[RFA] gdb.base/shlib-call.exp changes for IRIX

[RFA] gdb.base/shlib-call.exp changes for IRIX

[Kevin Buettner]

Some more IRIX related changes.  I can break this change up into two
parts if necessary.  (The linker flag change is probably reasonably
obvious, but the change to the "step out of shr2" test should be
reviewed.)

Anyway, regarding the "step out of shr2" test changes, the IRIX
compiler doesn't emit line number information for the epilogue, so we
don't see the step onto the final ``}'' of the function.  I don't
see a problem with this behavior.  I've revised the test to accept
either behavior.

I've also revised the messages a little bit to account for the
restructuring of the test.

Okay to commit?

	* gdb.base/shlib-call.exp (additional_flags): Add linker flags
	so that shared libraries created by test will be found.
	(step inside shr2): Don't fail if first step ends up stepping
	out of the function instead of stopping on the epilogue.

Index: testsuite/gdb.base/shlib-call.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.base/shlib-call.exp,v
retrieving revision 1.5
diff -u -p -r1.5 shlib-call.exp
--- testsuite/gdb.base/shlib-call.exp	18 Apr 2002 21:07:20 -0000	1.5
+++ testsuite/gdb.base/shlib-call.exp	29 Jul 2002 21:37:35 -0000
@@ -103,6 +103,8 @@ if { ($gcc_compiled 
       &&  ([istarget "powerpc*-*-aix*"]
 	|| [istarget "rs6000*-*-aix*"] )) } {
     set additional_flags "additional_flags=-L${objdir}/${subdir}"
+} elseif { [istarget "mips-sgi-irix*"] } {
+    set additional_flags "additional_flags=-rpath ${objdir}/${subdir}"
 } else {
     set additional_flags ""
 }
@@ -234,18 +236,27 @@ gdb_expect {
 
 #step -return
     send_gdb "step\n"
+    # A step at this point will either take us entirely out of
+    # the function or into the function's epilogue.  The exact
+    # behavior will differ depending upon upon whether or not
+    # the compiler emits line number information for the epilogue.
     gdb_expect {
-        -re ".*\\\}.*$gdb_prompt $" { pass "step inside shr2 (shlib func)"}
-        -re ".*$gdb_prompt $" { fail "step inside shr2 (shlib func)" }
-        timeout { fail "step inside shr2 (shlib func) (timeout)" }
-    }
-    
-    send_gdb "step\n"
-    gdb_expect {
-        -re "main \\(\\) at.*g = mainshr1\\(g\\);.*$gdb_prompt $" { pass "step out of shr2 to main"}
-        -re ".*$gdb_prompt $" { fail "step out of shr2 to main" }
+        -re "main \\(\\) at.*g = mainshr1\\(g\\);.*$gdb_prompt $" { 
+	    pass "step out of shr2 to main"
+	}
+        -re ".*\\\}.*$gdb_prompt $" {
+	    pass "step out of shr2 to main (stopped in shr2 epilogue)"
+	    send_gdb "step\n"
+	    gdb_expect {
+		-re "main \\(\\) at.*g = mainshr1\\(g\\);.*$gdb_prompt $" { pass "step out of shr2 epilogue to main"}
+		-re ".*$gdb_prompt $" { fail "step out of shr2 epilogue to main" }
+		timeout { fail "step out of shr2 epilogue to main (timeout)" }
+	    }
+	}
+        -re ".*$gdb_prompt $" { fail "step out of shr2" }
         timeout { fail "step out of shr2 to main (timeout)" }
     }
+    
 
 #print mainshr1(1)
 send_gdb "print mainshr1(1)\n"

Re: [RFA] gdb.base/shlib-call.exp changes for IRIX

[Kevin Buettner]

On Jul 29,  3:15pm, Kevin Buettner wrote:

> Some more IRIX related changes.  I can break this change up into two
> parts if necessary.  (The linker flag change is probably reasonably
> obvious, but the change to the "step out of shr2" test should be
> reviewed.)
> 
> Anyway, regarding the "step out of shr2" test changes, the IRIX
> compiler doesn't emit line number information for the epilogue, so we
> don't see the step onto the final ``}'' of the function.  I don't
> see a problem with this behavior.  I've revised the test to accept
> either behavior.
> 
> I've also revised the messages a little bit to account for the
> restructuring of the test.
> 
> Okay to commit?
> 
> 	* gdb.base/shlib-call.exp (additional_flags): Add linker flags
> 	so that shared libraries created by test will be found.
> 	(step inside shr2): Don't fail if first step ends up stepping
> 	out of the function instead of stopping on the epilogue.

I'm also (temporarily) withdrawing this patch since I now have some
other changes to make to shlib-call.exp.  I'll commit the obvious
portions and then repost the part that I want reviewed.

Kevin