headers
Patrick Welche | 1 Feb 2004 17:06
Picon
Picon

Re: [PATCH] Shell selection bug fixes

On Tue, Jan 13, 2004 at 10:15:31AM -0800, Paul Eggert wrote:
> Patrick Welche <prlw1 <at> newn.cam.ac.uk> writes:
> 
> > Is there any chance this patch could be applied, and the cvs files
> > it affects updated?
> 
> Not until we get the paperwork from Eric Sunshine.  These things take
> time, unfortunately.
> 
> Is there a quick short-term patch that would fix the problem for you?
> 
> (In the long run, as Eric mentioned, we seem to prefer a different approach.)

Presumably it is what appeared Friday :-) All tests pass for me, and it
so happens that ksh is chosen over sh (fine by me)

Thank you,

Patrick
Permalink | Reply | 0 comments |
headers
Bonzini | 2 Feb 2004 13:11
Picon
Gravatar

Re: Be done with AU_DEFUN's third argument

This one seems quite obvious, so I'll install this tomorrow
if I get no reaction (for example with respect to my English).

Fixing autoupdate's warnings is actually very easy.  The bug
is in macros whose AU_DEFUN third argument includes $0, it
is fixed by expanding the warning in the AU_DEFUN'd macro
instead of expanding it in _au_warn_XXX.  I hope this is
clearer with the new documentation.

Paolo

2004-02-02  Paolo Bonzini  <bonzini <at> gnu.org>

	* bin/autoupdate.in: Define __file__ so that warnings
	refer to the correct file.
	* doc/autoconf.texi (AU_DEFUN): Describe more correctly
	the behavior of the third argument.
	* lib/autoconf/autoupdate.m4 (AU_DEFUN): Describe more
	correctly the behavior of the third argument.  Document
	what the three macros that AU_DEFUN defines do.  Fix
	warning message when the third argument includes $0
	(reported by Alexandre Duret-Lutz).

Index: bin/autoupdate.in
===================================================================
RCS file: /cvsroot/autoconf/autoconf/bin/autoupdate.in,v
retrieving revision 1.50
diff -u -r1.50 autoupdate.in
--- bin/autoupdate.in	30 Jan 2004 14:21:35 -0000	1.50
+++ bin/autoupdate.in	2 Feb 2004 11:39:49 -0000
(Continue reading)

Permalink | Reply | 668 comments |
headers
Paul Eggert | 2 Feb 2004 19:35
Favicon

Re: Be done with AU_DEFUN's third argument

Bonzini <bonzini <at> gnu.org> writes:

> This one seems quite obvious, so I'll install this tomorrow
> if I get no reaction (for example with respect to my English).

Looks good to me.  One minor English point.  In this text:

> +replaced by the modern  <at> var{implementation}.   <at> var{message} should
> +include information on what to do after running  <at> command{autoupdate},
> +and  <at> command{autoupdate} will print it as a warning,  <at> emph{and} include
> +it in the updated  <at> file{configure.ac} file.

Change: "after running  <at> command{autoupdate}, and  <at> command{autoupdate}"
to:     "after running  <at> command{autoupdate};  <at> command{autoupdate}".

Change: " <at> emph{and}"
to:     "and".
Permalink | Reply | 0 comments |
headers
Alexandre Duret-Lutz | 3 Feb 2004 19:55
X-Face
Picon

Re: Be done with AU_DEFUN's third argument

>>> "Bonzini" == Bonzini  <bonzini <at> gnu.org> writes:

[...]

 Bonzini> --- lib/autoconf/autoupdate.m4	30 Jan 2004 14:21:36 -0000	1.7
 Bonzini> +++ lib/autoconf/autoupdate.m4	2 Feb 2004 11:39:53 -0000
 Bonzini>  <at>  <at>  -67,8 +67,9  <at>  <at> 
 Bonzini> # AU_DEFUN(NAME, NEW-CODE, [MESSAGE])
 Bonzini> # -----------------------------------
 Bonzini> # Declare that the macro NAME is now obsoleted, and should be replaced
 Bonzini> -# by NEW-CODE.  Tell the user she should run autoupdate, and include
 Bonzini> -# the additional MESSAGE.
 Bonzini> +# by NEW-CODE.  Tell the user she should run autoupdate, and when
 Bonzini> +# autoupdate is run, emit MESSAGE as a warning and include it in
 Bonzini> +# the updated configure.ac file.
 Bonzini> #
 Bonzini> # Also define NAME as a macro which code is NEW-CODE.
 Bonzini> #
 Bonzini>  <at>  <at>  -76,15 +77,31  <at>  <at> 
 Bonzini> # and to update a configure.ac.
 Bonzini> # See `acobsolete.m4' for a longer description.
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 Bonzini> m4_define([AU_DEFUN],
[...]

