[Buildroot] [PATCH 1/3] gpsd: proper usage of prefix and DESTDIR

[Buildroot] [PATCH 1/3] gpsd: proper usage of prefix and DESTDIR

[Thomas Petazzoni]

prefix should always be /usr, and destdir must be passed as DESTDIR,
and in the environment, not as a scons argument. Finally, we pass the
sysroot= argument to scons so that it doesn't add -L/usr/lib
parameters when compiling.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 package/gpsd/gpsd.mk |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/package/gpsd/gpsd.mk b/package/gpsd/gpsd.mk
index 7d1dbcd..759cd9c 100644
--- a/package/gpsd/gpsd.mk
+++ b/package/gpsd/gpsd.mk
@@ -17,8 +17,9 @@ GPSD_SCONS_ENV = $(TARGET_CONFIGURE_OPTS)
 
 GPSD_SCONS_OPTS = \
 	arch=$(ARCH)\
-	prefix=$(TARGET_DIR)/usr\
+	prefix=/usr\
 	chrpath=no\
+	sysroot=$(STAGING_DIR)\
 	strip=no
 
 ifeq ($(BR2_PACKAGE_NCURSES),y)
@@ -210,9 +211,9 @@ endef
 define GPSD_INSTALL_TARGET_CMDS
 	(cd $(@D); \
 		$(GPSD_SCONS_ENV) \
+		DESTDIR=$(TARGET_DIR) \
 		$(SCONS) \
 		$(GPSD_SCONS_OPTS) \
-		destdir=$(TARGET_DIR) \
 		install)
 	if [ ! -f $(TARGET_DIR)/etc/init.d/S50gpsd ]; then \
 		$(INSTALL) -m 0755 -D package/gpsd/S50gpsd $(TARGET_DIR)/etc/init.d/S50gpsd; \
@@ -223,10 +224,9 @@ endef
 define GPSD_INSTALL_STAGING_CMDS
 	(cd $(@D); \
 		$(GPSD_SCONS_ENV) \
+		DESTDIR=$(TARGET_DIR) \
 		$(SCONS) \
 		$(GPSD_SCONS_OPTS) \
-		destdir=$(STAGING_DIR) \
-		includedir="$(STAGING_DIR)/usr/include" \
 		install)
 endef
 
-- 
1.7.9.5

[Buildroot] [PATCH 2/3] gpsd: add patch to disable rpath

[Thomas Petazzoni]

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 .../gpsd/gpsd-03-sconstruct-disable-rpath.patch    |   17 +++++++++++++++++
 1 file changed, 17 insertions(+)
 create mode 100644 package/gpsd/gpsd-03-sconstruct-disable-rpath.patch

diff --git a/package/gpsd/gpsd-03-sconstruct-disable-rpath.patch b/package/gpsd/gpsd-03-sconstruct-disable-rpath.patch
new file mode 100644
index 0000000..5476450
--- /dev/null
+++ b/package/gpsd/gpsd-03-sconstruct-disable-rpath.patch
@@ -0,0 +1,17 @@
+Disable addition of RPATH to the generated libraries
+
+Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
+
+Index: b/SConstruct
+===================================================================
+--- a/SConstruct
++++ b/SConstruct
+@@ -273,8 +273,6 @@
+     env.Prepend(LIBPATH=[installdir('libdir')])
+ else:
+     env.Prepend(LIBPATH=[env["sysroot"] + installdir('libdir')])
+-if env["shared"]:
+-    env.Prepend(RPATH=[installdir('libdir')])
+ 
+ # Give deheader a way to set compiler flags
+ if 'MORECFLAGS' in os.environ:
-- 
1.7.9.5

[Buildroot] [PATCH 2/3] gpsd: add patch to disable rpath

[Simon Dawson]

On 24 July 2012 23:10, Thomas Petazzoni
<thomas.petazzoni@free-electrons.com> wrote:
> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
> ---
>  .../gpsd/gpsd-03-sconstruct-disable-rpath.patch    |   17 +++++++++++++++++
>  1 file changed, 17 insertions(+)
>  create mode 100644 package/gpsd/gpsd-03-sconstruct-disable-rpath.patch

Acked-by: Simon Dawson <spdawson@gmail.com>

[Buildroot] [PATCH 3/3] qt: tune .pc files after installation in staging directory

[Thomas Petazzoni]

The Qt configuration and installation process is a bit strange, and it
leads the .pc files to contain absolute paths to the staging
directory, while it should only contains paths relative to the staging
directory, because pkg-config already adds the path of the sysroot
automatically to the headers and libraries locations.

