public inbox for gcc@gcc.gnu.org
 help / color / mirror / Atom feed
* testsuite results 971201
@ 1997-12-02  6:35 schmid
  1997-12-02  9:23 ` Joe Buck
  1997-12-02 11:25 ` Manfred.Hollstein
  0 siblings, 2 replies; 4+ messages in thread
From: schmid @ 1997-12-02  6:35 UTC (permalink / raw)
  To: egcs

These are the results of the gcc, g++ and g77 testsuite for egcs-2.90.20.
I am using libc-5.4.7 and binutils 2.8.1.0.15.

The XPASS:g++.jason/destruct3.C - (test for bogus messages, line 38)
is an internal compiler error in disguise:

Native configuration is i586-pc-linux-gnulibc1
 on host: /home/schmid/ftp/egcs-971201/gcc/testsuite/../xgcc -B/home/schmid/ftp/egcs-971201/gcc/testsuite/../ /home/schmid/ftp/egcs-971201/gcc/testsuite/g++.old-deja/g++.jason/destruct3.C  -ansi -pedantic-errors  -I/home/schmid/ftp/egcs-971201/libio -I/home/schmid/ftp/egcs-971201/libraries/libio -I/home/schmid/ftp/egcs-971201/libstdc++ -I/home/schmid/ftp/egcs-971201/libstdc++/stl -I/home/schmid/ftp/egcs-971201/gcc/testsuite/../include -c  -o /tmp/destruct3.o   
compiler exited with status 1
output is:
/home/schmid/ftp/egcs-971201/gcc/testsuite/g++.old-deja/g++.jason/destruct3.C:35:
warning: direct base `ccUnwind' inaccessible in `ccVSTool' due to
ambiguity 
/home/schmid/ftp/egcs-971201/gcc/testsuite/g++.old-deja/g++.jason/destruct3.C:
35:Internal compiler error 369.
/home/schmid/ftp/egcs-971201/gcc/testsuite/g++.old-deja/g++.jason/destruct3.C:
35: Please submit a full bug report to `egcs-bugs@cygnus.com'.
Are you going to fix this?
At least the compiler should not crash.

Peter Schmid 


Native configuration is i586-pc-linux-gnulibc1

		=== gcc tests ===


		=== gcc Summary ===

# of expected passes		4883
# of expected failures		5
# of unsupported tests		7
/home/schmid/ftp/egcs-971201/gcc/xgcc version egcs-2.90.20 971201 (gcc2-970802 experimental)

		=== g++ tests ===

		=== g++ Summary ===

# of expected passes		3400
# of unexpected successes	3
# of expected failures		80
# of untested testcases		6
/home/schmid/ftp/egcs-971201/gcc/testsuite/../xgcc version egcs-2.90.20 971201 (gcc2-970802 experimental)
XPASS: g++.jason/destruct3.C - (test for bogus messages, line 38)
XPASS: g++.mike/dyncast1.C  Execution test
XPASS: g++.mike/dyncast2.C  Execution test





		=== g77 tests ===

# of expected passes		130
# of unexpected failures	2
/home/schmid/ftp/egcs-971201/gcc/g77 version egcs-2.90.20 971201 (gcc2-970802 experimental)
FAIL: g77.f-torture/execute/dnrm2.f execution,  -O2 -fomit-frame-pointer -finline-functions -funroll-loops 
FAIL: g77.f-torture/execute/dnrm2.f execution,  -O2 -fomit-frame-pointer -finline-functions -funroll-all-loops 


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: testsuite results 971201
  1997-12-02  6:35 testsuite results 971201 schmid
@ 1997-12-02  9:23 ` Joe Buck
  1997-12-02 11:25 ` Manfred.Hollstein
  1 sibling, 0 replies; 4+ messages in thread
From: Joe Buck @ 1997-12-02  9:23 UTC (permalink / raw)
  To: schmid; +Cc: egcs

> The XPASS:g++.jason/destruct3.C - (test for bogus messages, line 38)
> is an internal compiler error in disguise:

Yikes!  dejagnu should never allow this to happen for tests that
expect an error (though, knowing nothing about dejagnu it's hard
to say what to do).

This was noted because it is an XPASS, but would not be noticed
at all if it were a PASS.

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: testsuite results 971201
  1997-12-02  6:35 testsuite results 971201 schmid
  1997-12-02  9:23 ` Joe Buck
@ 1997-12-02 11:25 ` Manfred.Hollstein
  1 sibling, 0 replies; 4+ messages in thread
From: Manfred.Hollstein @ 1997-12-02 11:25 UTC (permalink / raw)
  To: schmid; +Cc: egcs

On Tue, 2 December 1997, 15:07:44, schmid@ltoi.iap.physik.tu-darmstadt.de wrote:

 > /home/schmid/ftp/egcs-971201/gcc/testsuite/g++.old-deja/g++.jason/destruct3.C:
 > 35:Internal compiler error 369.
 > /home/schmid/ftp/egcs-971201/gcc/testsuite/g++.old-deja/g++.jason/destruct3.C:
 > 35: Please submit a full bug report to `egcs-bugs@cygnus.com'.
 > Are you going to fix this?
 > At least the compiler should not crash.
 > 

This one has been fixed for me by Mike Stump's patch:

On Thu, 6 November 1997, 15:37:39, mrs@wrs.com wrote:

    This fixes up my previous patch to handle an additional case, so that
    we are almost never limited in complex MI cases for vtable names.
    This preserves backwards compatibility.