I finally found what this comments was referring to.  The file
was in fact named `acoldnames.m4', and all its introductory
comment has now been moved to the end of `autoupdate.in'.

Could you update this comment to
(Continue reading)

Permalink | Reply | 668 comments |
headers
Paolo Bonzini | 4 Feb 2004 00:06
Picon
Gravatar

Re: Be done with AU_DEFUN's third argument

Both comments addressed thus:

2004-02-04   Paolo Bonzini  <bonzini <at> gnu.org>

        * doc/autoconf.texi (AU_DEFUN): Fix English,
        suggested by Paul Eggert.
        * lib/autoconf/autoupdate.m4: Correct reference to
        acobsolete.m4, suggested by Alexandre Duret-Lutz.

Paolo

Index: doc/autoconf.texi
===================================================================
RCS file: /cvsroot/autoconf/autoconf/doc/autoconf.texi,v
retrieving revision 1.799
diff -u -r1.799 autoconf.texi
--- doc/autoconf.texi   3 Feb 2004 09:00:27 -0000       1.799
+++ doc/autoconf.texi   3 Feb 2004 23:03:16 -0000
 <at>  <at>  -8941,9 +8941,9  <at>  <at> 

 If she then uses  <at> command{autoupdate}, the call to  <at> var{old-macro} will be
 replaced by the modern  <at> var{implementation}.   <at> var{message} should
-include information on what to do after running  <at> command{autoupdate},
-and  <at> command{autoupdate} will print it as a warning,  <at> emph{and} include
-it in the updated  <at> file{configure.ac} file.
+include information on what to do after running  <at> command{autoupdate};
+ <at> command{autoupdate} will print it as a warning, and include it
+in the updated  <at> file{configure.ac} file.
  <at> end defmac
(Continue reading)

Permalink | Reply | 0 comments |
headers
Andreas Waechter | 5 Feb 2004 20:57
Picon
Favicon

fix for detecting fortran libraries on Darwin

Hi,

Autoconf 2.59 doesn't seem to handle the detection of fortran libraries
correctly on newer Darwin (MacOS) systems.  Typical error message is

checking for Fortran libraries of f77...  -lcrt2.o
-L/finkinstalled/lib/gcc/powerpc-apple-darwin7.0.0/3.4
-L/finkinstalled/lib/gcc/powerpc-apple-darwin7.0.0/3.4/../../..
-lfrtbegin -lg2c -lSystem
checking for dummy main to link with Fortran libraries... unknown
configure: error: linking to Fortran libraries from C fails

Problem is that '-lcrt2.o' is thought to be part of the fortran libraries.

I seem to have fixed it by changing line 650 in lib/autoconf/fortran.m4
from

<         -lang* | -lcrt[[01]].o | -lcrtbegin.o | -lc | -lgcc | -libmil | -LANG:=*)

to

>         -lang* | -lcrt[[012]].o | -lcrtbegin.o | -lc | -lgcc | -libmil | -LANG:=*)

This seems also to be the cause for the failure reported in

http://mail.gnu.org/archive/html/bug-autoconf/2003-12/msg00090.html

Hope this helps.

Andreas
(Continue reading)

Permalink | Reply | 501 comments |
headers
Rainer Orth | 6 Feb 2004 23:17
Picon
Picon

Work around IRIX 6 ld bug creating executables

The following patch arised when building current GCC mainline on IRIX 6.5
with GNU as:

	http://gcc.gnu.org/ml/gcc-patches/2004-02/msg00592.html

Consider the following configure.ac:

AC_INIT(aout_perm, 0.1)
touch a.out
AC_PROG_CC

On IRIX 6, configure fails like this:

$ ./configure 
checking for gcc... gcc
checking for C compiler default output... a.out
checking whether the C compiler works... configure: error: cannot run C compiled programs.
If you meant to cross compile, use `--host'.
See `config.log' for more details.

