do_final_cleanups -> do_my_cleanups -> gdbtk_cleanup ->
Tcl_Finalize -> (many more) -> quit_force

By the time the second (recursive) call to quit_force calls remote_kill,
the scb has already been cleared.  The SIGSEGV only occurs at the end of
the program, and is thus usually invisible, but it might be causing memory
leaks in NT, etc.  I don't know that much about the internals of either
Insight or gdb... is this a known problem?  Any solutions?  Would it help
to see the complete (53 function) backtrace?

Nicholas Karagas
Associate Software Engineer
Cirilium Corporation
Tempe, AZ
(480) 317-1144

Re: SIGSEGV in Insight cleanup

[Fernando Nasser]

I've never seem this.  I believe you are implying this happens on a cygwin NT host,
isn't it?

Is it in the latest cygwin and gdb snapshots?

I will keep this in mind while I use the debugger and see if it happens to me
so I can take a look.

Thanks for reporting it.

Fernando



Nicholas_Karagas@cirilium.com wrote:
> 
> I've noticed, while debugging Insight with gdb, that Insight always (to the
> best of my knowledge, at least with remote targets) exits with a SIGSEGV.
> Specifically,
> 
> Program received signal SIGSEGV, Segmentation fault.
> 0x47354e in serial_write (scb=0x0, str=0x291d1f0 "$k#6b", len=5)
>      at serial.c:387
> 387       return (scb->ops->write (scb, str, len));
> 
> This is in remote_kill, when gdb sends the kill packet to the remote
> target.  I've traced the reason to the following series of function calls
> (the first quit_force originates from exiting Insight and answering yes to
> the "A debugging session is active.  Do you still want to close the
> debugger?" dialog box):
> 
> quit_force -> do_final_cleanups -> do_my_cleanups -> gdbtk_cleanup ->
> Tcl_Finalize -> (many more) -> quit_force
> 
> By the time the second (recursive) call to quit_force calls remote_kill,
> the scb has already been cleared.  The SIGSEGV only occurs at the end of
> the program, and is thus usually invisible, but it might be causing memory
> leaks in NT, etc.  I don't know that much about the internals of either
> Insight or gdb... is this a known problem?  Any solutions?  Would it help
> to see the complete (53 function) backtrace?
> 

-- 
Fernando Nasser
Red Hat - Toronto                       E-Mail:  fnasser@cygnus.com
2323 Yonge Street, Suite #300           Tel:  416-482-2661 ext. 311
Toronto, Ontario   M4P 2C9              Fax:  416-482-6299