[Bug libobjc/21992] New: [4.1 regression] mmix-knuth-mmixware testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[Bug libobjc/21992] New: [4.1 regression] mmix-knuth-mmixware testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[hp at gcc dot gnu dot org]

With LAST_UPDATED: "Thu Jun  9 21:38:48 UTC 2005" I get:
FAIL: objc/execute/_cmd.m execution,  -O0
FAIL: objc/execute/_cmd.m execution,  -O1
FAIL: objc/execute/_cmd.m execution,  -O2
FAIL: objc/execute/_cmd.m execution,  -O3 -fomit-frame-pointer
FAIL: objc/execute/_cmd.m execution,  -O3 -g
FAIL: objc/execute/_cmd.m execution,  -Os
FAIL: objc/execute/accessing_ivars.m execution,  -O2
FAIL: objc/execute/accessing_ivars.m execution,  -O3 -fomit-frame-pointer
FAIL: objc/execute/accessing_ivars.m execution,  -O3 -g
FAIL: objc/execute/accessing_ivars.m execution,  -Os
FAIL: objc/execute/bycopy-2.m execution,  -O2
FAIL: objc/execute/bycopy-2.m execution,  -O3 -fomit-frame-pointer
FAIL: objc/execute/bycopy-2.m execution,  -O3 -g
FAIL: objc/execute/bycopy-2.m execution,  -Os
FAIL: objc/execute/bycopy-3.m execution,  -O2
FAIL: objc/execute/bycopy-3.m execution,  -O3 -fomit-frame-pointer
FAIL: objc/execute/bycopy-3.m execution,  -O3 -g
FAIL: objc/execute/bycopy-3.m execution,  -Os
FAIL: objc/execute/cascading-1.m execution,  -O2
FAIL: objc/execute/cascading-1.m execution,  -O3 -fomit-frame-pointer
FAIL: objc/execute/cascading-1.m execution,  -O3 -g
FAIL: objc/execute/cascading-1.m execution,  -Os
FAIL: objc/execute/class-1.m execution,  -O2
FAIL: objc/execute/class-1.m execution,  -O3 -fomit-frame-pointer
FAIL: objc/execute/class-1.m execution,  -O3 -g
FAIL: objc/execute/class-1.m execution,  -Os
FAIL: objc/execute/class-10.m execution,  -O2
FAIL: objc/execute/class-10.m execution,  -O3 -fomit-frame-pointer
FAIL: objc/execute/class-10.m execution,  -O3 -g
FAIL: objc/execute/class-10.m execution,  -Os
FAIL: objc/execute/class-11.m execution,  -O2
FAIL: objc/execute/class-11.m execution,  -O3 -fomit-frame-pointer
FAIL: objc/execute/class-11.m execution,  -O3 -g
FAIL: objc/execute/class-11.m execution,  -Os
FAIL: objc/execute/class-12.m execution,  -O2
FAIL: objc/execute/class-12.m execution,  -O3 -fomit-frame-pointer
FAIL: objc/execute/class-12.m execution,  -O3 -g
FAIL: objc/execute/class-12.m execution,  -Os
FAIL: objc/execute/class-13.m execution,  -O2
FAIL: objc/execute/class-13.m execution,  -O3 -fomit-frame-pointer
FAIL: objc/execute/class-13.m execution,  -O3 -g
FAIL: objc/execute/class-13.m execution,  -Os
FAIL: objc/execute/class-14.m execution,  -O2
FAIL: objc/execute/class-14.m execution,  -O3 -fomit-frame-pointer
FAIL: objc/execute/class-14.m execution,  -O3 -g
FAIL: objc/execute/class-14.m execution,  -Os
FAIL: objc/execute/class-2.m execution,  -O2
FAIL: objc/execute/class-2.m execution,  -O3 -fomit-frame-pointer
FAIL: objc/execute/class-2.m execution,  -O3 -g
FAIL: objc/execute/class-2.m execution,  -Os
FAIL: objc/execute/class-3.m execution,  -O2
FAIL: objc/execute/class-3.m execution,  -O3 -fomit-frame-pointer
FAIL: objc/execute/class-3.m execution,  -O3 -g
FAIL: objc/execute/class-3.m execution,  -Os
FAIL: objc/execute/class-4.m execution,  -O2
(list pruned, 279 unexpected fails where none previously)

With the message in the .log being for all or most:
objc runtime: cannot find class Object
*** EXIT code 4242

Last known to work on LAST_UPDATED: "Thu Jun  9 10:30:36 UTC 2005".

