Re: PATCH: Re: other/9031

Re: PATCH: Re: other/9031

[Alexandre Oliva]

The following reply was made to PR other/9031; it has been noted by GNATS.

From: Alexandre Oliva <aoliva@redhat.com>
To: John David Anglin <dave@hiauly1.hia.nrc.ca>
Cc: gcc-gnats@gcc.gnu.org, nobody@gcc.gnu.org, dave.anglin@nrc.ca,
   gcc-patches@gcc.gnu.org, binutils@sources.redhat.com, neroden@twcny.rr.com
Subject: Re: PATCH: Re: other/9031
Date: 28 Dec 2002 07:15:53 -0200

 On Dec 28, 2002, Alexandre Oliva <aoliva@redhat.com> wrote:
 
 >> 2002-12-28  John David Anglin  <dave.anglin@nrc.ca>
 
 >> * configure (LD): Improve test for gcc and try to set LD to the ld
 >> used by gcc when using gcc and not doing a Canadian Cross.
 
 > Ok, but I wouldn't mind if you waited for Nathan's patch to go in
 > before adjusting yours.  If you agree, consider the updated patch
 > pre-approved, but please post it, for the record.
 
 In fact, consider now sinclude()ing libtool.m4 and using its
 AC_PROG_LD macro.
 
 -- 
 Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
 Red Hat GCC Developer                 aoliva@{redhat.com, gcc.gnu.org}
 CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
 Free Software Evangelist                Professional serial bug killer

Re: PATCH: Re: other/9031

[John David Anglin]

The following reply was made to PR other/9031; it has been noted by GNATS.

From: "John David Anglin" <dave@hiauly1.hia.nrc.ca>
To: aoliva@redhat.com (Alexandre Oliva)
Cc: gcc-gnats@gcc.gnu.org, nobody@gcc.gnu.org, dave.anglin@nrc.ca,
   gcc-patches@gcc.gnu.org, binutils@sources.redhat.com, neroden@twcny.rr.com
Subject: Re: PATCH: Re: other/9031
Date: Sat, 4 Jan 2003 15:58:55 -0500 (EST)

 > > Ok, but I wouldn't mind if you waited for Nathan's patch to go in
 > > before adjusting yours.  If you agree, consider the updated patch
 > > pre-approved, but please post it, for the record.
 > 
 > In fact, consider now sinclude()ing libtool.m4 and using its
 > AC_PROG_LD macro.
 
 I have installed the following revised version of the patch on the main.
 The code now checks to see if LD is set.  There are some minor variable
 name changes and other revisions due to the fact that the code is now
 processed by autoconf.
 
 Tested on hppa64-hp-hpux11.11 and hppa2.0w-hp-hpux11.11.
 
 Dave
 -- 
 J. David Anglin                                  dave.anglin@nrc.ca
 National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)
 
 2003-01-04 John David Anglin  <dave.anglin@nrc.ca>
 
 	* configure.in (LD): Improve test for gcc.  Try to set LD to the ld used
 	by gcc if LD is not defined and we are not doing a Canadian Cross.
 	* configure: Rebuilt.
 
 Index: configure.in
 ===================================================================
 RCS file: /cvsroot/gcc/gcc/configure.in,v
 retrieving revision 1.202
 diff -u -3 -p -r1.202 configure.in
 --- configure.in	28 Dec 2002 17:57:14 -0000	1.202
 +++ configure.in	29 Dec 2002 19:01:47 -0000
 @@ -1456,12 +1456,14 @@ else
    YACC="\$(USUAL_YACC)"
  
