[RFA/testsuite] Test printing complex numbers in C

[RFA/testsuite] Test printing complex numbers in C

[Daniel Jacobowitz]

Fernando,

I think this test submission got lost, so here it is again.  It's a test for
the printing of complex numbers in C, for those compilers which support it. 
The test is GCC specific right now.

[By the way, it fails on GCC 3.0/stabs because of a (probably) GCC bug in
debug information.  Still not fixed in 3.2.  It also fails in GCC
2.95/DWARF-2 because of a bug with __attribute__((packed)).  We can sort out
XFAILs later, I have some ideas on that front.]

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer


2002-03-20  Daniel Jacobowitz  <drow@mvista.com>

	* gdb.base/complex.c: New file.
	* gdb.base/complex.exp: New file.

--- /dev/null	Wed Dec 31 19:00:00 1969
+++ gdb/testsuite/gdb.base/complex.c	Wed Mar 20 19:51:21 2002
@@ -0,0 +1,49 @@
+/* Test taken from GCC.  Verify that we can print a structure containing
+   a complex number.  */
+
+typedef __complex__ float cf;
+struct x { char c; cf f; } __attribute__ ((__packed__));
+struct unpacked_x { char c; cf f; };
+extern void f4 (struct unpacked_x*);
+extern void f3 (void);
+extern void f2 (struct x*);
+extern void f1 (void);
+int
+main (void)
+{
+  f1 ();
+  f3 ();
+  exit (0);
+}
+
+void
+f1 (void)
+{
+  struct x s;
+  s.f = 1;
+  s.c = 42;
+  f2 (&s);
+}
+
+void
+f2 (struct x *y)
+{
+  if (y->f != 1 || y->c != 42)
+    abort ();
+}
+
+void
+f3 (void)
+{
+  struct unpacked_x s;
+  s.f = 1;
+  s.c = 42;
+  f4 (&s);
+}
+
+void
+f4 (struct unpacked_x *y)
+{
+  if (y->f != 1 || y->c != 42)
+    abort ();
+}
--- /dev/null	Wed Dec 31 19:00:00 1969
+++ gdb/testsuite/gdb.base/complex.exp	Wed Mar 20 19:50:45 2002
@@ -0,0 +1,56 @@
+# Copyright 2002 Free Software Foundation, Inc.
+
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2 of the License, or
+# (at your option) any later version.
+# 
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+# 
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software
+# Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.  
+
+# Please email any bugs, comments, and/or additions to this file to:
+# bug-gdb@prep.ai.mit.edu
+
+if $tracelevel then {
+	strace $tracelevel
+}
+
+set bug_id 0
+
+set testfile complex
+set srcfile ${testfile}.c
+set binfile ${objdir}/${subdir}/${testfile}
+set options debug
+
+if  { [gdb_compile "${srcdir}/${subdir}/${srcfile}" "${binfile}" executable $options] != "" } {
+  # No support for __complex__, presumably.
+  unsupported "print complex packed value in C"
+  unsupported "print complex value in C"
+  return 0
+}
+
+# Start with a fresh gdb.
+
+gdb_exit
+gdb_start
+gdb_reinitialize_dir $srcdir/$subdir
+gdb_load $binfile
+
+if [runto f2] then {
+  gdb_test "p *y" "\\\$\[0-9\]* = \{c = 42 '\\*', f = 1 \\+ 0 \\* I\}" \
+	"print complex packed value in C"
+}
+
+if [runto f4] then {
+  gdb_test "p *y" "\\\$\[0-9\]* = \{c = 42 '\\*', f = 1 \\+ 0 \\* I\}" \
+	"print complex value in C"
+
+}
+
+return 0

Re: [RFA/testsuite] Test printing complex numbers in C

[Fernando Nasser]

Sorry Daniel, due to a problem with my mail it got lost again.

I am OK with the new tests (thanks!), but I would like to 
"setup_xfail_format" them for gcc stabs.  It is a compiler problem, not 
GDB's, so we should not add the FAILs to our output.