    Let me know when it goes in, thanks.

    See g++.jason/destruct3.C for testcase.

        sethi %hi(_vt.8ccVSTool.8ccUnwind),%o1
        or %o1,%lo(_vt.8ccVSTool.8ccUnwind),%o0
        st %o0,[%l0+4]
        sethi %hi(_vt.8ccVSTool.8ccUnwind.2),%o1
        or %o1,%lo(_vt.8ccVSTool.8ccUnwind.2),%o0
        st %o0,[%l0+8]
        sethi %hi(_vt.8ccVSTool),%o1
        or %o1,%lo(_vt.8ccVSTool),%o0


1997-11-06  Mike Stump  <mrs@wrs.com>

	* class.c (prepare_fresh_vtable): Enable even more complex MI
	vtable names.

Doing diffs in cp:
*** cp/class.c.~1~	Fri Oct 31 14:10:08 1997
--- cp/class.c	Thu Nov  6 15:13:21 1997
*************** prepare_fresh_vtable (binfo, for_type)
*** 788,794 ****
  
    while (1)
      {
!       char *buf1 = (char *) alloca (TYPE_ASSEMBLER_NAME_LENGTH (for_type) + 1 + i);
        char *new_buf2;
  
        sprintf (buf1, "%s%c%s", TYPE_ASSEMBLER_NAME_STRING (for_type), joiner,
--- 788,795 ----
  
    while (1)
      {
!       char *buf1 = (char *) alloca (TYPE_ASSEMBLER_NAME_LENGTH (for_type)
! 				    + 1 + i);
        char *new_buf2;
  
        sprintf (buf1, "%s%c%s", TYPE_ASSEMBLER_NAME_STRING (for_type), joiner,
*************** prepare_fresh_vtable (binfo, for_type)
*** 812,819 ****
  
        basetype = TYPE_MAIN_VARIANT (BINFO_TYPE (path));
  
!       /* We better not run out of stuff to make it unique.  */
!       my_friendly_assert (for_type != basetype, 369);
  
        i = TYPE_ASSEMBLER_NAME_LENGTH (basetype) + 1 + i;
        new_buf2 = (char *) alloca (i);
--- 813,846 ----
  
        basetype = TYPE_MAIN_VARIANT (BINFO_TYPE (path));
  
!       if (for_type == basetype)
! 	{
! 	  /* If we run out of basetypes in the path, we have already
! 	     found created a vtable with that name before, we now
! 	     resort to tacking on _%d to distinguish them.  */
! 	  int j = 2;
! 	  i = TYPE_ASSEMBLER_NAME_LENGTH (basetype) + 1 + i + 1 + 3;
! 	  buf1 = (char *) alloca (i);
! 	  do {
! 	    sprintf (buf1, "%s%c%s%c%d",
! 		     TYPE_ASSEMBLER_NAME_STRING (basetype), joiner,
! 		     buf2, joiner, j);
! 	    buf = (char *) alloca (strlen (VTABLE_NAME_FORMAT)
! 				   + strlen (buf1) + 1);
! 	    sprintf (buf, VTABLE_NAME_FORMAT, buf1);
! 	    name = get_identifier (buf);
! 
! 	    /* If this name doesn't clash, then we can use it,
! 	       otherwise we add something different to the name until
! 	       it is unique.  */
! 	  } while (++j <= 999 && IDENTIFIER_GLOBAL_VALUE (name));
! 
! 	  /* Hey, they really like MI don't they?  Increase the 3
!              above to 6, and the 999 to 999999.  :-)  */
! 	  my_friendly_assert (j <= 999, 369);
! 
! 	  break;
! 	}
  
        i = TYPE_ASSEMBLER_NAME_LENGTH (basetype) + 1 + i;
        new_buf2 = (char *) alloca (i);
--------------
Doing diffs in testsuite:
*** testsuite/g++.old-deja/g++.jason/destruct3.C.~1~	Mon Aug 25 08:35:39 1997
--- testsuite/g++.old-deja/g++.jason/destruct3.C	Thu Nov  6 15:31:08 1997
***************
*** 1,3 ****
--- 1,4 ----
+ // Special g++ Options: -w
  // PRMS Id: 4342 (second testcase)
  // Bug: g++ still can't deal with ambiguous inheritance in destructor calls.
  // Build don't link:
*************** struct ccScreenObj : public ccScreenObjR
*** 32,41 ****
  {};
  
  struct ccVSTool : public ccImpExp, public ccUnwind 
! {};	// gets bogus error - XFAIL *-*-*
  
  struct ccSCCP : public ccVSTool
! {};	// gets bogus error - XFAIL *-*-*
  
  void foo ()
  {
--- 33,42 ----
  {};
  
  struct ccVSTool : public ccImpExp, public ccUnwind 
! {};
  
  struct ccSCCP : public ccVSTool
! {};
  
  void foo ()
  {

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: testsuite results 971201
@ 1997-12-03  0:21 Mike Stump
  0 siblings, 0 replies; 4+ messages in thread
From: Mike Stump @ 1997-12-03  0:21 UTC (permalink / raw)
  To: egcs, schmid

I think this fixes it, I sent it in before
( http://www.cygnus.com/ml/egcs/1997-Nov/0234.html ), and haven't heard
back.  Jason?  As you can see, this patch is very safe, it cannot hit
unless it would have dumped before.  It'd be nice to have it in the
release, but it isn't critical.

