[Bug ada/56178] New: [4.8 Regression] Conversion of Long_Float to Integer triggers CONSTRAINT_ERROR overflow check failed

[Bug ada/56178] New: [4.8 Regression] Conversion of Long_Float to Integer triggers CONSTRAINT_ERROR overflow check failed

[orion at cora dot nwra.com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56178

             Bug #: 56178
           Summary: [4.8 Regression] Conversion of Long_Float to Integer
                    triggers CONSTRAINT_ERROR overflow check failed
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Severity: major
          Priority: P3
         Component: ada
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: orion@cora.nwra.com


Description of problem:

$ cat test_overflow.adb
procedure Test_Overflow is
    r : Integer;
    r1 : Long_Float := 0.3;
begin
    r := Integer(r1);
end Test_Overflow;
$ gnatmake test_overflow.adb 
gcc -c test_overflow.adb
gnatbind -x test_overflow.ali
gnatlink test_overflow.ali
$ ./test_overflow 

raised CONSTRAINT_ERROR : test_overflow.adb:5 overflow check failed

gcc-4.8.0-0.7.fc19.x86_64

[Bug ada/56178] [4.8 Regression] Conversion of Long_Float to Integer triggers CONSTRAINT_ERROR overflow check failed

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56178

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2013-02-04
                 CC|                            |jakub at gcc dot gnu.org
     Ever Confirmed|0                           |1

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> 2013-02-04 07:32:46 UTC ---
This only seems to occur in profiledbootstrap, e.g. x86_64-linux
--enable-checking=release profiledbootstrap, stagefeedback gnat1 fails, while
stageprofile and stage1 gnat1 work (and in normal bootstrap all of stage[123]
work).  The issue is somewhere in the Ada FE, already in the *.original dump
the difference is between if (0) vs. if (1) for the overflow check call.
Compared to normal bootstrap (also --enable-checking=yes,rtl) with
profiledbootstrap (--enable-checking=release) also shows:
+FAIL:  c34002a
+FAIL:  c460007
+FAIL:  c490001
+FAIL:  cc1221d
+FAIL:  cxa5a04
+FAIL:  cxg2001
+FAIL:  cxg2015
regressions, not sure if that is in any way relased to this issue.

[Bug ada/56178] [4.8 Regression] Conversion of Long_Float to Integer triggers CONSTRAINT_ERROR overflow check failed

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56178

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> 2013-02-04 09:15:49 UTC ---
bisected to ada/urealp.o, if this comes from -fprofile-use build, even if all
other ada/*.o come from stage1, it still fails, if ada/urealp.o is from stage1,
even if all other ada/*.o come from -fprofile-use build, it works fine.
Not familiar enough with Ada to know where to look at though.

[Bug ada/56178] [4.8 Regression] Conversion of Long_Float to Integer triggers CONSTRAINT_ERROR overflow check failed

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56178

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.8.0

[Bug ada/56178] [4.8 Regression] Conversion of Long_Float to Integer triggers CONSTRAINT_ERROR overflow check failed

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56178

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

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

--- Comment #3 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2013-02-04 10:42:15 UTC ---
Thanks for the analysis, Jakub, I'll take it from there.

[Bug ada/56178] [4.8 Regression] Miscompilation of Ada front-end with profiled bootstrap

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56178

--- Comment #4 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2013-02-05 21:41:30 UTC ---
ureal.adb:UR_Add is miscompiled with profile feedback and this is yet another
instance of the infamous webizer bug:
  http://gcc.gnu.org/ml/gcc-patches/2012-10/msg01259.html
but the origin seems to be in fwprop1 this time...

[Bug rtl-optimization/56178] [4.8 Regression] Miscompilation of Ada front-end with profiled bootstrap

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56178

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|ada                         |rtl-optimization

--- Comment #5 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2013-02-06 17:24:45 UTC ---
I confirm that fixing the ureal.adb miscompilation fixes the ACATS failures.

[Bug rtl-optimization/56178] [4.8 Regression] Miscompilation of Ada front-end with profiled bootstrap

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56178

--- Comment #6 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2013-02-07 08:07:19 UTC ---
Author: ebotcazou
Date: Thu Feb  7 08:07:04 2013
New Revision: 195841

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=195841
Log:
    PR rtl-optimization/56178
    * cse.c (cse_insn): Do not create a REG_EQUAL note if the source is a
    SUBREG of a register.  Tidy up related block of code.
    * fwprop.c (forward_propagate_and_simplify): Do not create a REG_EQUAL
    note if the source is a register or a SUBREG of a register.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/cse.c
    trunk/gcc/fwprop.c

[Bug rtl-optimization/56178] [4.8 Regression] Miscompilation of Ada front-end with profiled bootstrap

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56178

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

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

--- Comment #7 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2013-02-07 08:27:55 UTC ---
Thanks for reporting the problem.