Can you do that?

Regards,
Fernando

P.S.: You must call "get_debug_format" somewhere before using 
"setup_xfail_format" if I remember correctly.



Daniel Jacobowitz wrote:
> Fernando,
> 
> I think this test submission got lost, so here it is again.  It's a test for
> the printing of complex numbers in C, for those compilers which support it. 
> The test is GCC specific right now.
> 
> [By the way, it fails on GCC 3.0/stabs because of a (probably) GCC bug in
> debug information.  Still not fixed in 3.2.  It also fails in GCC
> 2.95/DWARF-2 because of a bug with __attribute__((packed)).  We can sort out
> XFAILs later, I have some ideas on that front.]
> 


-- 
Fernando Nasser
Red Hat Canada Ltd.                     E-Mail:  fnasser@redhat.com
2323 Yonge Street, Suite #300
Toronto, Ontario   M4P 2C9

Re: [RFA/testsuite] Test printing complex numbers in C

[Daniel Jacobowitz]

On Tue, Dec 03, 2002 at 11:34:32AM -0500, Fernando Nasser wrote:
> Sorry Daniel, due to a problem with my mail it got lost again.
> 
> I am OK with the new tests (thanks!), but I would like to 
> "setup_xfail_format" them for gcc stabs.  It is a compiler problem, not 
> GDB's, so we should not add the FAILs to our output.
> 
> Can you do that?
> 
> Regards,
> Fernando
> 
> P.S.: You must call "get_debug_format" somewhere before using 
> "setup_xfail_format" if I remember correctly.

Here's the problem...
		stabs		DWARF-2
GCC 2.95	pass		fail
GCC 3.x		fail		pass

So what to do?  I suppose I could add a check for the major version of
GCC in get_compiler_info (?) and condition on that.

> 
> 
> 
> Daniel Jacobowitz wrote:
> >Fernando,
> >
> >I think this test submission got lost, so here it is again.  It's a test 
> >for
> >the printing of complex numbers in C, for those compilers which support 
> >it. The test is GCC specific right now.
> >
> >[By the way, it fails on GCC 3.0/stabs because of a (probably) GCC bug in
> >debug information.  Still not fixed in 3.2.  It also fails in GCC
> >2.95/DWARF-2 because of a bug with __attribute__((packed)).  We can sort 
> >out
> >XFAILs later, I have some ideas on that front.]
> >
> 
> 
> -- 
> Fernando Nasser
> Red Hat Canada Ltd.                     E-Mail:  fnasser@redhat.com
> 2323 Yonge Street, Suite #300
> Toronto, Ontario   M4P 2C9
> 
> 

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: [RFA/testsuite] Test printing complex numbers in C

[Fernando Nasser]

Daniel Jacobowitz wrote:
> (...) I suppose I could add a check for the major version of
> GCC in get_compiler_info (?) and condition on that.
> 

I wonder if we wouldn't do something with regards to compiler (GCC) versions as 
well.  We seem to have several failures depending on the compiler version (as 
Michael pointed in an unrelated thread, the number of failures depends on the 
compiler version used, in addition to the debug format).

Although in many cases we can add a pattern looking for the wrong output, 
conditioning on the version would be more reliable (so we don't XFAIL for other 
versions and correctly flag a regression).

What do you think Michael?

-- 
Fernando Nasser
Red Hat - Toronto                       E-Mail:  fnasser@redhat.com
2323 Yonge Street, Suite #300
Toronto, Ontario   M4P 2C9

Re: [RFA/testsuite] Test printing complex numbers in C

[Fernando Nasser]

Hi Daniel,

Can you make the change to xfail conditionally on compiler version (see Michael 
note) and format?

Thanks.