    # If CC is still not set, try to get gcc.
 +  cc_prog_is_gcc=
    if test -z "${CC}" ; then
      IFS="${IFS= 	}"; save_ifs="$IFS"; IFS="${IFS}:"
      for dir in $PATH; do
        test -z "$dir" && dir=.
        if test -f $dir/gcc; then
  	CC="gcc"
 +	cc_prog_is_gcc=yes
  	echo 'void f(){}' > conftest.c
  	if test -z "`${CC} -g -c conftest.c 2>&1`"; then
  	  CFLAGS=${CFLAGS-"-g -O2"}
 @@ -1477,11 +1479,20 @@ else
      IFS="$save_ifs"
      CC=${CC-cc}
    else
 +    # Determine if we are using gcc.
 +    cat > conftest.c <<EOF
 +#ifdef __GNUC__
 +  yes;
 +#endif
 +EOF
 +    if ${CC} -E conftest.c | grep yes >/dev/null 2>&1; then
 +      cc_prog_is_gcc=yes
 +    fi
 +    rm -f conftest.c
      if test -z "${CFLAGS}"; then
        # Here CC is set but CFLAGS is not.  Use a quick hack to use -O2 if CC
        # is set to a version of gcc.
 -      case "${CC}" in
 -      *gcc)
 +      if test "$cc_prog_is_gcc" = yes; then
  	echo 'void f(){}' > conftest.c
  	if test -z "`${CC} -g -c conftest.c 2>&1`"; then
  	  CFLAGS=${CFLAGS-"-g -O2"}
 @@ -1491,7 +1502,25 @@ else
  	  CXXFLAGS=${CXXFLAGS-"-O2"}
  	fi
  	rm -f conftest*
 -	;;
 +      fi
 +    fi
 +  fi
 +
 +  # We must set the default linker to the linker used by gcc for the correct
 +  # operation of libtool.  If LD is not defined and we are using gcc, try to
 +  # set the LD default to the ld used by gcc.
 +  if test -z "$LD"; then
 +    if test "$cc_prog_is_gcc" = yes; then
 +      case $build in
 +      *-*-mingw*)
 +	gcc_prog_ld=`$CC -print-prog-name=ld 2>&1 | tr -d '\015'` ;;
 +      *)
 +	gcc_prog_ld=`$CC -print-prog-name=ld 2>&1` ;;
 +      esac
 +      case $gcc_prog_ld in
 +      # Accept absolute paths.
 +      [[\\/]* | [A-Za-z]:[\\/]*)]
 +        LD="$gcc_prog_ld" ;;
        esac
      fi
    fi

Re: PATCH: Re: other/9031

[Alexandre Oliva]

The following reply was made to PR other/9031; it has been noted by GNATS.

From: Alexandre Oliva <aoliva@redhat.com>
To: John David Anglin <dave@hiauly1.hia.nrc.ca>
Cc: gcc-gnats@gcc.gnu.org, nobody@gcc.gnu.org, dave.anglin@nrc.ca,
   gcc-patches@gcc.gnu.org, binutils@sources.redhat.com, neroden@twcny.rr.com
Subject: Re: PATCH: Re: other/9031
Date: 28 Dec 2002 04:31:35 -0200

 On Dec 28, 2002, John David Anglin <dave@hiauly1.hia.nrc.ca> wrote:
 
 > Here is a revised patch.
 
 Heh.  Thanks.  Just as I was about to approve the original patch :-)
 
 > 2002-12-28  John David Anglin  <dave.anglin@nrc.ca>
 
 > 	* configure (LD): Improve test for gcc and try to set LD to the ld
 > 	used by gcc when using gcc and not doing a Canadian Cross.
 
 Ok, but I wouldn't mind if you waited for Nathan's patch to go in
 before adjusting yours.  If you agree, consider the updated patch
 pre-approved, but please post it, for the record.
 
 Thanks!
 
 -- 
 Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
 Red Hat GCC Developer                 aoliva@{redhat.com, gcc.gnu.org}
 CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
 Free Software Evangelist                Professional serial bug killer

Re: PATCH: Re: other/9031

[John David Anglin]

The following reply was made to PR other/9031; it has been noted by GNATS.

From: John David Anglin <dave@hiauly1.hia.nrc.ca>
To: Alexandre Oliva <aoliva@redhat.com>
Cc: gcc-gnats@gcc.gnu.org, nobody@gcc.gnu.org, dave.anglin@nrc.ca,
   gcc-patches@gcc.gnu.org, binutils@sources.redhat.com, neroden@twcny.rr.com
Subject: Re: PATCH: Re: other/9031
Date: Sat, 28 Dec 2002 01:22:25 -0500

 Here is a revised patch.  I have improved the gcc detection, removed the
 extra double quotes from the commands used to print the ld used by gcc,
 and the path canonicalization.  The mingw CR removal is still in as I
 have no way to determine whether this isn't still needed.
 
 I have tested the patch with builds on hppa64-hp-hpux11.11 and
 hppa2.0w-hp-hpux11.00.
 
 The patch will be obsolete if Nanthanael's autoconfiscation patch is
 installed, although I see his patch still includes the code that I
 am attempting to improve.
 
 Dave
 
 2002-12-28  John David Anglin  <dave.anglin@nrc.ca>
 
 	* configure (LD): Improve test for gcc and try to set LD to the ld
 	used by gcc when using gcc and not doing a Canadian Cross.
 
