diff mbox

Handle hitting a recording extent boundary as a normal trap-caused stop

Message ID 20190104213758.3477-1-autkin@undo.io
State New, archived
Headers

Commit Message

Andrey Utkin Jan. 4, 2019, 9:37 p.m. UTC 
  Because that's what it is.

This enables reporting of a breakpoint hitting on the first replayable
instruction when reverse-stepping to the beginning of recorded history.
---
 gdb/infrun.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

Comments

Andrey Utkin Jan. 7, 2019, 4:07 p.m. UTC | #1 
Raised a related bugticket
https://sourceware.org/bugzilla/show_bug.cgi?id=24073 (Breakpoint on the
first insn of recorded history is not reported in replay).
Tom Tromey Jan. 17, 2019, 3:34 p.m. UTC | #2 
>>>>> "Andrey" == Andrey Utkin <autkin@undo.io> writes:

Andrey> Because that's what it is.
Andrey> This enables reporting of a breakpoint hitting on the first replayable
Andrey> instruction when reverse-stepping to the beginning of recorded history.

Hi.  Thanks for the patch.

See the contribution checklist for some info on how to submit a patch:

    https://sourceware.org/gdb/wiki/ContributionChecklist

This one is missing a ChangeLog and perhaps a test case, assuming the
latter is reasonably possible to write.

Tom
diff mbox

Patch

diff --git a/gdb/infrun.c b/gdb/infrun.c
index 150288264f..a72f06129b 100644
--- a/gdb/infrun.c
+++ b/gdb/infrun.c
@@ -5212,11 +5212,12 @@  Cannot fill $_exitsignal with the correct signal number.\n"));
       ecs->event_thread->suspend.stop_pc
 	= regcache_read_pc (get_thread_regcache (inferior_thread ()));
 
-      if (handle_stop_requested (ecs))
-	return;
-
       gdb::observers::no_history.notify ();
-      stop_waiting (ecs);
+
+      // The end of history is always a stop after a stepping/running op
+      ecs->ws.kind = TARGET_WAITKIND_STOPPED;
+      ecs->ws.value.sig = GDB_SIGNAL_TRAP;
+      handle_signal_stop (ecs);
       return;
     }
 }