Regards,
Fernando

Daniel Jacobowitz wrote:> Fernando,
> 
> I think this test submission got lost, so here it is again.  It's a test for
> the printing of complex numbers in C, for those compilers which support it. 
> The test is GCC specific right now.
> 
> [By the way, it fails on GCC 3.0/stabs because of a (probably) GCC bug in
> debug information.  Still not fixed in 3.2.  It also fails in GCC
> 2.95/DWARF-2 because of a bug with __attribute__((packed)).  We can sort out
> XFAILs later, I have some ideas on that front.]
> 



-- 
Fernando Nasser
Red Hat - Toronto                       E-Mail:  fnasser@redhat.com
2323 Yonge Street, Suite #300
Toronto, Ontario   M4P 2C9

Re: [RFA/testsuite] Test printing complex numbers in C

[Michael Elizabeth Chastain]

		  stabs           DWARF-2
  GCC 2.95        pass            fail
  GCC 3.x         fail            pass

Do gcc 2.95.3/dwarf-2 and gcc 3.x/stabs+ have identical output,
or different output?

If they are different output, you can do:

  send_gdb "blah"
  gdb_expect {
    -re "good output" {
      pass ...
    }
    -re "flaky gcc 2.95.3/dwarf-2 output" {
      setup_xfail_format "dwarf-2"
      fail ...
    }
    -re "flaky gcc 3.x/stabs+ output" {
      setup_xfail_format "stabs+"
      fail ...
    }
    -re ".*" {
      fail ...
    }
  }

You can also dynamically poke at the executable to see what gcc compiled
it.  See "probe_demangler" in gdb.c++/cplusfuncs.exp for one way to do it.

Michael C

Re: [RFA/testsuite] Test printing complex numbers in C

[Fernando Nasser]

Thanks Michael.  Nice to have you back!

Michael Elizabeth Chastain wrote:> 		  stabs           DWARF-2
>   GCC 2.95        pass            fail
>   GCC 3.x         fail            pass
> 
> Do gcc 2.95.3/dwarf-2 and gcc 3.x/stabs+ have identical output,
> or different output?
> 
> If they are different output, you can do:
> 
>   send_gdb "blah"
>   gdb_expect {
>     -re "good output" {
>       pass ...
>     }
>     -re "flaky gcc 2.95.3/dwarf-2 output" {
>       setup_xfail_format "dwarf-2"
>       fail ...
>     }
>     -re "flaky gcc 3.x/stabs+ output" {
>       setup_xfail_format "stabs+"
>       fail ...
>     }
>     -re ".*" {
>       fail ...
>     }
>   }
> 
> You can also dynamically poke at the executable to see what gcc compiled
> it.  See "probe_demangler" in gdb.c++/cplusfuncs.exp for one way to do it.
> 
> Michael C
> 



-- 
Fernando Nasser
Red Hat - Toronto                       E-Mail:  fnasser@redhat.com
2323 Yonge Street, Suite #300
Toronto, Ontario   M4P 2C9

Re: [RFA/testsuite] Test printing complex numbers in C

[Michael Elizabeth Chastain]

Fernando Nasser writes:
> I wonder if we wouldn't do something with regards to compiler (GCC)
> versions as well.

I think we should bite the bullet and add something to lib/gdb.exp
to figure out the compiler version.

In fact, I see that we already do!  The variable "gcc_compiled" is set
to __GNUC__ if that is defined, or 0 if nothing is defined.  "info cpp"
says that __GNUC__ equals the major version of the compiler: 2 for gcc
2.x and 3 for gcc 3.x.  This is a documented feature so we can rely on
it being the major version number for all gcc's forever.

So "gcc_compiled" is more than a boolean, it actually has the values:

  0  not gnu c
  2  gcc v2
  3  gcc v3

I haven't actually tested this because I'm kinda tired and my test bed
is cranking on 5.2.91 and I don't want to get close to its playpen
and I don't want to start a different test bed.

As far as using it goes, I agree with Fernando, it's usually a good idea,
because it increases the specificity of the test suite.  Also it makes
the test suite a lot easier to maintain because it's clear which results
are expected from which compilers, so that people can read it easier,
and obsolete compilers can be de-supported more easily.

Michael C