[Bug ada/17793] New: [4.0 Regression] Ada front-end causing bootstrap failure

[Bug ada/17793] New: [4.0 Regression] Ada front-end causing bootstrap failure

[pinskia at gcc dot gnu dot org]

../../xgcc -B../../ -c -g -O2 -W -Wall -gnatpg g-socket.adb -o g-socket.o -fdump-tree-all -v
Reading specs from ../../specs
Configured with: /Volumes/UFS_Partition/pinskia/src/devel/new-fold/gcc/configure --enable-
threads=posix --enable-languages=ada --prefix=/Volumes/UFS_Partition/pinskia/new-fold
Thread model: posix
gcc version 4.0.0 20041001 (experimental)
 ../../gnat1 -feliminate-unused-debug-symbols -fPIC -quiet -dumpbase g-socket.adb -O2 -W -Wall 
-fdump-tree-all -g -gnatpg -gnatO g-socket.o g-socket.adb -o /var/tmp//ccTSUdQN.s
+===========================GNAT BUG 
DETECTED==============================+
| 4.0.0 20041001 (experimental) (powerpc-apple-darwin7.4.1) GCC error:     |
| in check_pointer_types_r, at gimplify.c:4218                             |

Here is the tree which gimplifier is complaining about:
gdb) p debug_tree(t)
 <addr_expr 0x437481a0
    type <pointer_type 0x43747d14
        type <integer_type 0x436682e8 gnat__sockets__to_in_addrGP5369__sourceXn type <integer_type 
0x43653b24 interfaces__c__TintB>
            visited SI
            size <integer_cst 0x4260a500 constant invariant 32>
            unit size <integer_cst 0x4260a140 constant invariant 4>
            user align 32 symtab 0 alias set -1 precision 32 min <integer_cst 0x4364ffc0 -2147483648> 
max <integer_cst 0x436550c0 2147483647> RM size <integer_cst 0x4260a500 32>
            pointer_to_this <pointer_type 0x43747d14>>
        unsigned SI size <integer_cst 0x4260a500 32> unit size <integer_cst 0x4260a140 4>
        align 32 symtab 0 alias set -1>
    invariant
    arg 0 <array_ref 0x436badd0
        type <integer_type 0x43653aa8 interfaces__c__int type <integer_type 0x43653b24 
interfaces__c__TintB>
            public visited SI size <integer_cst 0x4260a500 32> unit size <integer_cst 0x4260a140 4>
            user align 32 symtab 0 alias set -1 precision 32 min <integer_cst 0x4364ffc0 -2147483648> 
max <integer_cst 0x436550c0 2147483647> RM size <integer_cst 0x4260a500 32>
            pointer_to_this <pointer_type 0x43653d14 gnat__sockets__thin__int_access>>
       
        arg 0 <var_decl 0x437396c8 v8 type <array_type 0x4373945c gnat__sockets__thin__two_int>
            addressable DI file g-socket.adb line 1026
            size <integer_cst 0x4260a760 constant invariant 64>
            unit size <integer_cst 0x4260a780 constant invariant 8>
            align 32 context <function_decl 0x426c4d14 gnat__sockets__get_socket_option> chain 
<var_decl 0x43739744 v4>>
        arg 1 <integer_cst 0x4373fa60 constant invariant 0>
        g-socket.adb:1090>>

The integer_type's are not compatible by defined by the front-end, why?

-- 
           Summary: [4.0 Regression] Ada front-end causing bootstrap failure
           Product: gcc
           Version: 4.0.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code, build, ice-checking
          Severity: critical
          Priority: P2
         Component: ada
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: pinskia at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org
GCC target triplet: powerpc-apple-darwin


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17793

[Bug ada/17793] [4.0 Regression] Ada front-end causing bootstrap failure

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.0.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17793

[Bug ada/17793] [4.0 Regression] Ada front-end causing bootstrap failure

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-07 00:16 -------
The line number is 1086:
            Opt.Multicast_Address := To_Inet_Addr (To_In_Addr (V8 (V8'First)));


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17793

[Bug ada/17793] [4.0 Regression] Ada front-end causing bootstrap failure

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
 GCC target triplet|powerpc-apple-darwin        |powerpc-darwin, powerpc-
                   |                            |linux


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17793

[Bug ada/17793] [4.0 Regression] Ada front-end causing bootstrap failure

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-13 23:16 -------
*** Bug 17983 has been marked as a duplicate of this bug. ***

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |schwab at suse dot de


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17793

[Bug ada/17793] [4.0 Regression] Ada front-end causing bootstrap failure

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-13 23:16 -------
Confirmed via the dup bug.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2004-10-13 23:16:36
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17793

[Bug middle-end/17793] [4.0 Regression] Ada bootstrap failure

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-10-14 19:18 -------
> The integer_type's are not compatible by defined by the front-end, why?

The front-end generates a correct tree (with a NOP_EXPR) but it is stripped by
the gimplifier:

      /* Strip away as many useless type conversions as possible
	 at the toplevel.  */
      STRIP_USELESS_TYPE_CONVERSION (*expr_p);


A quick fix is therefore the following:

Index: gimplify.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/gimplify.c,v
retrieving revision 2.82
diff -u -p -r2.82 gimplify.c
--- gimplify.c	30 Sep 2004 01:22:05 -0000	2.82
+++ gimplify.c	14 Oct 2004 15:24:57 -0000
@@ -4156,7 +4156,7 @@ gimplify_one_sizepos (tree *expr_p, tree
 static bool
 cpt_same_type (tree a, tree b)
 {
-  if (lang_hooks.types_compatible_p (a, b))
+  if (a == b || lang_hooks.types_compatible_p (a, b))
     return true;
 
   /* ??? The C++ FE decomposes METHOD_TYPES to FUNCTION_TYPES and doesn't
@@ -4179,7 +4179,8 @@ cpt_same_type (tree a, tree b)
   if (POINTER_TYPE_P (a) && POINTER_TYPE_P (b))
     return cpt_same_type (TREE_TYPE (a), TREE_TYPE (b));
 
-  return false;
+  /* STRIP_USELESS_TYPE_CONVERSION uses this predicate.  */
+  return tree_ssa_useless_type_conversion_1 (a, b);
 }
 
 /* Check for some cases of the front end missing cast expressions.


