Re: bootstrap/7143: real.c: In function `make_nan': `TFbignan' undeclared

Re: bootstrap/7143: real.c: In function `make_nan': `TFbignan' undeclared

[John David Anglin]

The following reply was made to PR bootstrap/7143; it has been noted by GNATS.

From: "John David Anglin" <dave@hiauly1.hia.nrc.ca>
To: neil@gcc.gnu.org, dave@hiauly1.hia.nrc.ca, gcc-bugs@gcc.gnu.org,
   gcc-prs@gcc.gnu.org, neil@gcc.gnu.org, gcc-gnats@gcc.gnu.org
Cc: gcc-bugs@gcc.gnu.org, gcc-prs@gcc.gnu.org, neil@gcc.gnu.org
Subject: Re: bootstrap/7143: real.c: In function `make_nan': `TFbignan' undeclared
Date: Sat, 9 Nov 2002 11:06:31 -0500 (EST)

 > Synopsis: real.c: In function `make_nan': `TFbignan' undeclared
 > 
 > State-Changed-From-To: open->feedback
 > State-Changed-By: neil
 > State-Changed-When: Sat Nov  9 04:17:59 2002
 > State-Changed-Why:
 >     Can I close this?  I find it really hard to believe a CPP bug is biting in this case.
 > 
 > http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=7143
 
 Go ahead.
 
 Dave
 -- 
 J. David Anglin                                  dave.anglin@nrc.ca
 National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)

Re: bootstrap/7143: real.c: In function `make_nan': `TFbignan' undeclared

[neil]

Synopsis: real.c: In function `make_nan': `TFbignan' undeclared

State-Changed-From-To: feedback->closed
State-Changed-By: neil
State-Changed-When: Sat Nov  9 08:08:47 2002
State-Changed-Why:
    Submitter agreed to close, presumably can no longer reproduce.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=7143

Re: bootstrap/7143: real.c: In function `make_nan': `TFbignan' undeclared

[neil]

Synopsis: real.c: In function `make_nan': `TFbignan' undeclared

State-Changed-From-To: open->feedback
State-Changed-By: neil
State-Changed-When: Sat Nov  9 04:17:59 2002
State-Changed-Why:
    Can I close this?  I find it really hard to believe a CPP bug is biting in this case.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=7143

Re: bootstrap/7143: real.c: In function `make_nan': `TFbignan' undeclared

[John David Anglin]

The following reply was made to PR bootstrap/7143; it has been noted by GNATS.

From: "John David Anglin" <dave@hiauly1.hia.nrc.ca>
To: neil@daikokuya.co.uk (Neil Booth)
Cc: gcc-gnats@gcc.gnu.org, gcc-bugs@gcc.gnu.org
Subject: Re: bootstrap/7143: real.c: In function `make_nan': `TFbignan' undeclared
Date: Mon, 1 Jul 2002 15:41:28 -0400 (EDT)

 > > I'm going to retry the bootstrap using the native compiler.  If that
 > > fails, I'll try to debug the problem.
 > 
 > What's the situation here now?
 
 I ran into a few problems with the above.  I am currently bootstrapping
 3.1.1.  I'm far enough to know that the preprocessing problem isn't
 present in 3.1.1.  When 3.1.1 completes, I will retry 3.2.  This will
 take awhile (sometime next week).
 
 I also built a complete cross compile environment with 3.2 under
 hppa1.1-hp-hpux10.20.  It also doesn't have the problem, so I suspect
 that my 3.0 version of gcc on the vax has a bug which results in
 the preprocessor being miscompiled.
 
 Dave
 -- 
 J. David Anglin                                  dave.anglin@nrc.ca
 National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)

Re: bootstrap/7143: real.c: In function `make_nan': `TFbignan' undeclared

[Neil Booth]

The following reply was made to PR bootstrap/7143; it has been noted by GNATS.

