[testsuite] kfail sigaltstack.exp on i386/gnu/linux

[testsuite] kfail sigaltstack.exp on i386/gnu/linux

[Daniel Jacobowitz]

See PR 1736 and the link to gdb@ for more information.  I've kfailed
this bug because, while the bug is external, GDB could handle it more
gracefully...

Tested on i686-pc-linux-gnu, where six of the tests FAIL->KFAIL.  OK?

-- 
Daniel Jacobowitz

2004-08-08  Daniel Jacobowitz  <dan@debian.org>

	PR gdb/1736
	* gdb.base/sigaltstack.exp (finish_test): New procedure.  KFAIL
	for i?86-*-linux*.

Index: testsuite/gdb.base/sigaltstack.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.base/sigaltstack.exp,v
retrieving revision 1.1
diff -u -p -r1.1 sigaltstack.exp
--- testsuite/gdb.base/sigaltstack.exp	5 Apr 2004 21:00:21 -0000	1.1
+++ testsuite/gdb.base/sigaltstack.exp	8 Aug 2004 22:00:02 -0000
@@ -78,11 +78,26 @@ gdb_expect_list "backtrace" ".*$gdb_prom
     "\[\r\n\]+.7 \[^\r\n\]* main .*"
 }
 
+proc finish_test { pattern msg } {
+    gdb_test_multiple "finish" $msg {
+	-re "Cannot insert breakpoint 0" {
+	    # Some platforms use a special read-only page for signal
+	    # trampolines.  We can't set a breakpoint there, and we
+	    # don't gracefully fall back to single-stepping.
+	    setup_kfail "i?86-*-linux*" gdb/1736
+	    fail "$msg (could not set breakpoint)"
+	}
+	-re "$pattern" {
+	    pass $msg
+	}
+    }
+}
+
 # Finish?
-gdb_test "finish" "signal handler called." "finish from catch LEAF"
-gdb_test "finish" "thrower .next_level=INNER, .*" "finish to throw INNER"
-gdb_test "finish" "catcher .*" "finish to catch INNER"
-gdb_test "finish" "signal handler called.*" "finish from catch INNER"
-gdb_test "finish" "thrower .next_level=OUTER, .*" "finish to OUTER"
-gdb_test "finish" "catcher .*" "finish to catch MAIN"
-gdb_test "finish" "main .*" "finish to MAIN"
+finish_test "signal handler called." "finish from catch LEAF"
+finish_test "thrower .next_level=INNER, .*" "finish to throw INNER"
+finish_test "catcher .*" "finish to catch INNER"
+finish_test "signal handler called.*" "finish from catch INNER"
+finish_test "thrower .next_level=OUTER, .*" "finish to OUTER"
+finish_test "catcher .*" "finish to catch MAIN"
+finish_test "main .*" "finish to MAIN"

Re: [testsuite] kfail sigaltstack.exp on i386/gnu/linux

[Michael Chastain]

This patch is approved.

It's okay to use a KFAIL even for an external bug.  The gdb PR can be
put in class "external" and refer to some external bug report against a
different product, and the gdb PR can still serve as a locus of
information about the bug.

===

2004-08-08  Daniel Jacobowitz  <dan@debian.org>

	PR gdb/1736
 	* gdb.base/sigaltstack.exp (finish_test): New procedure.  KFAIL
	for i?86-*-linux*.

Re: [testsuite] kfail sigaltstack.exp on i386/gnu/linux

[Daniel Jacobowitz]

On Sun, Aug 08, 2004 at 06:22:12PM -0400, Michael Chastain wrote:
> This patch is approved.
> 
> It's okay to use a KFAIL even for an external bug.  The gdb PR can be
> put in class "external" and refer to some external bug report against a
> different product, and the gdb PR can still serve as a locus of
> information about the bug.

I thought we had agreed that was the definition of XFAIL for GDB, an
expected external failure?

> 2004-08-08  Daniel Jacobowitz  <dan@debian.org>
> 
> 	PR gdb/1736
>  	* gdb.base/sigaltstack.exp (finish_test): New procedure.  KFAIL
> 	for i?86-*-linux*.

Checked in.

-- 
Daniel Jacobowitz

Re: [testsuite] kfail sigaltstack.exp on i386/gnu/linux

[Michael Chastain]

Daniel Jacobowitz <drow@false.org> wrote:
> I thought we had agreed that was the definition of XFAIL for GDB, an
> expected external failure?

That is the definition of XFAIL.  I'm saying that I accept a KFAIL
instead, if you have a reason.

