Hi,
I have noted a APR question in TC user list, if what is said is true then we
have to arrange build/apr_common.m4
Any hints?
Cheers
Jean-frederic
--- Begin Message ---
Hi all.
This is more for developers, but here goes. OS is Tru64 UNIX 4.0F (a.k.a.
Digital UNIX (a.k.a. DEC OSF/1)). I'm using native CC of Tru64. "configure"
script in APR erroneously detects the presence of <stdint.h> on the system and
that consequently bugs the whole mod_webapp.so.
mod_webapp.so, although built wouldn't load with a message:
Legba:/usr/local/apache/conf# /usr/local/apache/bin/apachectl configtest
Syntax error on line 218 of /usr/local/apache/conf/httpd.conf:
Cannot load /usr/local/apache/libexec/mod_webapp.so into server: Unresolved
symbol in /usr/local/apache/libexec/mod_webapp.so: INT64_C
I have run the portion which detects "INT64_C in stdint.h" of the "configure"
manually and it produces a wrong conclusion.
THE PROBLEM is in this:
cat >conftest.$ac_ext <<_ACEOF
#line 18159 "configure"
#include "confdefs.h"
#include <stdint.h>
#ifdef INT64_C
YES_IS_DEFINED
#endif
_ACEOF
Please note that the portion of CPP directives (those three in the middle) is
*indented*.
Well, this is not supported in Tru64's CC and frankly I don't know of a C
language specification which allows it. To make myself perfectly clear, all C
language specifications I've seen (K&R C and ANSI C), insist on "#" sign of the
CPP directive being in the first column or the first character on the line. You
can indent the directive itself, but the "#" MUST be in the first column. DEC
CC insists on this and ignores lines that have whitespace before "#". So,
something like this would be correct:
cat >conftest.$ac_ext <<_ACEOF
#line 18159 "configure"
#include "confdefs.h"
# include <stdint.h>
# ifdef INT64_C
YES_IS_DEFINED
# endif
_ACEOF
I'm guessing that GCC allows for indentation of CPP lines, but I still think it
is against the specification. Reasoning like "GCC is the major CC in the world,
comply with it" is something I would expect from Microsoft - and not from GCC,
Apache or Tomcat supporters.
I know I could build GCC for Tru64 UNIX and "save myself a lot of trouble in
the future", but that would be 50-th "meander" in building some final-stage
product on my machine, which is not only time-, but also nerve-consuming, as
well.
For the time being I'll manually "weed out all of you non-hackers, who are not
fit to serve in my beloved corps" ("Full metal jacket"-Stanley Kubrick). Of
course, untaring sources from scratch...
Nix.
smime.p7s
Description: S/MIME cryptographic signature
--- End Message ---
Index: apr_common.m4
===================================================================
RCS file: /home/cvs/apache/apr/build/apr_common.m4,v
retrieving revision 1.21
diff -u -r1.21 apr_common.m4
--- apr_common.m4 2001/07/24 10:05:36 1.21
+++ apr_common.m4 2001/10/03 15:59:01
@@ -208,10 +208,10 @@
for curhdr in $2
do
AC_EGREP_CPP(YES_IS_DEFINED, [
- #include <$curhdr>
- #ifdef $1
+#include <$curhdr>
+#ifdef $1
YES_IS_DEFINED
- #endif
+#endif
], ac_cv_define_$1=yes)
done
])
@@ -227,10 +227,10 @@
AC_DEFUN(APR_CHECK_DEFINE,[
AC_CACHE_CHECK([for $1 in $2],ac_cv_define_$1,[
AC_EGREP_CPP(YES_IS_DEFINED, [
- #include <$2>
- #ifdef $1
+#include <$2>
+#ifdef $1
YES_IS_DEFINED
- #endif
+#endif
], ac_cv_define_$1=yes, ac_cv_define_$1=no)
])
if test "$ac_cv_define_$1" = "yes"; then
@@ -243,10 +243,10 @@
dnl
AC_DEFUN(APR_CHECK_APR_DEFINE,[
AC_EGREP_CPP(YES_IS_DEFINED, [
- #include "$2/include/apr.h"
- #if $1
+#include "$2/include/apr.h"
+#if $1
YES_IS_DEFINED
- #endif
+#endif
], ac_cv_define_$1=yes, ac_cv_define_$1=no)
])
@@ -437,8 +437,8 @@
AC_MSG_CHECKING(for type of inbuf parameter to iconv)
if test "x$apr_iconv_inbuf_const" = "x"; then
APR_TRY_COMPILE_NO_WARNING([
- #include <stddef.h>
- #include <iconv.h>
+#include <stddef.h>
+#include <iconv.h>
],[
iconv(0,(char **)0,(size_t *)0,(char **)0,(size_t *)0);
], apr_iconv_inbuf_const="0", apr_iconv_inbuf_const="1")