From: Neil Booth <neil@daikokuya.co.uk>
To: John David Anglin <dave@hiauly1.hia.nrc.ca>
Cc: gcc-gnats@gcc.gnu.org, gcc-bugs@gcc.gnu.org
Subject: Re: bootstrap/7143: real.c: In function `make_nan': `TFbignan' undeclared
Date: Mon, 1 Jul 2002 17:01:46 +0100

 John David Anglin wrote:-
 
 > > I've got a clue.  You mentioned you were compiling with -pedantic.
 > 
 > The only new clue that I have is tradcpp0 correctly processed the
 > test file.  I can't duplicate the problem with a cross compiler.
 > 
 > > Is DEC a built-in?  If so, -pedantic won't define it because it's in
 > > the user's namespace.  If DEC is defined in a file, then it doesn't
 > > matter of course.
 > 
 > No, it isn't.  I changed DEC to DECXXX and it didn't affect the problem.
 > -pedantic also didn't affect the situation.
 > 
 > I'm going to retry the bootstrap using the native compiler.  If that
 > fails, I'll try to debug the problem.
 
 What's the situation here now?
 
 Neil.

Re: bootstrap/7143: real.c: In function `make_nan': `TFbignan' undeclared

[John David Anglin]

The following reply was made to PR bootstrap/7143; it has been noted by GNATS.

From: "John David Anglin" <dave@hiauly1.hia.nrc.ca>
To: neil@daikokuya.co.uk (Neil Booth)
Cc: gcc-gnats@gcc.gnu.org, gcc-bugs@gcc.gnu.org
Subject: Re: bootstrap/7143: real.c: In function `make_nan': `TFbignan' undeclared
Date: Fri, 28 Jun 2002 11:30:36 -0400 (EDT)

 > I've got a clue.  You mentioned you were compiling with -pedantic.
 
 The only new clue that I have is tradcpp0 correctly processed the
 test file.  I can't duplicate the problem with a cross compiler.
 
 > Is DEC a built-in?  If so, -pedantic won't define it because it's in
 > the user's namespace.  If DEC is defined in a file, then it doesn't
 > matter of course.
 
 No, it isn't.  I changed DEC to DECXXX and it didn't affect the problem.
 -pedantic also didn't affect the situation.
 
 I'm going to retry the bootstrap using the native compiler.  If that
 fails, I'll try to debug the problem.
 
 Dave
 -- 
 J. David Anglin                                  dave.anglin@nrc.ca
 National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)

Re: bootstrap/7143: real.c: In function `make_nan': `TFbignan' undeclared

[Neil Booth]

The following reply was made to PR bootstrap/7143; it has been noted by GNATS.

From: Neil Booth <neil@daikokuya.co.uk>
To: John David Anglin <dave@hiauly1.hia.nrc.ca>
Cc: gcc-gnats@gcc.gnu.org, gcc-bugs@gcc.gnu.org
Subject: Re: bootstrap/7143: real.c: In function `make_nan': `TFbignan' undeclared
Date: Fri, 28 Jun 2002 07:20:43 +0100

 I've got a clue.  You mentioned you were compiling with -pedantic.
 
 Is DEC a built-in?  If so, -pedantic won't define it because it's in
 the user's namespace.  If DEC is defined in a file, then it doesn't
 matter of course.
 
 We didn't used to get this stuff right, but if your target has been
 converted to the new CPP builtins recently then we've suddenly
 started getting it right and it's prevented your bootstrap.
 
 Maybe real.c should not be -pedantic, or should check for __DEC__
 too (assuming that's defined).
 
 Neil.

Re: bootstrap/7143: real.c: In function `make_nan': `TFbignan' undeclared

[neil]

Synopsis: real.c: In function `make_nan': `TFbignan' undeclared

Responsible-Changed-From-To: unassigned->neil
Responsible-Changed-By: neil
Responsible-Changed-When: Thu Jun 27 23:19:59 2002
Responsible-Changed-Why:
    Mine, I guess, though I can't reproduce it as a CPP issue.
    Maybe the stage1 compiler for your target is somehow miscompiling the preprocessor?

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=7143

Re: bootstrap/7143: real.c: In function `make_nan': `TFbignan' undeclared

[Neil Booth]

The following reply was made to PR bootstrap/7143; it has been noted by GNATS.

From: Neil Booth <neil@daikokuya.co.uk>
To: John David Anglin <dave@hiauly1.hia.nrc.ca>
Cc: gcc-gnats@gcc.gnu.org, gcc-bugs@gcc.gnu.org
Subject: Re: bootstrap/7143: real.c: In function `make_nan': `TFbignan' undeclared
Date: Fri, 28 Jun 2002 07:11:40 +0100

 John David Anglin wrote:-
 
 > The code below reproduces the problem.  The problem has something
 > to do with the processing of '!' and defined.  If I remove '!', then
 > defining/undefing DEC yields the correct result.
 