Michael C

Re: [testsuite] kfail sigaltstack.exp on i386/gnu/linux

[Mark Kettenis]

   Date: Sun, 8 Aug 2004 18:02:20 -0400
   From: Daniel Jacobowitz <drow@false.org>

   2004-08-08  Daniel Jacobowitz  <dan@debian.org>

	   PR gdb/1736
	   * gdb.base/sigaltstack.exp (finish_test): New procedure.  KFAIL
	   for i?86-*-linux*.

Sorry Daniel, this patch isn't right.  The tests get out of sync on my
i386-unknown-freebsd4.7 system.  You can't blindly convert from
gdb_test to gdb_test_multiple.  The latter doesn't provide proper
anchoring.  The attached patch fixes that.  Is this OK Michael?

Index: testsuite/ChangeLog
from  Mark Kettenis  <kettenis@gnu.org>

	* gdb.base/sigaltstack.exp: Provide proper anchoring.

Index: testsuite/gdb.base/sigaltstack.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.base/sigaltstack.exp,v
retrieving revision 1.3
diff -u -p -r1.3 sigaltstack.exp
--- testsuite/gdb.base/sigaltstack.exp 9 Aug 2004 13:16:16 -0000 1.3
+++ testsuite/gdb.base/sigaltstack.exp 9 Aug 2004 17:07:51 -0000
@@ -89,7 +89,7 @@ proc finish_test { pattern msg } {
 	    setup_kfail "i?86-*-linux*" gdb/1736
 	    fail "$msg (could not set breakpoint)"
 	}
-	-re "$pattern" {
+	-re "$pattern.*${gdb_prompt} $" {
 	    pass $msg
 	}
     }

Re: [testsuite] kfail sigaltstack.exp on i386/gnu/linux

[Daniel Jacobowitz]

On Mon, Aug 09, 2004 at 07:08:54PM +0200, Mark Kettenis wrote:
>    Date: Sun, 8 Aug 2004 18:02:20 -0400
>    From: Daniel Jacobowitz <drow@false.org>
> 
>    2004-08-08  Daniel Jacobowitz  <dan@debian.org>
> 
> 	   PR gdb/1736
> 	   * gdb.base/sigaltstack.exp (finish_test): New procedure.  KFAIL
> 	   for i?86-*-linux*.
> 
> Sorry Daniel, this patch isn't right.  The tests get out of sync on my
> i386-unknown-freebsd4.7 system.  You can't blindly convert from
> gdb_test to gdb_test_multiple.  The latter doesn't provide proper
> anchoring.  The attached patch fixes that.  Is this OK Michael?

Glarg - I apologize profusely.  You are correct.  I remembered to
anchor the new pattern and forgot to anchor the old one.

-- 
Daniel Jacobowitz

Re: [testsuite] kfail sigaltstack.exp on i386/gnu/linux

[Michael Chastain]

This patch is approved.

Like Daniel said, glarg.  I should be proof-reading for this
in every patch.  That's part of my role.  :-(

Fixes for missing "$gdb_prompt" anchors are obvious, if you hit
any more of them.  (Just say where you tested them, even for
an obvious patch).

Michael C

===

Index: testsuite/ChangeLog
from  Mark Kettenis  <kettenis@gnu.org>

	* gdb.base/sigaltstack.exp: Provide proper anchoring.

Re: [testsuite] kfail sigaltstack.exp on i386/gnu/linux

[Mark Kettenis]

   Date: Sun, 8 Aug 2004 18:02:20 -0400
   From: Daniel Jacobowitz <drow@false.org>

   2004-08-08  Daniel Jacobowitz  <dan@debian.org>

	   PR gdb/1736
	   * gdb.base/sigaltstack.exp (finish_test): New procedure.  KFAIL
	   for i?86-*-linux*.

Sorry Daniel, this patch isn't right.  The tests get out of sync on my
i386-unknown-freebsd4.7 system.  You can't blindly convert from
gdb_test to gdb_test_multiple.  The latter doesn't provide proper
anchoring.  The attached patch fixes that.  Is this OK Michael?

Index: testsuite/ChangeLog
from  Mark Kettenis  <kettenis@gnu.org>

	* gdb.base/sigaltstack.exp: Provide proper anchoring.

Index: testsuite/gdb.base/sigaltstack.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.base/sigaltstack.exp,v
retrieving revision 1.3
diff -u -p -r1.3 sigaltstack.exp
--- testsuite/gdb.base/sigaltstack.exp 9 Aug 2004 13:16:16 -0000 1.3
+++ testsuite/gdb.base/sigaltstack.exp 9 Aug 2004 17:07:51 -0000
@@ -89,7 +89,7 @@ proc finish_test { pattern msg } {
 	    setup_kfail "i?86-*-linux*" gdb/1736
 	    fail "$msg (could not set breakpoint)"
 	}
-	-re "$pattern" {
+	-re "$pattern.*${gdb_prompt} $" {
 	    pass $msg
 	}
     }