[patch] early abort if child spawning failed

[patch] early abort if child spawning failed

[Mike Frysinger]

[-- Attachment #1.1: Type: text/plain, Size: 596 bytes --]

if you use gdbserver to listen on a port and the specified program fails to 
launch properly, the server process will continue on to the accept() call and 
sit there forever ... you cant control+c the process as the signals are taken 
over which means you have to connect with gdb to force the server to re-check 
itself and then abort ... all in all, a pita for a simple typo.

the attached patch checks the status field immediately after launch rather 
than falling down into the main loop and waiting for the remote debugger to 
connect for the status to get properly checked.
-mike

[-- Attachment #1.2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 827 bytes --]

[-- Attachment #2: gdbserver-check-initial-startup.patch --]
[-- Type: text/x-diff, Size: 997 bytes --]

2007-09-21  Mike Frysinger  <vapier@gentoo.org>

	* server.c (main): Check status after start_inferior.

Index: server.c
===================================================================
RCS file: /cvs/src/src/gdb/gdbserver/server.c,v
retrieving revision 1.56
diff -u -p -r1.56 server.c
--- server.c	23 Aug 2007 18:08:48 -0000	1.56
+++ server.c	22 Sep 2007 02:30:10 -0000
@@ -900,9 +900,14 @@ main (int argc, char *argv[])
       /* Wait till we are at first instruction in program.  */
       signal = start_inferior (&argv[2], &status);
 
+      if (status == 'W' || status == 'X')
+	{
+	  fprintf (stderr, "Aborting server; child exited with %i\n", signal);
+	  exit (signal);
+	}
+
       /* We are now (hopefully) stopped at the first instruction of
-	 the target process.  This assumes that the target process was
-	 successfully created.  */
+	 the target process.  */
 
       /* Don't report shared library events on the initial connection,
 	 even if some libraries are preloaded.  */

Re: [patch] early abort if child spawning failed

[Daniel Jacobowitz]

On Fri, Sep 21, 2007 at 10:31:11PM -0400, Mike Frysinger wrote:
> if you use gdbserver to listen on a port and the specified program fails to 
> launch properly, the server process will continue on to the accept() call and 
> sit there forever ... you cant control+c the process as the signals are taken 
> over which means you have to connect with gdb to force the server to re-check 
> itself and then abort ... all in all, a pita for a simple typo.
> 
> the attached patch checks the status field immediately after launch rather 
> than falling down into the main loop and waiting for the remote debugger to 
> connect for the status to get properly checked.
> -mike

Did you port this from an earlier version?  It shouldn't be necessary
any more - Michael fixed this 2007-07-27.

-- 
Daniel Jacobowitz
CodeSourcery

Re: [patch] early abort if child spawning failed

[Mike Frysinger]

On 9/21/07, Daniel Jacobowitz <drow@false.org> wrote:
> On Fri, Sep 21, 2007 at 10:31:11PM -0400, Mike Frysinger wrote:
> > if you use gdbserver to listen on a port and the specified program fails to
> > launch properly, the server process will continue on to the accept() call and
> > sit there forever ... you cant control+c the process as the signals are taken
> > over which means you have to connect with gdb to force the server to re-check
> > itself and then abort ... all in all, a pita for a simple typo.
> >
> > the attached patch checks the status field immediately after launch rather
> > than falling down into the main loop and waiting for the remote debugger to
> > connect for the status to get properly checked.
>
> Did you port this from an earlier version?

yep

> It shouldn't be necessary any more - Michael fixed this 2007-07-27.

when i looked at the CVS HEAD version, the comment "we assume success"
led me to believe that the situation was unchanged ... but looking at
the diff, the patch i proposed is pretty much the same as the one
Michael proposed so yeah, what i posted can just be dropped ...

thanks
-mike