-- 
           Summary: [4.1 regression] mmix-knuth-mmixware testsuite failure:
                    many objc execution tests fail, first
                    objc/execute/_cmd.m execution
           Product: gcc
           Version: 4.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: libobjc
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: hp at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: mmix-knuth-mmixware


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] mmix-knuth-mmixware testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-06-10 00:41 -------
Hmm, there were no libobjc changes.
There two patches which might have caused it:
-2005-06-09  Jan Hubicka  <jh@suse.cz>
-
-       * cgraphunit.c (cgraph_create_edges): Do not walk BLOCK; finalize
-       local statics when doing unit-at-a-time.
-       (cgraph_varpool_assemble_pending_decls): Output debug info.
-       * dwarf2out.c (decls_for_scope): Skip local statics.
-       (dwarf2out_decl): Handle local statics.
-       * passes.c (rest_of_decl_compilation): Do not differentiate
-       local and global statics in unit-at-a-time.
-       * tree-inline.c (remap_decls): Put local static into
-       unexpanded_vars_list rather than introducing duplicated VAR_DECL
-       node.
-


And:
-2005-06-08  Daniel Berlin  <dberlin@dberlin.org>
-
-       * Makefile.in (OBJS-common): Add tree-ssa-structalias.o.
-       * tree-flow.h (find_what_p_points_to): Add prototype.
-       (push_fields_onto_fieldstack): Ditto.
-       (sort_fieldstack): Ditto.
-       * tree-optimize.c (init_tree_optimization_passes): Add
-       pass_build_pta and pass_del_pta.
-       * tree-pass.h (pass_build_pta): New structure.
-       (pass_del_pta): Ditto.
-       * tree-ssa-alias.c (compute_flow_sensitive_aliasing): Disambiguate
-       using new alias analyzer.
-       (push_fields_onto_fieldstack): Removed from here.
-       (bitpos_of_field): Ditto.
-       (fieldoff_compare): Ditto.
-       * tree-ssa-structalias.c: New file.
-       * tree-ssa-structalias.h: Ditto.
-


I really doubt it was caused by DannyB's patch though

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pinskia at gcc dot gnu dot
                   |                            |org, hubicka at gcc dot gnu
                   |                            |dot org
          Component|libobjc                     |middle-end
           Keywords|                            |wrong-code
   Target Milestone|---                         |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[hp at gcc dot gnu dot org]

------- Additional Comments From hp at gcc dot gnu dot org  2005-06-18 16:10 -------
Also observed on cris-elf and cris-axis-linux-gnu, same timeframe and ever
since.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2005-06-18 16:10:15
               date|                            |
            Summary|[4.1 regression] mmix-knuth-|[4.1 regression] testsuite
                   |mmixware testsuite failure: |failure: many objc execution
                   |many objc execution tests   |tests fail, first
                   |fail, first                 |objc/execute/_cmd.m
                   |objc/execute/_cmd.m         |execution
                   |execution                   |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[hp at gcc dot gnu dot org]

------- Additional Comments From hp at gcc dot gnu dot org  2005-07-04 23:21 -------
There was a change of behavior between
LAST_UPDATED "Sun Jun 19 20:05:45 UTC 2005" and "Mon Jun 27 21:26:27 UTC 2005";
some tests now fail at -O1.  So whatever pass now runs at -O1 that used to run
at -O2 and higher seems the culprit.  With "Mon Jul  4 10:24:52 UTC 2005" these
tests still fail.  Complete list of new objc failures introduced between the two
first dates:
objc/execute/accessing_ivars.m execution,  -O1
objc/execute/bycopy-2.m execution,  -O1
objc/execute/bycopy-3.m execution,  -O1
objc/execute/cascading-1.m execution,  -O1
objc/execute/class-1.m execution,  -O1
objc/execute/class-10.m execution,  -O1
objc/execute/class-11.m execution,  -O1
objc/execute/class-12.m execution,  -O1
objc/execute/class-13.m execution,  -O1
objc/execute/class-14.m execution,  -O1
objc/execute/class-2.m execution,  -O1
objc/execute/class-3.m execution,  -O1
objc/execute/class-4.m execution,  -O1
objc/execute/class-5.m execution,  -O1
objc/execute/class-6.m execution,  -O1
objc/execute/class-7.m execution,  -O1
objc/execute/class-8.m execution,  -O1
objc/execute/class-9.m execution,  -O1
objc/execute/compatibility_alias.m execution,  -O1
objc/execute/enumeration-1.m execution,  -O1
objc/execute/enumeration-2.m execution,  -O1
objc/execute/formal_protocol-1.m execution,  -O1
objc/execute/formal_protocol-2.m execution,  -O1
objc/execute/formal_protocol-3.m execution,  -O1
objc/execute/formal_protocol-4.m execution,  -O1
objc/execute/formal_protocol-7.m execution,  -O1
objc/execute/function-message-1.m execution,  -O1
objc/execute/no_clash.m execution,  -O1
objc/execute/object_is_class.m execution,  -O1
objc/execute/object_is_meta_class.m execution,  -O1


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|2005-06-18 16:10:15         |2005-07-04 23:21:14
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-07-04 23:32 -------
(In reply to comment #3)
> There was a change of behavior between
> LAST_UPDATED "Sun Jun 19 20:05:45 UTC 2005" and "Mon Jun 27 21:26:27 UTC 2005";
> some tests now fail at -O1.  So whatever pass now runs at -O1 that used to run
> at -O2 and higher seems the culprit.  With "Mon Jul  4 10:24:52 UTC 2005" these
> tests still fail.  Complete list of new objc failures introduced between the two
> first dates:

That is the date which unit-at-a-time was enabled at -O1.  So it looks like it was Jan's change which 
caused it.  Though I don't know why it passes on i686-pc-linux-gnu.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-07-26 14:32 -------
I think I see the same thing at -O2 on a couple of tests on ppc-darwin.