1997-11-06  Mike Stump  <mrs@wrs.com>

	* class.c (prepare_fresh_vtable): Enable even more complex MI
	vtable names.

Doing diffs in cp:
*** cp/class.c.~1~	Fri Oct 31 14:10:08 1997
--- cp/class.c	Thu Nov  6 15:13:21 1997
*************** prepare_fresh_vtable (binfo, for_type)
*** 788,794 ****
  
    while (1)
      {
!       char *buf1 = (char *) alloca (TYPE_ASSEMBLER_NAME_LENGTH (for_type) + 1 + i);
        char *new_buf2;
  
        sprintf (buf1, "%s%c%s", TYPE_ASSEMBLER_NAME_STRING (for_type), joiner,
--- 788,795 ----
  
    while (1)
      {
!       char *buf1 = (char *) alloca (TYPE_ASSEMBLER_NAME_LENGTH (for_type)
! 				    + 1 + i);
        char *new_buf2;
  
        sprintf (buf1, "%s%c%s", TYPE_ASSEMBLER_NAME_STRING (for_type), joiner,
*************** prepare_fresh_vtable (binfo, for_type)
*** 812,819 ****
  
        basetype = TYPE_MAIN_VARIANT (BINFO_TYPE (path));
  
!       /* We better not run out of stuff to make it unique.  */
!       my_friendly_assert (for_type != basetype, 369);
  
        i = TYPE_ASSEMBLER_NAME_LENGTH (basetype) + 1 + i;
        new_buf2 = (char *) alloca (i);
--- 813,846 ----
  
        basetype = TYPE_MAIN_VARIANT (BINFO_TYPE (path));
  
!       if (for_type == basetype)
! 	{
! 	  /* If we run out of basetypes in the path, we have already
! 	     found created a vtable with that name before, we now
! 	     resort to tacking on _%d to distinguish them.  */
! 	  int j = 2;
! 	  i = TYPE_ASSEMBLER_NAME_LENGTH (basetype) + 1 + i + 1 + 3;
! 	  buf1 = (char *) alloca (i);
! 	  do {
! 	    sprintf (buf1, "%s%c%s%c%d",
! 		     TYPE_ASSEMBLER_NAME_STRING (basetype), joiner,
! 		     buf2, joiner, j);
! 	    buf = (char *) alloca (strlen (VTABLE_NAME_FORMAT)
! 				   + strlen (buf1) + 1);
! 	    sprintf (buf, VTABLE_NAME_FORMAT, buf1);
! 	    name = get_identifier (buf);
! 
! 	    /* If this name doesn't clash, then we can use it,
! 	       otherwise we add something different to the name until
! 	       it is unique.  */
! 	  } while (++j <= 999 && IDENTIFIER_GLOBAL_VALUE (name));
! 
! 	  /* Hey, they really like MI don't they?  Increase the 3
!              above to 6, and the 999 to 999999.  :-)  */
! 	  my_friendly_assert (j <= 999, 369);
! 
! 	  break;
! 	}
  
        i = TYPE_ASSEMBLER_NAME_LENGTH (basetype) + 1 + i;
        new_buf2 = (char *) alloca (i);
--------------
Doing diffs in testsuite:
*** testsuite/g++.old-deja/g++.jason/destruct3.C.~1~	Mon Aug 25 08:35:39 1997
--- testsuite/g++.old-deja/g++.jason/destruct3.C	Thu Nov  6 15:31:08 1997
***************
*** 1,3 ****
--- 1,4 ----
+ // Special g++ Options: -w
  // PRMS Id: 4342 (second testcase)
  // Bug: g++ still can't deal with ambiguous inheritance in destructor calls.
  // Build don't link:
*************** struct ccScreenObj : public ccScreenObjR
*** 32,41 ****
  {};
  
  struct ccVSTool : public ccImpExp, public ccUnwind 
! {};	// gets bogus error - XFAIL *-*-*
  
  struct ccSCCP : public ccVSTool
! {};	// gets bogus error - XFAIL *-*-*
  
  void foo ()
  {
--- 33,42 ----
  {};
  
  struct ccVSTool : public ccImpExp, public ccUnwind 
! {};
  
  struct ccSCCP : public ccVSTool
! {};
  
  void foo ()
  {
--------------

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~1997-12-03  0:21 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
1997-12-02  6:35 testsuite results 971201 schmid
1997-12-02  9:23 ` Joe Buck
1997-12-02 11:25 ` Manfred.Hollstein
1997-12-03  0:21 Mike Stump

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).