Re: RFA: don't xfail ptype test

Re: RFA: don't xfail ptype test

[Michael Elizabeth Chastain]

Whoops, you've got the setup_kfail one line early in the tests.
Other than that, this patch is approved.

setup_kfail is our new facility for indicating that a test will
fail due to a known bug in gdb (k == "k"nown).  This way, we can
distinguish between known bugs and new, unknown bugs (which are
presumably regressions).

It's not perfect because the more setup_kfail's exist in the test corpus,
the more likely that a regression will manifest as a KFAIL.  But it's
a lot better than those hundreds of setup_xfail's that are not really
external bugs.

FernandoN has written a DejaGnu patch with doco.  Here's a pointer:

  http://sources.redhat.com/ml/gdb-patches/2002-04/msg00201.html

We're waiting on Rob Savoye right now to release a new DejaGnu
with this patch.  Also I don't know if Fernando's patch is in
sourceware dejagnu or not.  I'm using DejaGnu 1.42 + this patch,
so I am ready for kfail's whenever we start going live with them.

Michael C

Re: RFA: don't xfail ptype test

[Jim Blandy]

Michael Elizabeth Chastain <mec@shout.net> writes:
> Whoops, you've got the setup_kfail one line early in the tests.
> Other than that, this patch is approved.

Whoops!  Thanks for reading carefully.

> We're waiting on Rob Savoye right now to release a new DejaGnu
> with this patch.  Also I don't know if Fernando's patch is in
> sourceware dejagnu or not.  I'm using DejaGnu 1.42 + this patch,
> so I am ready for kfail's whenever we start going live with them.

So you're going to uncomment the calls to setup_kfail once there's a
released DejaGnu that supports them?

Re: RFA: don't xfail ptype test

[Michael Elizabeth Chastain]

Jim Blandy asks:
> So you're going to uncomment the calls to setup_kfail once there's a
> released DejaGnu that supports them?

That's more or less my intention; it's really up to Fernando.

The user base for running gdb tests is small, but it has several dejagnu's
to choose from.  There's an FSF dejagnu and a sourceware dejagnu, and
Cygwin has a very old dejagnu.

Michael C

Re: RFA: don't xfail ptype test

[Jim Blandy]

Michael Elizabeth Chastain <mec@shout.net> writes:
> This patch is returned for more work.
> 
> Hmmm.  I agree that this is a bug:
> 
>   (gdb) ptype &*"foo"
>   type = char [4]
> 
> I get that behavior in all of my configurations.
> 
> The problem is simply converting XFAIL -> FAIL.  That's correct, but it
> doesn't leave a good enough trail for people doing regression analysis.
> 
> Could you please file a bug report and then add a comment to the
> test script with the bug ID number in it:
> 
>   # setup_kfail "gdb/1234"
> 
> The bug report can say that this bug is also present in 5.2.
> 
> I am willing to file the bug report this evening if you want me to
> do that part.
> 
> Michael C

Okay, how's the below?  What's setup_kfail?  Where can I find a
description of it?

2002-05-10  Jim Blandy  <jimb@redhat.com>

	* gdb.base/printcmds.exp: Don't xfail the ptype command.  This is
	a bug.

Index: gdb/testsuite/gdb.base/printcmds.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.base/printcmds.exp,v
retrieving revision 1.7
diff -c -r1.7 printcmds.exp
*** gdb/testsuite/gdb.base/printcmds.exp	4 May 2002 15:18:21 -0000	1.7
--- gdb/testsuite/gdb.base/printcmds.exp	10 May 2002 18:05:04 -0000
***************
*** 626,633 ****
      gdb_test "ptype \"foo\""		" = char \\\[4\\\]"
      gdb_test "p *\"foo\""		" = 102 'f'"
      gdb_test "ptype *\"foo\""		" = char"
      gdb_test "p &*\"foo\""		" = \"foo\""
-     setup_xfail "*-*-*"
      gdb_test "ptype &*\"foo\""	"type = char \\*"
      gdb_test "p (char *)\"foo\""	" = \"foo\""
  }
--- 626,633 ----
      gdb_test "ptype \"foo\""		" = char \\\[4\\\]"
      gdb_test "p *\"foo\""		" = 102 'f'"
      gdb_test "ptype *\"foo\""		" = char"
+     # setup_kfail "gdb/538"
      gdb_test "p &*\"foo\""		" = \"foo\""
      gdb_test "ptype &*\"foo\""	"type = char \\*"
      gdb_test "p (char *)\"foo\""	" = \"foo\""
  }

Re: RFA: don't xfail ptype test

[Michael Elizabeth Chastain]

This patch is returned for more work.

Hmmm.  I agree that this is a bug:

  (gdb) ptype &*"foo"
  type = char [4]

I get that behavior in all of my configurations.

The problem is simply converting XFAIL -> FAIL.  That's correct, but it
doesn't leave a good enough trail for people doing regression analysis.

Could you please file a bug report and then add a comment to the
test script with the bug ID number in it:

  # setup_kfail "gdb/1234"

The bug report can say that this bug is also present in 5.2.

I am willing to file the bug report this evening if you want me to
do that part.

Michael C

RFA: don't xfail ptype test

[Jim Blandy]

2002-05-10  Jim Blandy  <jimb@redhat.com>

	* gdb.base/printcmds.exp: Don't xfail the ptype command.  This is
	a bug.

Index: gdb/testsuite/gdb.base/printcmds.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.base/printcmds.exp,v
retrieving revision 1.7
diff -c -r1.7 printcmds.exp
*** gdb/testsuite/gdb.base/printcmds.exp	4 May 2002 15:18:21 -0000	1.7
--- gdb/testsuite/gdb.base/printcmds.exp	10 May 2002 16:13:09 -0000
***************
*** 627,633 ****
      gdb_test "p *\"foo\""		" = 102 'f'"
      gdb_test "ptype *\"foo\""		" = char"
      gdb_test "p &*\"foo\""		" = \"foo\""
-     setup_xfail "*-*-*"
      gdb_test "ptype &*\"foo\""	"type = char \\*"
      gdb_test "p (char *)\"foo\""	" = \"foo\""
  }
--- 627,632 ----