Re: Avoid timeouts in call-sc.exp

[Paul Gilliam <pgilliam@us.ibm.com>]

Here is another try.  
-=# Paul #=-

PS: I only did the ${....} because that seemed to be the usage in call-sc.exp

2004-08-25  Paul Gilliam  <pgilliam@us.ibm.com>

        * gdb.base/call-sc: Make sure PC is syncronized after the "return".


diff -Naur old/call-sc.exp new/call-sc.exp
--- old/call-sc.exp     2004-08-24 01:11:59.000000000 +0000
+++ new/call-sc.exp     2004-08-25 22:27:04.089926216 +0000
@@ -281,6 +281,30 @@
        }
     }

+    # If the previous test did not work, the program counter might
+    # still be inside foo() rather than main().  Make sure the program
+    # counter is is main().
+    #
+    # This happens on ppc64 GNU/Linux with gcc 3.4.1 and a buggy GDB
+
+    set test "return foo; syncronize pc to main()"
+    for {set loop_count 0} {$loop_count < 2} {incr loop_count} {
+      gdb_test_multiple "backtrace 1" $test {
+        -re "#0.*main \\(\\).*${gdb_prompt} $" {
+          pass $test
+          set loop_count 2
+        }
+        -re "#0.*fun \\(\\).*${gdb_prompt} $" {
+          if {$loop_count < 1} {
+            gdb_test "finish" ".*" ""
+          } else {
+            fail $test
+            set loop_count 2
+          }
+        }
+      }
+    }
+
     # Check that the return-value is as expected.  At this stage we're
     # just checking that GDB has returned a value consistent with
     # "return_value_unknown" set above.



On Wednesday 25 August 2004 11:11, Michael Chastain wrote:
> It's a little weird, but it's much better.
>
> . "synchronize" is spelled with an h.
> . include a changelog entry with the submission
> . you don't need ${...} for each variable access.
>   just $loop_count, $test, and so on.
> . this is the fsf version of gdb, it's better to work with
>   fsf sources -- either gdb 6.2 or the cvs version of gdb,
>   not a red hat gdb.
>
> Can you fix those things up, then download gdb 6.2 or the
> cvs version of gdb, and do your work against that.