Re: optimization/9123: [3.2/3.3/3.4 regression ] Internal compiler error in do_SUBST at combine.c:434

Re: optimization/9123: [3.2/3.3/3.4 regression ] Internal compiler error in do_SUBST at combine.c:434

[ebotcazou]

Synopsis: [3.2/3.3/3.4 regression ] Internal compiler error in do_SUBST at combine.c:434

State-Changed-From-To: analyzed->closed
State-Changed-By: ebotcazou
State-Changed-When: Fri Mar 28 21:33:57 2003
State-Changed-Why:
    Fixed.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=9123

Re: optimization/9123: [3.2/3.3/3.4 regression ] Internal compiler error in do_SUBST at combine.c:434

[Eric Botcazou]

The following reply was made to PR optimization/9123; it has been noted by GNATS.

From: Eric Botcazou <ebotcazou@libertysurf.fr>
To: Jason Merrill <jason@redhat.com>
Cc: gcc-bugs@gcc.gnu.org,
 khindenburg@cherrynebula.net,
 nobody@gcc.gnu.org,
 gcc-gnats@gcc.gnu.org
Subject: Re: optimization/9123: [3.2/3.3/3.4 regression ] Internal compiler error in do_SUBST at combine.c:434
Date: Fri, 28 Mar 2003 20:54:18 +0100

 > Hmm, I seem to have attached the wrong C testcase.  To fail, it needs
 > consts.  It fails with 3.2 and 3.3, but seems to be fixed on the trunk.
 >
 > typedef unsigned char T;
 > inline T clamp (const T in, const T min, const T max)
 > {
 >   return in <= min ? min : in >= max ? max : in;
 > }
 > unsigned char f (int i)
 > {
 >   return clamp (i, 0, 255);
 > }
 
 I can't reproduce on any branch on i586-redhat-linux-gnu. I strongly suspect 
 that I fixed it everywhere two weeks ago with:
 
 2003-03-14  Eric Botcazou  <ebotcazou@libertysurf.fr>
 
 	PR optimization/8396
 	* tree-inline.c (initialize_inlined_parameters): Make sure the value
 	of read-only constant arguments is passed with the right type.
 
 Are your 3.2 and 3.3 compilers up-to-date?
 
 -- 
 Eric Botcazou

Re: optimization/9123: [3.2/3.3/3.4 regression ] Internal compiler error in do_SUBST at combine.c:434

[Jason Merrill]

The following reply was made to PR optimization/9123; it has been noted by GNATS.

From: Jason Merrill <jason@redhat.com>
To: Eric Botcazou <ebotcazou@libertysurf.fr>
Cc: gcc-bugs@gcc.gnu.org, khindenburg@cherrynebula.net, nobody@gcc.gnu.org,
   gcc-gnats@gcc.gnu.org
Subject: Re: optimization/9123: [3.2/3.3/3.4 regression ] Internal compiler
 error in do_SUBST at combine.c:434
Date: Fri, 28 Mar 2003 11:57:02 -0500

 On Fri, 28 Mar 2003 11:17:31 +0100, Eric Botcazou <ebotcazou@libertysurf.fr> wrote:
 
 > On which versions and with which options does this C testcase fail?
 
 Hmm, I seem to have attached the wrong C testcase.  To fail, it needs
 consts.  It fails with 3.2 and 3.3, but seems to be fixed on the trunk.
 
 typedef unsigned char T;
 inline T clamp (const T in, const T min, const T max)
 {
   return in <= min ? min : in >= max ? max : in;
 }
 unsigned char f (int i)
 {
   return clamp (i, 0, 255);
 }

Re: optimization/9123: [3.2/3.3/3.4 regression ] Internal compiler error in do_SUBST at combine.c:434

[Eric Botcazou]

The following reply was made to PR optimization/9123; it has been noted by GNATS.

From: Eric Botcazou <ebotcazou@libertysurf.fr>
To: jason@gcc.gnu.org
Cc: gcc-bugs@gcc.gnu.org,
 khindenburg@cherrynebula.net,
 nobody@gcc.gnu.org,
 gcc-gnats@gcc.gnu.org
Subject: Re: optimization/9123: [3.2/3.3/3.4 regression ] Internal compiler error in do_SUBST at combine.c:434
Date: Fri, 28 Mar 2003 11:17:31 +0100

 >     Not a c++-specific bug; it's dying because it's trying to substitute
 > (unsigned char)-1 for a QI reg, but combine doesn't like that because
 > (HOST_WIDE_INT)255 != (HOST_WIDE_INT)-1.  Here's a C testcase:
 >
 >     typedef unsigned char T;
 >     inline T clamp (T in, T min, T max)
 >     {
 >       return in <= min ? min : in >= max ? max : in;
 >     }
 >     unsigned char f (int i)
 >     {
 >       return clamp (i, 0, 255);
 >     }
 
 On which versions and with which options does this C testcase fail?
 
 -- 
 Eric Botcazou

Re: optimization/9123: [3.2/3.3/3.4 regression ] Internal compiler error in do_SUBST at combine.c:434

[Janis Johnson]

The following reply was made to PR optimization/9123; it has been noted by GNATS.

From: Janis Johnson <janis187@us.ibm.com>
To: gcc-gnats@gcc.gnu.org, gcc-bugs@gcc.gnu.org, khindenburg@cherrynebula.net,
   nobody@gcc.gnu.org, gcc-prs@gcc.gnu.org, aoliva@redhat.com
Cc:  
Subject: Re: optimization/9123: [3.2/3.3/3.4 regression ] Internal compiler error 
 in do_SUBST at combine.c:434
Date: Sat, 22 Mar 2003 13:42:33 -0800

 The regression appeared with this patch:
 
 2001-04-12  Alexandre Oliva  <aoliva@redhat.com>
 
         * recog.c (general_operand, immediate_operand,
         nonmemory_operand): Require CONST_INTs to be sign-extended
         values for their modes.
 
 The search used the smaller C++ test case.
 
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=9123
 
 
 

Re: optimization/9123: [3.2/3.3/3.4 regression ] Internal compiler error in do_SUBST at combine.c:434

[jason]

Synopsis: [3.2/3.3/3.4 regression ] Internal compiler error in do_SUBST at combine.c:434

State-Changed-From-To: open->analyzed
State-Changed-By: jason
State-Changed-When: Mon Mar 10 22:38:35 2003
State-Changed-Why:
    analyzed->analyzed

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=9123

Re: optimization/9123: [3.2/3.3/3.4 regression ] Internal compiler error in do_SUBST at combine.c:434

[jason]

Synopsis: [3.2/3.3/3.4 regression ] Internal compiler error in do_SUBST at combine.c:434

State-Changed-From-To: analyzed->open
State-Changed-By: jason
State-Changed-When: Mon Mar 10 22:37:43 2003
State-Changed-Why:
    Not a c++-specific bug; it's dying because it's trying to substitute (unsigned char)-1 for a QI reg, but combine doesn't like that because (HOST_WIDE_INT)255 != (HOST_WIDE_INT)-1.  Here's a C testcase:
    
    typedef unsigned char T;
    inline T clamp (T in, T min, T max)
    {
      return in <= min ? min : in >= max ? max : in;
    }
    unsigned char f (int i)
    {
      return clamp (i, 0, 255);
    }

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=9123