However I'm not sure it is not papering over something else.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ebotcazou at gcc dot gnu dot
                   |                            |org
          Component|ada                         |middle-end
 GCC target triplet|powerpc-darwin, powerpc-    |powerpc-*, sparc-*
                   |linux                       |
            Summary|[4.0 Regression] Ada front- |[4.0 Regression] Ada
                   |end causing bootstrap       |bootstrap failure
                   |failure                     |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17793

[Bug middle-end/17793] [4.0 Regression] Ada bootstrap failure

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-10-15 07:30 -------
Investigating.


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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17793

[Bug ada/17793] [4.0 Regression] Ada bootstrap failure

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-16 03:09 -------
No this is an Ada front-end problem.
I think the main issue is that the tree-ssa expects that INTEGER types are close to the same and that 
pointers to them are compatible if the types are them selves compatible so maybe the Ada front-end 
should be producing types which are like that.  Really if types are compatiable so should be the 
pointers to them. 

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|ebotcazou at gcc dot gnu dot|unassigned at gcc dot gnu
                   |org                         |dot org
             Status|ASSIGNED                    |NEW
          Component|middle-end                  |ada


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17793

[Bug middle-end/17793] [4.0 Regression] Ada bootstrap failure

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-10-16 09:02 -------
> No this is an Ada front-end problem.

Did you investigate?

> I think the main issue is that the tree-ssa expects that INTEGER types are
> close to the same and that  pointers to them are compatible if the types are
> themselves compatible so maybe the Ada front-end should be producing types
> which are like that.  Really if types are compatiable so should be the 
> pointers to them. 

Did you read what I wrote?  The types are not compatible so the Ada front-end
correctly puts a NOP_EXPR between them.  But the gimplifier decides to strip it
and later thinks it should not have done so.

And, please, don't unassign PRs without discussing with the assignee beforehand.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |ebotcazou at gcc dot gnu dot
                   |dot org                     |org
             Status|NEW                         |ASSIGNED
          Component|ada                         |middle-end


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17793

[Bug middle-end/17793] [4.0 Regression] Ada bootstrap failure

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-16 13:55 -------
check_pointer_types_r is wrong but so is Ada assuming that integer types are not compatible (read the 
full bug report which I reported in the first place):
"The integer_type's are not compatible by defined by the front-end, why?"

See this is where Ada becomes wrong with respect to generic and gimple where it needs integer types 
which have the same PRECISION, UNSIGNEDness, and SIZE are consided compatible types (so are their
POINTERs).

For Kenner here, Generic and Gimple have the same type system which is partly described above.
Yes this is not documented but should be.  This again is an Ada bug not doing what the middle-end 
generic and gimple expects to happen.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17793

[Bug middle-end/17793] [4.0 Regression] Ada bootstrap failure

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-10-16 17:26 -------
[Thanks for elaborating on your position.]

> check_pointer_types_r is wrong but so is Ada assuming that integer types are
> not compatible (read the full bug report which I reported in the first place):
> "The integer_type's are not compatible by defined by the front-end, why?"

My understanding is that type compatibility is purely a front-end business, so
there is not really a notion of being right or wrong here.  The middle-end
should be able to cope with either situation.

> See this is where Ada becomes wrong with respect to generic and gimple where
> it needs integer types which have the same PRECISION, UNSIGNEDness, and SIZE
> are consided compatible types (so are their POINTERs).

Could you point me at where this requirement is documented?

The Ada front-end contains this comment (misc.c:560):

   ??? We may also want to generalize to considering lots of integer types
   compatible, but we need to understand the effects of alias sets first.

> For Kenner here, Generic and Gimple have the same type system which is partly
> described above. Yes this is not documented but should be.  This again is an
> Ada bug not doing what the middle-end generic and gimple expects to happen.

I think the issue is orthogonal to the discussion on the type systems of GENERIC
and GIMPLE.  The front-end generates a correct GENERIC tree AFAICS so the
gimplifier should be able to deal with it.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17793

[Bug middle-end/17793] [4.0 Regression] Ada bootstrap failure

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-10-23 09:16 -------
Subject: Bug 17793

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	ebotcazou@gcc.gnu.org	2004-10-23 09:16:47

Modified files:
	gcc            : ChangeLog gimplify.c 

Log message:
	PR middle-end/17793
	* gimplify.c (gimplify_addr_expr) <VIEW_CONVERT_EXPR>: Look
	through the operand if it is a useless type conversion.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.5994&r2=2.5995
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/gimplify.c.diff?cvsroot=gcc&r1=2.83&r2=2.84



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17793

[Bug middle-end/17793] [4.0 Regression] Ada bootstrap failure

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-10-23 09:19 -------
Initial patch: http://gcc.gnu.org/ml/gcc-patches/2004-10/msg01632.html
Final version: http://gcc.gnu.org/ml/gcc-patches/2004-10/msg01887.html


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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17793