 Index: configure
 ===================================================================
 RCS file: /cvsroot/gcc/gcc/configure,v
 retrieving revision 1.51
 diff -u -3 -p -r1.51 configure
 --- configure	22 Dec 2002 06:46:41 -0000	1.51
 +++ configure	27 Dec 2002 18:51:43 -0000
 @@ -1020,6 +1020,7 @@ else
    YACC='\$(USUAL_YACC)'
  
    # If CC is still not set, try to get gcc.
 +  cc_prog=
    if [ -z "${CC}" ]; then
      IFS="${IFS= 	}"; save_ifs="$IFS"; IFS="${IFS}:"
      for dir in $PATH; do
 @@ -1028,6 +1029,7 @@ else
  	CC="gcc"
  	echo 'void f(){}' > conftest.c
  	if test -z "`${CC} -g -c conftest.c 2>&1`"; then
 +	  cc_prog=gcc
  	  CFLAGS=${CFLAGS-"-g -O2"}
  	  CXXFLAGS=${CXXFLAGS-"-g -O2"}
  	else
 @@ -1044,8 +1046,16 @@ else
      if test -z "${CFLAGS}"; then
        # Here CC is set but CFLAGS is not.  Use a quick hack to use -O2 if CC
        # is set to a version of gcc.
 -      case "${CC}" in
 -      *gcc)
 +      cat > conftest.c <<EOF
 +#ifdef __GNUC__
 +  yes;
 +#endif
 +EOF
 +      if ${CC} -E conftest.c | grep yes >/dev/null 2>&1; then
 +	cc_prog=gcc
 +      fi
 +      rm -f conftest.c
 +      if [ "$cc_prog" = gcc ]; then
  	echo 'void f(){}' > conftest.c
  	if test -z "`${CC} -g -c conftest.c 2>&1`"; then
  	  CFLAGS=${CFLAGS-"-g -O2"}
 @@ -1055,9 +1065,23 @@ else
  	  CXXFLAGS=${CXXFLAGS-"-O2"}
  	fi
  	rm -f conftest*
 -	;;
 -      esac
 +      fi
      fi
 +  fi
 +
 +  # If we are using gcc, try to set the LD default to the ld used by gcc.
 +  if [ "$cc_prog" = gcc ]; then
 +    case $build in
 +    *-*-mingw*)
 +      gcc_prog_ld=`$CC -print-prog-name=ld 2>&1 | tr -d '\015'` ;;
 +    *)
 +      gcc_prog_ld=`$CC -print-prog-name=ld 2>&1` ;;
 +    esac
 +    case $gcc_prog_ld in
 +    # Accept absolute paths.
 +    [\\/]* | [A-Za-z]:[\\/]*)
 +      LD="$gcc_prog_ld" ;;
 +    esac
    fi
  
    CXX=${CXX-"c++"}

Re: PATCH: Re: other/9031

[John David Anglin]

The following reply was made to PR other/9031; it has been noted by GNATS.

From: "John David Anglin" <dave@hiauly1.hia.nrc.ca>
To: aoliva@redhat.com (Alexandre Oliva)
Cc: gcc-gnats@gcc.gnu.org, nobody@gcc.gnu.org, dave.anglin@nrc.ca,
   gcc-patches@gcc.gnu.org, binutils@sources.redhat.com
Subject: Re: PATCH: Re: other/9031
Date: Thu, 26 Dec 2002 23:10:28 -0500 (EST)

 > host.  So perhaps it would be more reasonable to change case $host to
 > case $build.  Besides, I suppose the problem may actually be caused by
 > the unnecessary use of double quotes in the assignment.  I *think*
 > taking them out may cause the needless CR to go away without affecting
 > anything else.
 
 I don't think so.  The double quotes were mine and not in the original
 libtool macro.  The original comment regarding the CR was that gcc leaves
 a trailing CR which upsets mingw.  I am unable to test whether this is
 actually a problem.  The libtool patch to strip carriage returns was made
 by Gary Vaughan on 1999-08-03.  The predates the libtool archive.  I agree
 that $host should be $build.
 
 > Also, pathname canonicalization is tricky business.  Replacing
 > foo/../ with / is only a good idea when foo is not a soft link.  Do we
 > really need this?
 
 I did the pathname canonicalization because the libtool macro did it
 and I was trying the provide a default value for LD comparible to what
 libtool would generate.  It's unnecessary in the configurations that
 I use and would have no objection in removing it.
 
 Dave
 -- 
 J. David Anglin                                  dave.anglin@nrc.ca
 National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)

Re: PATCH: Re: other/9031

[Alexandre Oliva]

The following reply was made to PR other/9031; it has been noted by GNATS.

From: Alexandre Oliva <aoliva@redhat.com>
To: "John David Anglin" <dave@hiauly1.hia.nrc.ca>
Cc: gcc-gnats@gcc.gnu.org, nobody@gcc.gnu.org, dave.anglin@nrc.ca,
   gcc-patches@gcc.gnu.org, binutils@sources.redhat.com
Subject: Re: PATCH: Re: other/9031
Date: 27 Dec 2002 01:11:11 -0200

 On Dec 26, 2002, "John David Anglin" <dave@hiauly1.hia.nrc.ca> wrote:
 
 > The enclosed patch fixes the problem reported in other/9031.  The
 > technique for determining LD is derived from that used by the
 > libtool AC_PROG_LD macro.
 
