7:17 a.m.
From: "Daniel P. Berrange" <berrange(a)redhat.com>
There are various problems building libvirt with clang,
which mostly revolve around compiler warning handling.
This series fixes all the problems I see with clang
3.2 on a Fedora 19 x86_64 host.
Daniel P. Berrange (6):
Ensure consistent enablement of gcc 'diagnostic' pragma
Workaround issue with clang and inline functions with static vars
Ignore cast alignment warnings in inotify code for Xen.
Correctly detect warning flags with clang
Only pass -export-dynamic to linker, not compiler
Don't duplicate compiler warning flags when linking
configure.ac | 4 ++--
daemon/Makefile.am | 1 -
m4/virt-compile-warnings.m4 | 29 +++++++++++++++++++++++++----
src/Makefile.am | 11 +++--------
src/internal.h | 2 +-
src/xen/xen_inotify.c | 3 +++
tests/vircgroupmock.c | 7 ++++++-
tools/Makefile.am | 4 +---
8 files changed, 41 insertions(+), 20 deletions(-)
--
1.8.2.1
7:17 a.m.
New subject: [libvirt] [PATCH 1/6] Ensure consistent enablement of gcc 'diagnostic' pragma
From: "Daniel P. Berrange" <berrange(a)redhat.com>
The virt-compile-warnings.m4 file would do an explicit
check for whether the compile could use the 'diagnostic'
pragma push/pop feature. The src/internal.h file would
then only enable it for GCC >= 4.6
This breaks with clang which supports the pragma but
does not claim GCC 4.6 compat. Export a variable from
the m4 check to the header file so they are consistent.
Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com>
---
m4/virt-compile-warnings.m4 | 3 +++
src/internal.h | 2 +-
2 files changed, 4 insertions(+), 1 deletion(-)
diff --git a/m4/virt-compile-warnings.m4 b/m4/virt-compile-warnings.m4
index e054913..fbeb3eb 100644
--- a/m4/virt-compile-warnings.m4
+++ b/m4/virt-compile-warnings.m4
@@ -94,6 +94,9 @@ AC_DEFUN([LIBVIRT_COMPILE_WARNINGS],[
dontwarn="$dontwarn -Wmissing-prototypes"
dontwarn="$dontwarn -Wmissing-declarations"
dontwarn="$dontwarn -Wcast-align"
+ else
+ AC_DEFINE_UNQUOTED([WORKING_PRAGMA_PUSH], 1,
+ [Define to 1 if gcc supports pragma push/pop])
fi
dnl Check whether strchr(s, char variable) causes a bogus compile
diff --git a/src/internal.h b/src/internal.h
index d819aa3..03c2493 100644
--- a/src/internal.h
+++ b/src/internal.h
@@ -215,7 +215,7 @@
# endif /* __GNUC__ */
-# if __GNUC_PREREQ (4, 6)
+# if WORKING_PRAGMA_PUSH
# define VIR_WARNINGS_NO_CAST_ALIGN \
_Pragma ("GCC diagnostic push") \
_Pragma ("GCC diagnostic ignored \"-Wcast-align\"")
--
1.8.2.1
2:16 p.m.
New subject: [libvirt] [PATCH 1/6] Ensure consistent enablement of gcc 'diagnostic' pragma
[adding bug-gnulib]
On 05/13/2013 06:17 AM, Daniel P. Berrange wrote:
From: "Daniel P. Berrange" <berrange(a)redhat.com>
The virt-compile-warnings.m4 file would do an explicit
check for whether the compile could use the 'diagnostic'
pragma push/pop feature. The src/internal.h file would
then only enable it for GCC >= 4.6
This breaks with clang which supports the pragma but
does not claim GCC 4.6 compat. Export a variable from
the m4 check to the header file so they are consistent.
Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com>
---
m4/virt-compile-warnings.m4 | 3 +++
src/internal.h | 2 +-
2 files changed, 4 insertions(+), 1 deletion(-)
ACK. Hmm, I wonder if gnulib should use something similar, for the
places where it uses pragma gcc push.
diff --git a/m4/virt-compile-warnings.m4 b/m4/virt-compile-warnings.m4
index e054913..fbeb3eb 100644
--- a/m4/virt-compile-warnings.m4
+++ b/m4/virt-compile-warnings.m4
@@ -94,6 +94,9 @@ AC_DEFUN([LIBVIRT_COMPILE_WARNINGS],[
dontwarn="$dontwarn -Wmissing-prototypes"
dontwarn="$dontwarn -Wmissing-declarations"
dontwarn="$dontwarn -Wcast-align"
+ else
+ AC_DEFINE_UNQUOTED([WORKING_PRAGMA_PUSH], 1,
+ [Define to 1 if gcc supports pragma push/pop])
fi
dnl Check whether strchr(s, char variable) causes a bogus compile
diff --git a/src/internal.h b/src/internal.h
index d819aa3..03c2493 100644
--- a/src/internal.h
+++ b/src/internal.h
@@ -215,7 +215,7 @@
# endif /* __GNUC__ */
-# if __GNUC_PREREQ (4, 6)
+# if WORKING_PRAGMA_PUSH
# define VIR_WARNINGS_NO_CAST_ALIGN \
_Pragma ("GCC diagnostic push") \
_Pragma ("GCC diagnostic ignored \"-Wcast-align\"")
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
2:48 p.m.
New subject: [libvirt] [PATCH 1/6] Ensure consistent enablement of gcc 'diagnostic' pragma
On 05/13/13 12:16, Eric Blake wrote:
I wonder if gnulib should use something similar, for the
places where it uses pragma gcc push.
I hope not. The only place gnulib does a push,
is to turn off bogus diagnostics that I hope clang
doesn't generate. If I'm wrong I suppose we can
add an "|| __clang__", but let's hope I'm right.
7:17 a.m.
New subject: [libvirt] [PATCH 2/6] Workaround issue with clang and inline functions with static vars
From: "Daniel P. Berrange" <berrange(a)redhat.com>
Clang does not like it when you pass a static variable to an
inline function
vircgroupmock.c:462:22: error: static variable 'fakesysfsdir' is
used in an inline function with external linkage [-Werror,-Wstatic-in-inline]
Just make the var non-static to avoid this
Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com>
---
tests/vircgroupmock.c | 7 ++++++-
1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/tests/vircgroupmock.c b/tests/vircgroupmock.c
index 17ea75f..f1a5700 100644
--- a/tests/vircgroupmock.c
+++ b/tests/vircgroupmock.c
@@ -37,7 +37,12 @@ static int (*realaccess)(const char *path, int mode);
static int (*reallstat)(const char *path, struct stat *sb);
static int (*real__lxstat)(int ver, const char *path, struct stat *sb);
static int (*realmkdir)(const char *path, mode_t mode);
-static char *fakesysfsdir;
+
+/* Don't make static, since it causes problems with clang
+ * when passed as an arg to asprintf()
+ * vircgroupmock.c:462:22: error: static variable 'fakesysfsdir' is used in an
inline function with external linkage [-Werror,-Wstatic-in-inline]
+ */
+char *fakesysfsdir;
# define SYSFS_PREFIX "/not/really/sys/fs/cgroup/"
--
1.8.2.1
2:26 p.m.
New subject: [libvirt] [PATCH 2/6] Workaround issue with clang and inline functions with static vars
On 05/13/2013 06:17 AM, Daniel P. Berrange wrote:
From: "Daniel P. Berrange" <berrange(a)redhat.com>
Clang does not like it when you pass a static variable to an
inline function
vircgroupmock.c:462:22: error: static variable 'fakesysfsdir' is
used in an inline function with external linkage [-Werror,-Wstatic-in-inline]
Just make the var non-static to avoid this
Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com>
---
tests/vircgroupmock.c | 7 ++++++-
1 file changed, 6 insertions(+), 1 deletion(-)
ACK. I don't know if that is a case where gcc is silently allowing a
non-conforming extension beyond C99, or whether it is a case of clang
just being brain-dead; but either way, the workaround is painless enough
that I don't mind the wider visibility.
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
7:17 a.m.
New subject: [libvirt] [PATCH 3/6] Ignore cast alignment warnings in inotify code for Xen.
From: "Daniel P. Berrange" <berrange(a)redhat.com>
The inotify Xen code causes a cast alignment warning, but this
is harmless since the kernel inotify interface will ensure
sufficient alignment of the inotify structs in the buffer being
read
Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com>
---
src/xen/xen_inotify.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/src/xen/xen_inotify.c b/src/xen/xen_inotify.c
index b032bba..e13c572 100644
--- a/src/xen/xen_inotify.c
+++ b/src/xen/xen_inotify.c
@@ -278,7 +278,10 @@ reread:
if (got < sizeof(struct inotify_event))
goto cleanup; /* bad */
+ VIR_WARNINGS_NO_CAST_ALIGN
e = (struct inotify_event *)tmp;
+ VIR_WARNINGS_RESET
+
tmp += sizeof(struct inotify_event);
got -= sizeof(struct inotify_event);
--
1.8.2.1
2:27 p.m.
New subject: [libvirt] [PATCH 3/6] Ignore cast alignment warnings in inotify code for Xen.
On 05/13/2013 06:17 AM, Daniel P. Berrange wrote:
From: "Daniel P. Berrange" <berrange(a)redhat.com>
The inotify Xen code causes a cast alignment warning, but this
is harmless since the kernel inotify interface will ensure
sufficient alignment of the inotify structs in the buffer being
read
Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com>
---
src/xen/xen_inotify.c | 3 +++
1 file changed, 3 insertions(+)
ACK.
diff --git a/src/xen/xen_inotify.c b/src/xen/xen_inotify.c
index b032bba..e13c572 100644
--- a/src/xen/xen_inotify.c
+++ b/src/xen/xen_inotify.c
@@ -278,7 +278,10 @@ reread:
if (got < sizeof(struct inotify_event))
goto cleanup; /* bad */
+ VIR_WARNINGS_NO_CAST_ALIGN
e = (struct inotify_event *)tmp;
+ VIR_WARNINGS_RESET
+
tmp += sizeof(struct inotify_event);
got -= sizeof(struct inotify_event);
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
7:17 a.m.
New subject: [libvirt] [PATCH 4/6] Correctly detect warning flags with clang
From: "Daniel P. Berrange" <berrange(a)redhat.com>
Clang will happily claim to support any warning flags
unless the -Werror and -Wunknown-warning-option flags
are set. Thus we need to make sure these are set when
testing for clags.
We must also set the clang specific warning flags
-Wno-unused-command-line-argument to avoid a warning
from the ssp-buffer-size flag when linking .o files.
Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com>
---
m4/virt-compile-warnings.m4 | 22 ++++++++++++++++++++++
1 file changed, 22 insertions(+)
diff --git a/m4/virt-compile-warnings.m4 b/m4/virt-compile-warnings.m4
index fbeb3eb..5803d48 100644
--- a/m4/virt-compile-warnings.m4
+++ b/m4/virt-compile-warnings.m4
@@ -60,6 +60,18 @@ AC_DEFUN([LIBVIRT_COMPILE_WARNINGS],[
# gcc 4.4.6 complains this is C++ only; gcc 4.7.0 implies this from -Wall
dontwarn="$dontwarn -Wenum-compare"
+ # clang rather horribly ignores unknown warning flags by
+ # default. Thus to get gl_WARN_ADD to reliably detect
+ # flags, we need to set '-Werror -Wunknown-warning-option'
+ # in CFLAGS while probing support
+ WARN_CFLAGS=
+ orig_CFLAGS="$CFLAGS"
+ gl_WARN_ADD([-Wunknown-warning-option])
+ if test -n "$WARN_CFLAGS" ; then
+ WARN_CFLAGS=
+ CFLAGS="-Werror -Wunknown-warning-option $CFLAGS"
+ fi
+
# gcc 4.2 treats attribute(format) as an implicit attribute(nonnull),
# which triggers spurious warnings for our usage
AC_CACHE_CHECK([whether gcc -Wformat allows NULL strings],
@@ -185,6 +197,13 @@ AC_DEFUN([LIBVIRT_COMPILE_WARNINGS],[
dnl gl_WARN_ADD([-fstack-protector])
gl_WARN_ADD([-fstack-protector-all])
gl_WARN_ADD([--param=ssp-buffer-size=4])
+ dnl Even though it supports it, clang complains about
+ dnl use of --param=ssp-buffer-size=4 unless used with
+ dnl the -c arg. It doesn't like it when used with args
+ dnl that just link together .o files. Unfortunately
+ dnl we can't avoid that with automake, so we must turn
+ dnl off the following clang specific warning
+ gl_WARN_ADD([-Wno-unused-command-line-argument])
;;
esac
gl_WARN_ADD([-fexceptions])
@@ -222,4 +241,7 @@ AC_DEFUN([LIBVIRT_COMPILE_WARNINGS],[
AC_DEFINE_UNQUOTED([BROKEN_GCC_WLOGICALOP], 1,
[Define to 1 if gcc -Wlogical-op reports false positives on strchr])
fi
+
+ # Remove stuff we set for clang
+ CFLAGS="$orig_CFLAGS"
])
--
1.8.2.1
2:28 p.m.
New subject: [libvirt] [PATCH 4/6] Correctly detect warning flags with clang
[adding bug-gnulib]
On 05/13/2013 06:17 AM, Daniel P. Berrange wrote:
From: "Daniel P. Berrange" <berrange(a)redhat.com>
Clang will happily claim to support any warning flags
unless the -Werror and -Wunknown-warning-option flags
are set. Thus we need to make sure these are set when
testing for clags.
We must also set the clang specific warning flags
-Wno-unused-command-line-argument to avoid a warning
from the ssp-buffer-size flag when linking .o files.
Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com>
---
m4/virt-compile-warnings.m4 | 22 ++++++++++++++++++++++
1 file changed, 22 insertions(+)
Some of this needs to be imported into upstream gnulib.
diff --git a/m4/virt-compile-warnings.m4 b/m4/virt-compile-warnings.m4
index fbeb3eb..5803d48 100644
--- a/m4/virt-compile-warnings.m4
+++ b/m4/virt-compile-warnings.m4
@@ -60,6 +60,18 @@ AC_DEFUN([LIBVIRT_COMPILE_WARNINGS],[
# gcc 4.4.6 complains this is C++ only; gcc 4.7.0 implies this from -Wall
dontwarn="$dontwarn -Wenum-compare"
+ # clang rather horribly ignores unknown warning flags by
+ # default. Thus to get gl_WARN_ADD to reliably detect
+ # flags, we need to set '-Werror -Wunknown-warning-option'
+ # in CFLAGS while probing support
+ WARN_CFLAGS=
+ orig_CFLAGS="$CFLAGS"
+ gl_WARN_ADD([-Wunknown-warning-option])
+ if test -n "$WARN_CFLAGS" ; then
+ WARN_CFLAGS=
+ CFLAGS="-Werror -Wunknown-warning-option $CFLAGS"
+ fi
+
# gcc 4.2 treats attribute(format) as an implicit attribute(nonnull),
# which triggers spurious warnings for our usage
AC_CACHE_CHECK([whether gcc -Wformat allows NULL strings],
@@ -185,6 +197,13 @@ AC_DEFUN([LIBVIRT_COMPILE_WARNINGS],[
dnl gl_WARN_ADD([-fstack-protector])
gl_WARN_ADD([-fstack-protector-all])
gl_WARN_ADD([--param=ssp-buffer-size=4])
+ dnl Even though it supports it, clang complains about
+ dnl use of --param=ssp-buffer-size=4 unless used with
+ dnl the -c arg. It doesn't like it when used with args
+ dnl that just link together .o files. Unfortunately
+ dnl we can't avoid that with automake, so we must turn
+ dnl off the following clang specific warning
+ gl_WARN_ADD([-Wno-unused-command-line-argument])
;;
esac
gl_WARN_ADD([-fexceptions])
@@ -222,4 +241,7 @@ AC_DEFUN([LIBVIRT_COMPILE_WARNINGS],[
AC_DEFINE_UNQUOTED([BROKEN_GCC_WLOGICALOP], 1,
[Define to 1 if gcc -Wlogical-op reports false positives on strchr])
fi
+
+ # Remove stuff we set for clang
+ CFLAGS="$orig_CFLAGS"
])
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
2:34 a.m.
New subject: [libvirt] [PATCH 4/6] Correctly detect warning flags with clang
On 05/13/2013 12:28 PM, Eric Blake wrote:
Some of this needs to be imported into upstream gnulib.
I gave it a try, by installing the following.
I'll follow up on bug-gnulib with 7 other clang-related patches.
From 45fc031e607cb82dce777228d9334cdac16ca648 Mon Sep 17 00:00:00 2001
From: Paul Eggert <eggert(a)cs.ucla.edu>
Date: Wed, 15 May 2013 00:07:15 -0700
Subject: [PATCH 1/8] warnings: port to clang
Problem reported by Daniel P. Berrange via Eric Blake in
<http://lists.gnu.org/archive/html/bug-gnulib/2013-05/msg00055.html>;.
* m4/warnings.m4 (gl_UNKNOWN_WARNINGS_ARE_ERRORS): New macro.
(gl_WARN_ADD): Use it.
---
ChangeLog | 8 ++++++++
m4/warnings.m4 | 15 ++++++++++++---
2 files changed, 20 insertions(+), 3 deletions(-)
diff --git a/ChangeLog b/ChangeLog
index 08f2fe0..b7aceaa 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,11 @@
+2013-05-15 Paul Eggert <eggert(a)cs.ucla.edu>
+
+ warnings: port to clang
+ Problem reported by Daniel P. Berrange via Eric Blake in
+ <http://lists.gnu.org/archive/html/bug-gnulib/2013-05/msg00055.html>;.
+ * m4/warnings.m4 (gl_UNKNOWN_WARNINGS_ARE_ERRORS): New macro.
+ (gl_WARN_ADD): Use it.
+
2013-05-11 Jim Meyering <meyering(a)fb.com>
quotearg: do not read beyond end of buffer
diff --git a/m4/warnings.m4 b/m4/warnings.m4
index 4b2ac38..1848732 100644
--- a/m4/warnings.m4
+++ b/m4/warnings.m4
@@ -1,4 +1,4 @@
-# warnings.m4 serial 7
+# warnings.m4 serial 8
dnl Copyright (C) 2008-2013 Free Software Foundation, Inc.
dnl This file is free software; the Free Software Foundation
dnl gives unlimited permission to copy and/or distribute it,
@@ -27,7 +27,7 @@ AC_DEFUN([gl_COMPILER_OPTION_IF],
AS_VAR_PUSHDEF([gl_Flags], [_AC_LANG_PREFIX[]FLAGS])dnl
AC_CACHE_CHECK([whether _AC_LANG compiler handles $1], m4_defn([gl_Warn]), [
gl_save_compiler_FLAGS="$gl_Flags"
- gl_AS_VAR_APPEND(m4_defn([gl_Flags]), [" $1"])
+ gl_AS_VAR_APPEND(m4_defn([gl_Flags]), [" $gl_unknown_warnings_are_errors
$1"])
AC_COMPILE_IFELSE([m4_default([$4], [AC_LANG_PROGRAM([])])],
[AS_VAR_SET(gl_Warn, [yes])],
[AS_VAR_SET(gl_Warn, [no])])
@@ -38,6 +38,14 @@ AS_VAR_POPDEF([gl_Flags])dnl
AS_VAR_POPDEF([gl_Warn])dnl
])
+# gl_UNKNOWN_WARNINGS_ARE_ERRORS
+# ------------------------------
+# Clang doesn't complain about unknown warning options unless one also
+# specifies -Wunknown-warning-option -Werror. Detect this.
+AC_DEFUN([gl_UNKNOWN_WARNINGS_ARE_ERRORS],
+[gl_COMPILER_OPTION_IF([-Werror -Wunknown-warning-option],
+ [gl_unknown_warnings_are_errors='-Wunknown-warning-option -Werror'],
+ [gl_unknown_warnings_are_errors=])])
# gl_WARN_ADD(OPTION, [VARIABLE = WARN_CFLAGS],
# [PROGRAM = AC_LANG_PROGRAM()])
@@ -47,7 +55,8 @@ AS_VAR_POPDEF([gl_Warn])dnl
#
# If VARIABLE is a variable name, AC_SUBST it.
AC_DEFUN([gl_WARN_ADD],
-[gl_COMPILER_OPTION_IF([$1],
+[AC_REQUIRE([gl_UNKNOWN_WARNINGS_ARE_ERRORS])
+gl_COMPILER_OPTION_IF([$1],
[gl_AS_VAR_APPEND(m4_if([$2], [], [[WARN_CFLAGS]], [[$2]]), [" $1"])],
[],
[$3])
--
1.7.11.7
7:17 a.m.
New subject: [libvirt] [PATCH 5/6] Only pass -export-dynamic to linker, not compiler
From: "Daniel P. Berrange" <berrange(a)redhat.com>
Clang does not like the -export-dynamic flag. The compiler does
not need it in the first place, so we can avoid the problem by
only setting it for the linker
Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com>
---
configure.ac | 4 ++--
src/Makefile.am | 6 +++---
2 files changed, 5 insertions(+), 5 deletions(-)
diff --git a/configure.ac b/configure.ac
index 53f78de..9e31c39 100644
--- a/configure.ac
+++ b/configure.ac
@@ -2154,7 +2154,7 @@ if test "$with_driver_modules" = "yes" || test
"$with_driver_modules" = "check";
fi
if test "$with_driver_modules" = "yes" ; then
- DRIVER_MODULE_CFLAGS="-export-dynamic"
+ DRIVER_MODULE_LDFLAGS="-export-dynamic"
case $ac_cv_search_dlopen in
no*) DRIVER_MODULE_LIBS= ;;
*) DRIVER_MODULE_LIBS=$ac_cv_search_dlopen ;;
@@ -2162,7 +2162,7 @@ if test "$with_driver_modules" = "yes" ; then
AC_DEFINE_UNQUOTED([WITH_DRIVER_MODULES], 1, [whether to build drivers as modules])
fi
AM_CONDITIONAL([WITH_DRIVER_MODULES], [test "$with_driver_modules" !=
"no"])
-AC_SUBST([DRIVER_MODULE_CFLAGS])
+AC_SUBST([DRIVER_MODULE_LDFLAGS])
AC_SUBST([DRIVER_MODULE_LIBS])
diff --git a/src/Makefile.am b/src/Makefile.am
index 4312c3c..9b9f9f2 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -14,13 +14,13 @@ INCLUDES = -I../gnulib/lib \
-DIN_LIBVIRT \
$(GETTEXT_CPPFLAGS)
-AM_CFLAGS = $(DRIVER_MODULE_CFLAGS) \
- $(LIBXML_CFLAGS) \
+AM_CFLAGS = $(LIBXML_CFLAGS) \
$(WARN_CFLAGS) \
$(LOCK_CHECKING_CFLAGS) \
$(WIN32_EXTRA_CFLAGS) \
$(COVERAGE_CFLAGS)
-AM_LDFLAGS = $(COVERAGE_LDFLAGS)
+AM_LDFLAGS = $(DRIVER_MODULE_LDFLAGS) \
+ $(COVERAGE_LDFLAGS)
EXTRA_DIST = $(conf_DATA) util/keymaps.csv
--
1.8.2.1
2:29 p.m.
New subject: [libvirt] [PATCH 5/6] Only pass -export-dynamic to linker, not compiler
On 05/13/2013 06:17 AM, Daniel P. Berrange wrote:
From: "Daniel P. Berrange" <berrange(a)redhat.com>
Clang does not like the -export-dynamic flag. The compiler does
not need it in the first place, so we can avoid the problem by
only setting it for the linker
Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com>
---
configure.ac | 4 ++--
src/Makefile.am | 6 +++---
2 files changed, 5 insertions(+), 5 deletions(-)
ACK.
diff --git a/configure.ac b/configure.ac
index 53f78de..9e31c39 100644
--- a/configure.ac
+++ b/configure.ac
@@ -2154,7 +2154,7 @@ if test "$with_driver_modules" = "yes" || test
"$with_driver_modules" = "check";
fi
if test "$with_driver_modules" = "yes" ; then
- DRIVER_MODULE_CFLAGS="-export-dynamic"
+ DRIVER_MODULE_LDFLAGS="-export-dynamic"
case $ac_cv_search_dlopen in
no*) DRIVER_MODULE_LIBS= ;;
*) DRIVER_MODULE_LIBS=$ac_cv_search_dlopen ;;
@@ -2162,7 +2162,7 @@ if test "$with_driver_modules" = "yes" ; then
AC_DEFINE_UNQUOTED([WITH_DRIVER_MODULES], 1, [whether to build drivers as modules])
fi
AM_CONDITIONAL([WITH_DRIVER_MODULES], [test "$with_driver_modules" !=
"no"])
-AC_SUBST([DRIVER_MODULE_CFLAGS])
+AC_SUBST([DRIVER_MODULE_LDFLAGS])
AC_SUBST([DRIVER_MODULE_LIBS])
diff --git a/src/Makefile.am b/src/Makefile.am
index 4312c3c..9b9f9f2 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -14,13 +14,13 @@ INCLUDES = -I../gnulib/lib \
-DIN_LIBVIRT \
$(GETTEXT_CPPFLAGS)
-AM_CFLAGS = $(DRIVER_MODULE_CFLAGS) \
- $(LIBXML_CFLAGS) \
+AM_CFLAGS = $(LIBXML_CFLAGS) \
$(WARN_CFLAGS) \
$(LOCK_CHECKING_CFLAGS) \
$(WIN32_EXTRA_CFLAGS) \
$(COVERAGE_CFLAGS)
-AM_LDFLAGS = $(COVERAGE_LDFLAGS)
+AM_LDFLAGS = $(DRIVER_MODULE_LDFLAGS) \
+ $(COVERAGE_LDFLAGS)
EXTRA_DIST = $(conf_DATA) util/keymaps.csv
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
7:17 a.m.
New subject: [libvirt] [PATCH 6/6] Don't duplicate compiler warning flags when linking
From: "Daniel P. Berrange" <berrange(a)redhat.com>
Automake already passes all CFLAGS to the linker too, so it
is not neccesary to set WARN_LDFLAGS in addition to the
WARN_CFLAGS variable.
Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com>
---
daemon/Makefile.am | 1 -
m4/virt-compile-warnings.m4 | 4 ----
src/Makefile.am | 5 -----
tools/Makefile.am | 4 +---
4 files changed, 1 insertion(+), 13 deletions(-)
diff --git a/daemon/Makefile.am b/daemon/Makefile.am
index 64126e5..f48dc85 100644
--- a/daemon/Makefile.am
+++ b/daemon/Makefile.am
@@ -111,7 +111,6 @@ libvirtd_CFLAGS = \
-DQEMUD_PID_FILE="\"$(QEMUD_PID_FILE)\""
libvirtd_LDFLAGS = \
- $(WARN_LDFLAGS) \
$(PIE_LDFLAGS) \
$(RELRO_LDFLAGS) \
$(COVERAGE_LDFLAGS)
diff --git a/m4/virt-compile-warnings.m4 b/m4/virt-compile-warnings.m4
index 5803d48..5174472 100644
--- a/m4/virt-compile-warnings.m4
+++ b/m4/virt-compile-warnings.m4
@@ -224,10 +224,6 @@ AC_DEFUN([LIBVIRT_COMPILE_WARNINGS],[
gl_WARN_ADD([-Werror])
fi
- WARN_LDFLAGS=$WARN_CFLAGS
- AC_SUBST([WARN_CFLAGS])
- AC_SUBST([WARN_LDFLAGS])
-
dnl Needed to keep compile quiet on python 2.4
save_WARN_CFLAGS=$WARN_CFLAGS
WARN_CFLAGS=
diff --git a/src/Makefile.am b/src/Makefile.am
index 9b9f9f2..6c626ac 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -1979,7 +1979,6 @@ if WITH_LIBVIRTD
libexec_PROGRAMS += libvirt_iohelper
libvirt_iohelper_SOURCES = $(UTIL_IO_HELPER_SOURCES)
libvirt_iohelper_LDFLAGS = \
- $(WARN_LDFLAGS) \
$(AM_LDFLAGS) \
$(PIE_LDFLAGS) \
$(RELRO_LDFLAGS) \
@@ -2003,7 +2002,6 @@ libexec_PROGRAMS += libvirt_parthelper
libvirt_parthelper_SOURCES = $(STORAGE_HELPER_DISK_SOURCES)
libvirt_parthelper_LDFLAGS = \
- $(WARN_LDFLAGS) \
$(AM_LDFLAGS) \
$(PIE_LDFLAGS) \
$(RELRO_LDFLAGS) \
@@ -2036,7 +2034,6 @@ libvirt_sanlock_helper_CFLAGS = \
$(PIE_CFLAGS) \
$(NULL)
libvirt_sanlock_helper_LDFLAGS = \
- $(WARN_LDFLAGS) \
$(AM_LDFLAGS) \
$(PIE_LDFLAGS) \
$(RELRO_LDFLAGS) \
@@ -2053,7 +2050,6 @@ libvirt_lxc_SOURCES = \
$(NODE_INFO_SOURCES) \
$(DATATYPES_SOURCES)
libvirt_lxc_LDFLAGS = \
- $(WARN_LDFLAGS) \
$(AM_LDFLAGS) \
$(PIE_LDFLAGS) \
$(RELRO_LDFLAGS) \
@@ -2098,7 +2094,6 @@ libexec_PROGRAMS += virt-aa-helper
virt_aa_helper_SOURCES = $(SECURITY_DRIVER_APPARMOR_HELPER_SOURCES)
virt_aa_helper_LDFLAGS = \
- $(WARN_LDFLAGS) \
$(AM_LDFLAGS) \
$(PIE_LDFLAGS) \
$(RELRO_LDFLAGS) \
diff --git a/tools/Makefile.am b/tools/Makefile.am
index 07c9f43..4136b84 100644
--- a/tools/Makefile.am
+++ b/tools/Makefile.am
@@ -98,7 +98,6 @@ virt_host_validate_SOURCES = \
$(NULL)
virt_host_validate_LDFLAGS = \
- $(WARN_LDFLAGS) \
$(PIE_LDFLAGS) \
$(RELRO_LDFLAGS) \
$(COVERAGE_LDFLAGS) \
@@ -131,10 +130,9 @@ virsh_SOURCES = \
virsh-volume.c virsh-volume.h \
$(NULL)
-virsh_LDFLAGS = $(WARN_LDFLAGS) $(COVERAGE_LDFLAGS)
+virsh_LDFLAGS = $(COVERAGE_LDFLAGS)
virsh_LDADD = \
$(STATIC_BINARIES) \
- $(WARN_LDFLAGS) \
$(PIE_LDFLAGS) \
$(RELRO_LDFLAGS) \
../src/libvirt.la \
--
1.8.2.1
2:30 p.m.
New subject: [libvirt] [PATCH 6/6] Don't duplicate compiler warning flags when linking
On 05/13/2013 06:17 AM, Daniel P. Berrange wrote:
From: "Daniel P. Berrange" <berrange(a)redhat.com>
Automake already passes all CFLAGS to the linker too, so it
is not neccesary to set WARN_LDFLAGS in addition to the
s/neccesary/necessary/
WARN_CFLAGS variable.
Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com>
---
daemon/Makefile.am | 1 -
m4/virt-compile-warnings.m4 | 4 ----
src/Makefile.am | 5 -----
tools/Makefile.am | 4 +---
4 files changed, 1 insertion(+), 13 deletions(-)
ACK.
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library http://libvirt.org
9:07 a.m.
Daniel P. Berrange wrote:
From: "Daniel P. Berrange" <berrange(a)redhat.com>
There are various problems building libvirt with clang,
which mostly revolve around compiler warning handling.
This series fixes all the problems I see with clang
3.2 on a Fedora 19 x86_64 host.
Daniel P. Berrange (6):
Ensure consistent enablement of gcc 'diagnostic' pragma
Workaround issue with clang and inline functions with static vars
Ignore cast alignment warnings in inotify code for Xen.
Correctly detect warning flags with clang
Only pass -export-dynamic to linker, not compiler
Don't duplicate compiler warning flags when linking
configure.ac | 4 ++--
daemon/Makefile.am | 1 -
m4/virt-compile-warnings.m4 | 29 +++++++++++++++++++++++++----
src/Makefile.am | 11 +++--------
src/internal.h | 2 +-
src/xen/xen_inotify.c | 3 +++
tests/vircgroupmock.c | 7 ++++++-
tools/Makefile.am | 4 +---
8 files changed, 41 insertions(+), 20 deletions(-)
Tested this patchset on FreeBSD with clang 3.1.
