Re: [patch 2/2] Fix watchpoints for multi-inferior #2

[Pedro Alves <palves@redhat.com>]

Here's what I'm applying.  I've added comments inspired on your follow-up patch.

This makes watchpoint-multi.exp pass cleanly with the extended-remote
board, on x86_64 Fedora 16.

gdb/gdbserver/
2012-03-16  Pedro Alves  <palves@redhat.com>
	    Jan Kratochvil  <jan.kratochvil@redhat.com>

	* server.c (cont_thread, general_thread): Add describing comments.
	(start_inferior): Clear `cont_thread'.
	(handle_v_cont): Don't set `cont_thread' if resuming all threads
	of a process.
---

 gdb/gdbserver/server.c |   24 ++++++++++++++++++++++--
 1 files changed, 22 insertions(+), 2 deletions(-)

diff --git a/gdb/gdbserver/server.c b/gdb/gdbserver/server.c
index 3c97dbd..a4e9e57 100644
--- a/gdb/gdbserver/server.c
+++ b/gdb/gdbserver/server.c
@@ -30,7 +30,19 @@
 #include <sys/wait.h>
 #endif

+/* The thread set with an `Hc' packet.  `Hc' is deprecated in favor of
+   `vCont'.  Note the multi-process extensions made `vCont' a
+   requirement, so `Hc pPID.TID' is pretty much undefined.  So
+   CONT_THREAD can be null_ptid for no `Hc' thread, minus_one_ptid for
+   resuming all threads of the process (again, `Hc' isn't used for
+   multi-process), or a specific thread ptid_t.
+
+   We also set this when handling a single-thread `vCont' resume, as
+   some places in the backends check it to know when (and for which
+   thread) single-thread scheduler-locking is in effect.  */
 ptid_t cont_thread;
+
+/* The thread set with an `Hg' packet.  */
 ptid_t general_thread;

 int server_waiting;
@@ -262,6 +274,10 @@ start_inferior (char **argv)
   signal (SIGTTIN, SIG_DFL);
 #endif

+  /* Clear this so the backend doesn't get confused, thinking
+     CONT_THREAD died, and it needs to resume all threads.  */
+  cont_thread = null_ptid;
+
   signal_pid = create_inferior (new_argv[0], new_argv);

   /* FIXME: we don't actually know at this point that the create
@@ -1962,9 +1978,13 @@ handle_v_cont (char *own_buf)
   if (i < n)
     resume_info[i] = default_action;

-  /* Still used in occasional places in the backend.  */
+  /* `cont_thread' is still used in occasional places in the backend,
+     to implement single-thread scheduler-locking.  Doesn't make sense
+     to set it if we see a stop request, or any form of wildcard
+     vCont.  */
   if (n == 1
-      && !ptid_equal (resume_info[0].thread, minus_one_ptid)
+      && !(ptid_equal (resume_info[0].thread, minus_one_ptid)
+	   || ptid_get_lwp (resume_info[0].thread) == -1)
       && resume_info[0].kind != resume_stop)
     cont_thread = resume_info[0].thread;
   else