[review v2] infrun: handle already-exited threads when attempting to stop

["Pedro Alves (Code Review)" <gerrit at gnutoolchain-gerrit dot osci dot io>]

Pedro Alves has posted comments on this change.

Change URL: https://gnutoolchain-gerrit.osci.io/r/c/binutils-gdb/+/133
......................................................................


Patch Set 2:

(1 comment)

| --- gdb/infrun.c
| +++ gdb/infrun.c
| @@ -4494,7 +4509,19 @@ stop_all_threads (void)
| +		  if (ws.kind == TARGET_WAITKIND_THREAD_EXITED)
| +		    delete_thread (t);
| +		  else
| +		    {
| +		      /* TARGET_WAITKIND_EXITED or
| +			 TARGET_WAITKIND_SIGNALLED.  */
| +		      /* Need to restore the context because
| +			 handle_inferior_exit switches it.  */
| +		      scoped_restore_current_pspace_and_thread restore;
| +		      handle_inferior_exit (event_ptid, ws);

PS2, Line 4518:

Sorry, but this doesn't look right.

We're inside stop_all_threads, processing some other event, and a
process exit event for one of the processes we're trying to stop comes
along, and this processes it right away.  Once the stop_all_threads
dance is done, we go back to handling the original event, and possibly
reporting a stop to the user.  Meanwhile, whatever state that was set
by handle_inferior_exit, like e.g. $_exitcode, is lost, or now
incorrect for the reported stop.  We also never present a stop on the
CLI for that "spurious" process exit.  Here:

 (gdb)
 continue
 Continuing.
 Executing on build: kill -9 29412 29417    (timeout = 300)
 spawn -ignore SIGHUP kill -9 29412 29417

 Program terminated with signal SIGKILL, Killed.
 The program no longer exists.
     <<<<<<<<<<< no user-visible stop / prompt here. <<<<<<<<<<<<<<<
 Program terminated with signal SIGKILL, Killed.
 The program no longer exists.
 (gdb) PASS: gdb.multi/multi-kill.exp: iteration 1: back to gdb prompt

The fix for this I think must be around leaving the
TARGET_WAITKIND_EXITED/TARGET_WAITKIND_SIGNALLED
event pending, so that it is processed later when we're out of the
stop_all_threads loop and back to
dequeuing the next event.

gdb/linux-nat.c also has its own "stop all threads temporarily" logic,
and that does that -- leaves process exits pending.  See wait_lwp:

              /* If this is the leader exiting, it means the whole
                 process is gone.  Store the status to report to the
                 core.  Store it in lp->waitstatus, because lp->status
                 would be ambiguous (W_EXITCODE(0,0) == 0).  */
              store_waitstatus (&lp->waitstatus, status);
              return 0;

| +		    }
| +		}
|  	    }
|  	  else
|  	    {
|  	      thread_info *t = find_thread_ptid (event_ptid);
|  	      if (t == NULL)
|  		t = add_thread (event_ptid);
|  

-- 
Gerrit-Project: binutils-gdb
Gerrit-Branch: master
Gerrit-Change-Id: I7cec98f40283773b79255d998511da434e9cd408
Gerrit-Change-Number: 133
Gerrit-PatchSet: 2
Gerrit-Owner: Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
Gerrit-Reviewer: Luis Machado <luis.machado@linaro.org>
Gerrit-Reviewer: Pedro Alves <palves@redhat.com>
Gerrit-Reviewer: Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
Gerrit-Comment-Date: Fri, 06 Dec 2019 16:19:22 +0000
Gerrit-HasComments: Yes
Gerrit-Has-Labels: No
Gerrit-MessageType: comment