 OK, thanks.  I can't reproduce it, though, with the compiler that
 contains my most recent changes (2 days ago now).  See below.  The
  file bug.c is exactly as contained in your mail.
 
 > It's likely that something has been miscompiled in stage1 but the same
 > compiler bootstrapped 3.1.  It's amazing that the build of the C compiler
 > nearly completed in stage2 before a problem showed up.
 [..]
 > #define DEC 1
 > 
 > #if !defined(DEC)
 > static void enan (void);
 > #endif
 
 neil@cat:~/src/gcc/build/gcc$ ./cc1 -E /tmp/bug.c -quiet
 # 1 "/tmp/bug.c"
 # 1 "<built-in>"
 # 1 "<command line>"
 # 1 "/tmp/bug.c"
 neil@cat:~/src/gcc/build/gcc$ ./cc1 -E -traditional-cpp /tmp/bug.c
 -quiet
 # 1 "/tmp/bug.c"
 # 1 "<built-in>"
 # 1 "<command line>"
 # 1 "/tmp/bug.c"
 
 
 
 
 
 
 neil@cat:~/src/gcc/build/gcc$ ./cc1 --version
 GNU C version 3.2 20020626 (experimental) (i686-pc-linux-gnu)
 	compiled by GNU C version 3.2 20020626 (experimental).
 neil@cat:~/src/gcc/build/gcc$ 
 
 So I'm not really sure what to do.  Actually, this compiler contains
 a local change that I'm about to commit, but it only affects the
 handling of #include by the traditional preprocessor.
 
 Neil.

bootstrap/7143: real.c: In function `make_nan': `TFbignan' undeclared

[dave]

>Number:         7143
>Category:       bootstrap
>Synopsis:       real.c: In function `make_nan': `TFbignan' undeclared
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Thu Jun 27 11:36:01 PDT 2002
>Closed-Date:
>Last-Modified:
>Originator:     Dave Anglin
>Release:        gcc 3.2 20020623 (experimental)
>Organization:
>Environment:
vax-dec-ultrix4.3
>Description:
The following error occurs in stage2 of the bootstrap:

stage1/xgcc -Bstage1/ -B/usr/local/vax-dec-ultrix4.3/bin/ -c -DIN_GCC    -g -O2 
-W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -Wtraditional 
-pedantic -Wno-long-long  -DHAVE_CONFIG_H    -I. -I. -I../../gcc/gcc -I../../gcc
/gcc/. -I../../gcc/gcc/config -I../../gcc/gcc/../include ../../gcc/gcc/real.c -o
 real.o
../../gcc/gcc/real.c: In function `make_nan':
../../gcc/gcc/real.c:6297: error: `TFbignan' undeclared (first use in this funct
ion)
../../gcc/gcc/real.c:6297: error: (Each undeclared identifier is reported only o
nce
../../gcc/gcc/real.c:6297: error: for each function it appears in.)
../../gcc/gcc/real.c:6299: error: `TFlittlenan' undeclared (first use in this fu
nction)
../../gcc/gcc/real.c:6307: error: `XFbignan' undeclared (first use in this funct
ion)
../../gcc/gcc/real.c:6309: error: `XFlittlenan' undeclared (first use in this fu
nction)
../../gcc/gcc/real.c:6315: error: `DFbignan' undeclared (first use in this funct
ion)
../../gcc/gcc/real.c:6317: error: `DFlittlenan' undeclared (first use in this fu
nction)
../../gcc/gcc/real.c:6324: error: `SFbignan' undeclared (first use in this funct
ion)
../../gcc/gcc/real.c:6326: error: `SFlittlenan' undeclared (first use in this fu
nction)
../../gcc/gcc/real.c: At top level:
../../gcc/gcc/real.c:1886: warning: `eiisinf' defined but not used

The puzzle is why this code is being compiled.  We have:

#define VAX_FLOAT_FORMAT 2
#define TARGET_FLOAT_FORMAT VAX_FLOAT_FORMAT
#define DEC 1

The code in question is inside

#if !defined(DEC) && !defined(IBM) && !defined(C4X)
...
#endif

It sure looks to me as if DEC is defined.  In stage1,
real.c compiles ok.
>How-To-Repeat:

>Fix:

>Release-Note:
>Audit-Trail:
>Unformatted: