On 2014-05-01 10:10:46 -0400, Tom Lane wrote:
> Andres Freund <and...@2ndquadrant.com> writes:
> > I still think we really need to fix this. I have three possible
> > solutions:
>
> > a) Add an external file (in the source tree) that's included in the
> > configure test.
> > b) Have a compiler specific override and specify USE_INLINE there.
> > c) Drop the requirement of quiet inlines.
>
> > a) would probably the best, but I haven't yet found a non ugly solution :(
>
> Why is (a) so hard again?
>
> (If you're wondering where to put the file, I'd vote for the config/
> directory.)
Patch attached.
Greetings,
Andres Freund
--
Andres Freund http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services
>From 67a8595f07a93d50276f1f4dc967b33a5137ae25 Mon Sep 17 00:00:00 2001
From: Andres Freund <and...@anarazel.de>
Date: Thu, 1 May 2014 20:40:31 +0200
Subject: [PATCH] Fix "quiet inline" configure test on newer clang compilers.
The test used to just define a static inline function inline and check
whether that causes a warning. But newer clang versions warn about
unused static inline functions when defined inside a .c file, but not
when defined in a included header. Change the test to cope.
This caused postgres compiled with newer clang versions to not use
inline functions, potentially leading to performance degradations.
---
config/c-compiler.m4 | 14 ++++++++++++--
config/test_quiet_include.h | 5 +++++
configure | 2 +-
3 files changed, 18 insertions(+), 3 deletions(-)
create mode 100644 config/test_quiet_include.h
diff --git a/config/c-compiler.m4 b/config/c-compiler.m4
index 4ba3236..1b9a2db 100644
--- a/config/c-compiler.m4
+++ b/config/c-compiler.m4
@@ -19,7 +19,17 @@ fi])# PGAC_C_SIGNED
# PGAC_C_INLINE
# -------------
-# Check if the C compiler understands inline functions.
+# Check if the C compiler understands inline functions without being
+# noisy about unused static inline functions. Some older compilers
+# understand inline functions (tested by AC_C_INLINE) but warn about
+# them if they aren't used in a translation unit.
+# This test used to just define a inline function, but some compilers
+# (notably clang) got too smart and now warn about unused static
+# inline functions when defined inside a .c file, but not when defined
+# in a included header. Since the latter is what we want to use, test
+# that by excluding a header defined outside autoconf's confines. Not
+# pretty, but it works.
+#
# Defines: inline, PG_USE_INLINE
AC_DEFUN([PGAC_C_INLINE],
[AC_C_INLINE
@@ -28,7 +38,7 @@ AC_CACHE_CHECK([for quiet inline (no complaint if unreferenced)], pgac_cv_c_inli
if test "$ac_cv_c_inline" != no; then
pgac_c_inline_save_werror=$ac_c_werror_flag
ac_c_werror_flag=yes
- AC_LINK_IFELSE([AC_LANG_PROGRAM([static inline int fun () {return 0;}],[])],
+ AC_LINK_IFELSE([AC_LANG_PROGRAM([#include "$srcdir/config/test_quiet_include.h"],[])],
[pgac_cv_c_inline_quietly=yes])
ac_c_werror_flag=$pgac_c_inline_save_werror
fi])
diff --git a/config/test_quiet_include.h b/config/test_quiet_include.h
new file mode 100644
index 0000000..d2060c1
--- /dev/null
+++ b/config/test_quiet_include.h
@@ -0,0 +1,5 @@
+/*
+ * For the raison d'etre of this file, check the comment above the definition
+ * of the PGAC_C_INLINE macro in config/c-compiler.m4.
+ */
+static inline int fun () {return 0;}
diff --git a/configure b/configure
index e0dbdfe..9953389 100755
--- a/configure
+++ b/configure
@@ -9725,7 +9725,7 @@ else
ac_c_werror_flag=yes
cat confdefs.h - <<_ACEOF >conftest.$ac_ext
/* end confdefs.h. */
-static inline int fun () {return 0;}
+#include "$srcdir/config/test_quiet_include.h"
int
main ()
{
--
1.8.5.rc2.dirty
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