Had a warning like that when building examples:
gmake[2]: Entering directory `/usr/home/novel/code/libvirt/examples/openauth'
CC openauth-openauth.o
CCLD openauth
clang: warning: argument unused during compilation: '-g'
[-Wunused-command-line-argument]
gmake[2]: Leaving directory `/usr/home/novel/code/libvirt/examples/openauth'
The problem was fixed by this change:
diff --git a/m4/virt-compile-warnings.m4 b/m4/virt-compile-warnings.m4
index 5174472..dc0e7d7 100644
--- a/m4/virt-compile-warnings.m4
+++ b/m4/virt-compile-warnings.m4
@@ -191,7 +191,7 @@ AC_DEFUN([LIBVIRT_COMPILE_WARNINGS],[
dnl -fstack-protector stuff passes gl_WARN_ADD with gcc
dnl on Mingw32, but fails when actually used
case $host in
- *-*-linux*)
+ *-*-linux*|*-*-freebsd*)
dnl Fedora only uses -fstack-protector, but doesn't seem to
dnl be great overhead in adding -fstack-protector-all instead
dnl gl_WARN_ADD([-fstack-protector])
I still have another warning in gnulib:
CC regex.lo
In file included from regex.c:70:
./regex_internal.c:1397:11: warning: comparison of unsigned expression < 0 is always
false [-Wtautological-compare]
if (idx < 0 || idx >= set->nelem)
~~~ ^ ~
1 warning generated.
But I'm not sure it affects the build with -Werror because I don't use
it since it still fails at gcrypt.h deprecated stuff anyways (and I
haven't test a fix for that with clang I posted in other thread).
Roman Bogorodskiy
11:09 a.m.
On Mon, May 13, 2013 at 01:17:21PM +0100, Daniel P. Berrange wrote:
From: "Daniel P. Berrange" <berrange(a)redhat.com>
There are various problems building libvirt with clang,
which mostly revolve around compiler warning handling.
This series fixes all the problems I see with clang
3.2 on a Fedora 19 x86_64 host.
FYI, I measured the following compile times for the src/
directory
GCC: 2 minutes 40 seconds
CLang: 1 minute 50 seconds
So Clang cuts approx 30% off the GCC compile time, which is
quite nice. I might have to switch to running clang by
default for libvirt :-)
Daniel
--
|: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org -o- http://virt-manager.org :|
|: http://autobuild.org -o- http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :|
12:53 p.m.
On 13.05.2013 18:09, Daniel P. Berrange wrote:
On Mon, May 13, 2013 at 01:17:21PM +0100, Daniel P. Berrange wrote:
> From: "Daniel P. Berrange" <berrange(a)redhat.com>
>
> There are various problems building libvirt with clang,
> which mostly revolve around compiler warning handling.
> This series fixes all the problems I see with clang
> 3.2 on a Fedora 19 x86_64 host.
FYI, I measured the following compile times for the src/
directory
GCC: 2 minutes 40 seconds
CLang: 1 minute 50 seconds
So Clang cuts approx 30% off the GCC compile time, which is
quite nice. I might have to switch to running clang by
default for libvirt :-)
Daniel
FYI:
./autogen --system: 4m40.064s
make -j5 all : 1m22.870s
so I think compilation is not the major time consumer here :)
Michal
4:44 a.m.
On Mon, May 13, 2013 at 07:53:49PM +0200, Michal Privoznik wrote:
On 13.05.2013 18:09, Daniel P. Berrange wrote:
> On Mon, May 13, 2013 at 01:17:21PM +0100, Daniel P. Berrange wrote:
>> From: "Daniel P. Berrange" <berrange(a)redhat.com>
>>
>> There are various problems building libvirt with clang,
>> which mostly revolve around compiler warning handling.
>> This series fixes all the problems I see with clang
>> 3.2 on a Fedora 19 x86_64 host.
>
> FYI, I measured the following compile times for the src/
> directory
>
> GCC: 2 minutes 40 seconds
> CLang: 1 minute 50 seconds
>
> So Clang cuts approx 30% off the GCC compile time, which is
> quite nice. I might have to switch to running clang by
> default for libvirt :-)
>
> Daniel
>
FYI:
./autogen --system: 4m40.064s
make -j5 all : 1m22.870s
so I think compilation is not the major time consumer here :)
I run 'autogen.sh' perhaps 2-3 times a day. I run 'make' as many
as 10-20 times an *hour*. So yes, compilation is the major time
consumer.
Daniel
--
|: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org -o- http://virt-manager.org :|
|: http://autobuild.org -o- http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :|
4239
days inactive
4241
days old
18 comments
5 participants
participants (5)
-
Daniel P. Berrange -
Eric Blake -
Michal Privoznik -
Paul Eggert -
Roman Bogorodskiy