 > +    case $host in
 > +    *-*-mingw*)
 > +      gcc_prog_ld="`$CC -print-prog-name=ld 2>&1 | tr -d '\015'`" ;;
 > +    *)
 > +      gcc_prog_ld="`$CC -print-prog-name=ld 2>&1`" ;;
 > +    esac
 
 I don't think this is appropriate.  If mingw's CRs are causing
 trouble, they'd cause trouble when mingw is the build machine, not the
 host.  So perhaps it would be more reasonable to change case $host to
 case $build.  Besides, I suppose the problem may actually be caused by
 the unnecessary use of double quotes in the assignment.  I *think*
 taking them out may cause the needless CR to go away without affecting
 anything else.
 
 Another issue is that CC's ending with gcc may not be a good test for
 whether the host compiler is CC.  I realize this is not the only
 occurrence of case $CC in *gcc) in the configure script, but even that
 one should IMHO be changed so as to use something more similar to
 autoconf's test for whether we are using GNU C.
 
 Also, pathname canonicalization is tricky business.  Replacing
 foo/../ with / is only a good idea when foo is not a soft link.  Do we
 really need this?
 
 -- 
 Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
 Red Hat GCC Developer                 aoliva@{redhat.com, gcc.gnu.org}
 CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
 Free Software Evangelist                Professional serial bug killer

PATCH: Re: other/9031

[John David Anglin]

The following reply was made to PR other/9031; it has been noted by GNATS.

From: "John David Anglin" <dave@hiauly1.hia.nrc.ca>
To: gcc-gnats@gcc.gnu.org, nobody@gcc.gnu.org
Cc: dave.anglin@nrc.ca, gcc-patches@gcc.gnu.org, binutils@sources.redhat.com
Subject: PATCH: Re: other/9031
Date: Thu, 26 Dec 2002 17:16:02 -0500 (EST)

 > Thank you very much for your problem report.
 > It has the internal identification `other/9031'.
 > The individual assigned to look at your
 > report is: unassigned. 
 > 
 > >Category:       other
 > >Responsible:    unassigned
 > >Synopsis:       Cygnus configure versus libtool AC_PROG_LD macro
 > >Arrival-Date:   Fri Dec 20 12:36:00 PST 2002
 
 The enclosed patch fixes the problem reported in other/9031.  The technique
 for determining LD is derived from that used by the libtool AC_PROG_LD macro.
 
 Tested on hppa64-hp-hpux11.11 and hppa2.0w-hp-hpux11.11.  Ok for main?
 
 Dave
 -- 
 J. David Anglin                                  dave.anglin@nrc.ca
 National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)
 
 2002-12-26  John David Anglin  <dave.anglin@nrc.ca>
 
 	* configure (LD): Try to set LD to the ld used by gcc when using gcc
 	and not doing a Canadian Cross.
 
 Index: configure
 ===================================================================
 RCS file: /cvsroot/gcc/gcc/configure,v
 retrieving revision 1.50
 diff -u -3 -p -r1.50 configure
 --- configure	16 Dec 2002 18:18:42 -0000	1.50
 +++ configure	21 Dec 2002 21:08:30 -0000
 @@ -1059,6 +1059,29 @@ else
      fi
    fi
  
 +  # If we are using gcc, try to set the LD default to the ld used by gcc.
 +  case "${CC}" in
 +  *gcc)
 +    case $host in
 +    *-*-mingw*)
 +      gcc_prog_ld="`$CC -print-prog-name=ld 2>&1 | tr -d '\015'`" ;;
 +    *)
 +      gcc_prog_ld="`$CC -print-prog-name=ld 2>&1`" ;;
 +    esac
 +    case $gcc_prog_ld in
 +    # Accept absolute paths.
 +    [\\/]* | [A-Za-z]:[\\/]*)
 +      re_direlt='/[^/][^/]*/\.\./'
 +      # Canonicalize the path of ld
 +      gcc_prog_ld=`echo $gcc_prog_ld| sed 's%\\\\%/%g'`
 +      while echo $gcc_prog_ld| grep "$re_direlt" > /dev/null 2>&1; do
 +	gcc_prog_ld=`echo $gcc_prog_ld| sed "s%$re_direlt%/%"`
 +      done
 +      LD="$gcc_prog_ld" ;;
 +    esac
 +    ;;
 +  esac
 +
    CXX=${CXX-"c++"}
    CFLAGS=${CFLAGS-"-g"}
    CXXFLAGS=${CXXFLAGS-"-g -O2"}