Is the execution a missing symbol?

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-07-26 15:31 -------
This fails on PPC-darwin also.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
 GCC target triplet|mmix-knuth-mmixware         |mmix-knuth-mmixware,
                   |                            |powerpc-darwin


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-07-26 15:34 -------
I wonder why this does not fail on i686-pc-linux-gnu.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-07-26 15:39 -------
                === objc Summary ===

# of expected passes            1309
# of unexpected failures        328
# of unresolved testcases       3
# of unsupported tests          1

Those are what I get on ppc-darwin, bad, very bad.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-07-26 15:52 -------
I think libobjc is also being miscompiled which is why some fails at -O0.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[hp at gcc dot gnu dot org]

------- Additional Comments From hp at gcc dot gnu dot org  2005-07-26 20:10 -------
Not sure it if matters, but for the record.  I can't really parse the question in
comment #5, but from the log quote in the description; "objc runtime: cannot
find class Object" I guess the answer is "yes".  (It's an execution failure, not
a link failure.)

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] [unit-at-a-time] testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-09 04:28 -------
The different betwen for the following code:
#include <objc/Object.h>
int main(void)
{
  [Object name];
}

at -funit-at-a-time and -fno-unit-at-a-time is:
the removal of at unit-at-a-time:
___objc_class_ref_Object:
	.long	___objc_class_name_Object


Now to figure out how to mark this reference as used.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
   GCC host triplet|i686-pc-linux-gnu           |
            Summary|[4.1 regression] testsuite  |[4.1 regression] [unit-at-a-
                   |failure: many objc execution|time] testsuite failure:
                   |tests fail, first           |many objc execution tests
                   |objc/execute/_cmd.m         |fail, first
                   |execution                   |objc/execute/_cmd.m
                   |                            |execution


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] [unit-at-a-time] testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-09 04:56 -------
Created an attachment (id=9449)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=9449&action=view)
patch which should fix it

This patch should fix the bug.	I am going to test on ppc-darwin but it would
be nice to double check, it also fixes it on mmix.  Could you test it for me?

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |pinskia at gcc dot gnu dot
                   |dot org                     |org
             Status|NEW                         |ASSIGNED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] [unit-at-a-time] testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-09 12:11 -------
(In reply to comment #13)
> This patch onle fixes the problem in comment #11 and not all the testcases.
The resulting Summary:
# of expected passes            1528
# of unexpected failures        109
# of unresolved testcases       3
# of unsupported tests          1

So it does in fact improve the situation.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] [unit-at-a-time] testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-09 16:43 -------
(In reply to comment #14)
> So it does in fact improve the situation.

Ok, the majority of the rest of the objc failures are caused by not initializing the runtime on ppc-
darwin, I think I can fix the ppc-darwin problem by making libobjc one module but that will not fix the 
failures on mmix.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] [unit-at-a-time] testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-08-09 17:53 -------
Subject: Bug 21992

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	pinskia@gcc.gnu.org	2005-08-09 17:53:28

Modified files:
	gcc/objc       : ChangeLog objc-act.c 

Log message:
	2005-08-09  Andrew Pinski  <pinskia@physics.uc.edu>
	
	part of PR objc/21992
	* objc-act.c (handle_class_ref): The ref decl is always referenced.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/objc/ChangeLog.diff?cvsroot=gcc&r1=1.54&r2=1.55
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/objc/objc-act.c.diff?cvsroot=gcc&r1=1.288&r2=1.289



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] [unit-at-a-time] testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-09 17:58 -------
I committed the fix for the ones which fail at optimizations level higher than -O0.

Hans-Peter, is the rest of the failures gone for mmix?

Also does mmix have shared libraries, if not then this is a problem as we then are removing things 
which should have forced the link with the Object class, there might be other ways around it still but I 
have to think of them.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] [unit-at-a-time] testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[hp at gcc dot gnu dot org]

------- Additional Comments From hp at gcc dot gnu dot org  2005-08-09 23:28 -------
The patch in comment #12 fixes *all* objc regressions
(actually no failures at all again now), on mmix-knuth-mmixware,
cris-axis-elf and cris-axis-linux-gnu.

No, MMIX does not have shared libraries.  It's not that different
from most of the *-elf "embedded" targets in that sense.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] [unit-at-a-time] testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-09 23:35 -------
Ok, since this was orginally about mmix, I am going to close this as fixed and I will open a new bug 
about the darwin failures once I really test on a clean tree, I know how to fix it if it still broken.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992

[Bug middle-end/21992] [4.1 regression] [unit-at-a-time] testsuite failure: many objc execution tests fail, first objc/execute/_cmd.m execution

[hp at gcc dot gnu dot org]

------- Additional Comments From hp at gcc dot gnu dot org  2005-08-11 16:59 -------
Clarification to comment #18; the newly-enabled-for-all-targets try-catch-8.m
(PR 23306) fails (so there was *one* failure in the objc testsuite at the time
the above commit was made).

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21992