bug#13872: 24.3-rc1; TERM_HEADER multiply defined during 'configure'

bug#13872: 24.3-rc1; TERM_HEADER multiply defined during 'configure'

[Paul Eggert]

Building Emacs 24.3-rc1 on Fedora 17 x86-64 generates a config.log with
lots of these diagnostics:

conftest.c:242:0: warning: "TERM_HEADER" redefined [enabled by default]
conftest.c:71:0: note: this is the location of the previous definition

The problem is that TERM_HEADER is put into conftest.c twice, with
disagreeing definitions:

  #define TERM_HEADER "xterm.h"
  ...
  #define TERM_HEADER "gtkutil.h"

On this platform it happens to work, but I worry that on other
platforms the C compiler will reject any such test program (as it is
entitled to do), which will cause later configuration decisions to
go wrong.

Here's a fix.  This looks pretty safe.

=== modified file 'ChangeLog'
--- ChangeLog	2013-03-05 03:43:52 +0000
+++ ChangeLog	2013-03-05 05:21:41 +0000
@@ -1,3 +1,8 @@
+2013-03-05  Paul Eggert  <eggert@cs.ucla.edu>
+
+	* configure.ac (TERM_HEADER): Remove duplicate definition
+	that can mess up 'configure' runs.
+
 2013-03-05  Glenn Morris  <rgm@gnu.org>
 
 	* Makefile.in (install-man): Ignore gzip exit status.

=== modified file 'configure.ac'
--- configure.ac	2013-03-05 03:53:34 +0000
+++ configure.ac	2013-03-05 04:50:24 +0000
@@ -1646,11 +1646,6 @@
   ;;
 esac
 
-if test -n "${term_header}"; then
-    AC_DEFINE_UNQUOTED(TERM_HEADER, "${term_header}",
-        [Define to the header for the built-in window system.])
-fi
-
 if test "$window_system" = none && test "X$with_x" != "Xno"; then
    AC_CHECK_PROG(HAVE_XSERVER, X, true, false)
    if test "$HAVE_XSERVER" = true ||

bug#13872: 24.3-rc1; TERM_HEADER multiply defined during 'configure'

[Glenn Morris]

How marvellous to hear about this now.
Sigh.

bug#13872: 24.3-rc1; TERM_HEADER multiply defined during 'configure'

[Ken Brown]

On 3/5/2013 12:26 AM, Paul Eggert wrote:
> Here's a fix.  This looks pretty safe.

Are you sure there are no tests between the two definitions of 
TERM_HEADER for which the latter should be defined?

I'm copying Daniel Colascione, since he's the one who introduced 
TERM_HEADER.

Ken

bug#13872: 24.3-rc1; TERM_HEADER multiply defined during 'configure'

[Paul Eggert]

On 03/05/2013 05:40 AM, Ken Brown wrote:
> Are you sure there are no tests between the two definitions of TERM_HEADER
> for which the latter should be defined?

Yes, because no test uses TERM_HEADER:

$ grep TERM_HEADER configure
#define TERM_HEADER "${term_header}"
#define TERM_HEADER "${term_header}"

bug#13872: 24.3-rc1; TERM_HEADER multiply defined during 'configure'

[Glenn Morris]

OK, please apply to emacs-24.

bug#13872: 24.3-rc1; TERM_HEADER multiply defined during 'configure'

[Paul Eggert]

On 03/05/13 09:20, Glenn Morris wrote:
> 
> OK, please apply to emacs-24.

Done, as emacs-24 bzr 111321.

bug#13872: 24.3-rc1; TERM_HEADER multiply defined during 'configure'

[Eli Zaretskii]

> From: Glenn Morris <rgm@gnu.org>
> Date: Tue, 05 Mar 2013 03:28:59 -0500
> 
> 
> How marvellous to hear about this now.
> Sigh.

There's nothing you or anyone else could do about that.  First Law of
Release Management states that no matter how long you delay the
release trying to make it bug-free, a critical bug will be reported
within hours of the release announcement.