Re: [PATCH] Don't use runto_main in i386-unwind.exp

[Mark Kettenis <kettenis@chello.nl>]

   Date: Sun, 23 Nov 2003 15:57:12 -0500
   From: Daniel Jacobowitz <drow@mvista.com>

   On Sun, Nov 23, 2003 at 09:50:27PM +0100, Mark Kettenis wrote:
   >    Please don't do this.  runto_main works (suboptimally, but it works)
   >    for remote targets which use a "target foo; load; continue" sequence,
   >    but "run" doesn't.
   > 
   >    How about gdb_run_cmd followed by a gdb_expect looking for the SIGTRAP?
   > 
   > Like the attached patch?  Works for me.

   Thanks - that's just what I had in mind.  I'll try to go back to making
   the testsuite work with gdbserver so that others can test this (I have
   some comments from Elena that I still need to address).

I never got it working properly :-(.  Anyway, I checked in the attached.

Index: testsuite/ChangeLog
from  Mark Kettenis  <kettenis@gnu.org>
 
	* gdb.arch/i386-unwind.exp: Use gdb_run_cmd and gdb_expect instead
	of gdb_test "run".

Index: testsuite/gdb.arch/i386-unwind.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.arch/i386-unwind.exp,v
retrieving revision 1.2
diff -u -p -r1.2 i386-unwind.exp
--- testsuite/gdb.arch/i386-unwind.exp 23 Nov 2003 19:30:52 -0000 1.2
+++ testsuite/gdb.arch/i386-unwind.exp 23 Nov 2003 21:12:51 -0000
@@ -47,8 +47,21 @@ gdb_load ${binfile}
 
 # Testcase for backtrace/1435.
 
-gdb_test "run" "Program received signal SIGTRAP.*" \
-         "run past gdb1435"
+# We use gdb_run_cmd so this stands a chance to work for remote
+# targets too.
+gdb_run_cmd
+
+gdb_expect {
+    -re "Program received signal SIGTRAP.*$gdb_prompt $" {
+	pass "run past gdb1435"
+    }
+    -re ".*$gdb_prompt $" {
+	fail "run past gdb1435"
+    }
+    timeout {
+	fail "run past gdb1435 (timeout)"
+    }
+}
 
 gdb_test "backtrace 10" \
 	"#1\[ \t]*$hex in gdb1435.*\r\n#2\[ \t\]*$hex in main.*" \