This happens because ld writes into an existing a.out instead of removing
it first or making it executable.

The following patch fixes this, and introduced no testsuite failures.

	Rainer

-----------------------------------------------------------------------------
Rainer Orth, Faculty of Technology, Bielefeld University
(Continue reading)

Permalink | Reply | 6 comments |
headers
Akim Demaille | 9 Feb 2004 14:52
Picon
Picon
Favicon

Re: Work around IRIX 6 ld bug creating executables


 > The following patch arised when building current GCC mainline on IRIX 6.5
 > with GNU as:

 > 	http://gcc.gnu.org/ml/gcc-patches/2004-02/msg00592.html

 > Consider the following configure.ac:

 > AC_INIT(aout_perm, 0.1)
 > touch a.out
 > AC_PROG_CC

 > On IRIX 6, configure fails like this:

 > $ ./configure 
 > checking for gcc... gcc
 > checking for C compiler default output... a.out
 > checking whether the C compiler works... configure: error: cannot run C compiled programs.
 > If you meant to cross compile, use `--host'.
 > See `config.log' for more details.

 > This happens because ld writes into an existing a.out instead of removing
 > it first or making it executable.

 > The following patch fixes this, and introduced no testsuite failures.

 > 	Rainer

 > -----------------------------------------------------------------------------
 > Rainer Orth, Faculty of Technology, Bielefeld University
(Continue reading)

Permalink | Reply | 1 comment |
headers
Rainer Orth | 9 Feb 2004 14:56
Picon
Picon

Re: Work around IRIX 6 ld bug creating executables

Akim Demaille writes:

>  > 	http://gcc.gnu.org/ml/gcc-patches/2004-02/msg00592.html

> It looks good, but isn't there another bug?  Where is this a.out
> coming from originally?

As described in the reference above, gas -v < /dev/null produces it.  But
the origin of a.out should be immaterial: if it exists for any reason, the
execution test will spuriously fail, so it should be removed first.

	Rainer

-----------------------------------------------------------------------------
Rainer Orth, Faculty of Technology, Bielefeld University
Permalink | Reply | 0 comments |
headers
Paul Eggert | 9 Feb 2004 19:39
Favicon

Re: fix for detecting fortran libraries on Darwin

Thanks for the bug report and patch.  I installed this more-generous
fix instead:

2004-02-09  Paul Eggert  <eggert <at> twinsun.com>

	* lib/autoconf/fortran.m4 (_AC_FC_LIBRARY_LDFLAGS): Ignore all
	flags of the form -lcrt*.o, not just -lcrt[01].o and -lcrtbegin.o.
	Darwin uses -lcrt2.o and there's little point to cataloging all
	the system variants.  Partial fix reported by Andreas Waechter in:
	http://mail.gnu.org/archive/html/autoconf-patches/2004-02/msg00006.html
	for bug reported by Nelson H. F. Beebe in:
	http://mail.gnu.org/archive/html/bug-autoconf/2003-12/msg00090.html

--- fortran.m4.~1.181.~	Fri Oct 31 00:44:47 2003
+++ fortran.m4	Mon Feb  9 10:28:40 2004
 <at>  <at>  -647,7 +647,7  <at>  <at>  while test $[ <at> %: <at> ] != 1; do
              [_AC_LINKER_OPTION([$ac_arg], ac_cv_[]_AC_LANG_ABBREV[]_libs)])
           ;;
           # Ignore these flags.
-        -lang* | -lcrt[[01]].o | -lcrtbegin.o | -lc | -lgcc | -libmil | -LANG:=*)
+        -lang* | -lcrt*.o | -lc | -lgcc | -libmil | -LANG:=*)
           ;;
         -lkernel32)
           test x"$CYGWIN" != xyes && ac_cv_[]_AC_LANG_ABBREV[]_libs="$ac_cv_[]_AC_LANG_ABBREV[]_libs $ac_arg"
Permalink | Reply | 501 comments |

Gmane