This allows the 'pkg-config --cflags' results for Qt libraries to be
correct.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
---
 package/qt/qt.mk |    9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/package/qt/qt.mk b/package/qt/qt.mk
index feb0d6f..3b40ea4 100644
--- a/package/qt/qt.mk
+++ b/package/qt/qt.mk
@@ -577,13 +577,20 @@ endef
 # everything in the STAGING_DIR), we move host programs such as qmake,
 # rcc or uic to the HOST_DIR so that they are available at the usual
 # location. A qt.conf file is generated to make sure that all host
-# programs still find all files they need.
+# programs still find all files they need. The .pc files are tuned to
+# remove the sysroot path from them, since pkg-config already adds it
+# automatically.
 define QT_INSTALL_STAGING_CMDS
 	$(MAKE) -C $(@D) install
 	mkdir -p $(HOST_DIR)/usr/bin
 	mv $(addprefix $(STAGING_DIR)/usr/bin/,$(QT_HOST_PROGRAMS)) $(HOST_DIR)/usr/bin
 	ln -sf $(STAGING_DIR)/usr/mkspecs $(HOST_DIR)/usr/mkspecs
 	$(QT_INSTALL_QT_CONF)
+	for i in moc uic rcc lupdate lrelease ; do \
+		$(SED) "s,^$${i}_location=.*,$${i}_location=$(HOST_DIR)/usr/bin/$${i}," \
+			$(STAGING_DIR)/usr/lib/pkgconfig/Qt*.pc ; \
+	done
+	$(SED) "s,$(STAGING_DIR)/,,g" $(STAGING_DIR)/usr/lib/pkgconfig/Qt*.pc
 endef
 
 # Library installation
-- 
1.7.9.5

[Buildroot] [PATCH 3/3] qt: tune .pc files after installation in staging directory

[Simon Dawson]

On 24 July 2012 23:10, Thomas Petazzoni
<thomas.petazzoni@free-electrons.com> wrote:
> This allows the 'pkg-config --cflags' results for Qt libraries to be
> correct.

Acked-by: Simon Dawson <spdawson@gmail.com>

[Buildroot] [PATCH 3/3] qt: tune .pc files after installation in staging directory

[Thomas Petazzoni]

Le Wed, 25 Jul 2012 08:26:43 +0100,
Simon Dawson <spdawson@gmail.com> a ?crit :

> On 24 July 2012 23:10, Thomas Petazzoni
> <thomas.petazzoni@free-electrons.com> wrote:
> > This allows the 'pkg-config --cflags' results for Qt libraries to be
> > correct.
> 
> Acked-by: Simon Dawson <spdawson@gmail.com>

Is this an Acked-by solely for PATCH 3/3, or for all patches?

Thanks for looking at them!

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] [PATCH 3/3] qt: tune .pc files after installation in staging directory

[Simon Dawson]

Hi Thomas,

On 25 July 2012 08:43, Thomas Petazzoni
<thomas.petazzoni@free-electrons.com> wrote:
> Is this an Acked-by solely for PATCH 3/3, or for all patches?

Just for PATCH 3/3. I'm still testing PATCH 1/3 and PATCH 2/3.

Simon.

[Buildroot] [PATCH 1/3] gpsd: proper usage of prefix and DESTDIR

[Simon Dawson]

On 24 July 2012 23:10, Thomas Petazzoni
<thomas.petazzoni@free-electrons.com> wrote:
> prefix should always be /usr, and destdir must be passed as DESTDIR,
> and in the environment, not as a scons argument. Finally, we pass the
> sysroot= argument to scons so that it doesn't add -L/usr/lib
> parameters when compiling.

Hi Thomas. The build is now falling over for me:

/opt/tfx/arch/arm/usr/bin/ccache /usr/lib/ccache/gcc -pthread -o
strl-py_2_7_2_final_0.so -c -fno-strict-aliasing -O2 -fPIC -DNDEBUG -g
-O3 -Wall -Wstrict-prototypes
-I/opt/tfx/arch/arm/usr/include/python2.7 strl.c
/opt/tfx/arch/arm/usr/bin/ccache /usr/lib/ccache/gcc -pthread -shared
-L/opt/tfx/arch/arm/lib -L/opt/tfx/arch/arm/usr/lib
-Wl,-rpath,/opt/tfx/arch/arm/usr/lib -o gps/clienthelpers.so
gpsclient-py_2_7_2_final_0.so geoid-py_2_7_2_final_0.so
gpsdclient-py_2_7_2_final_0.so strl-py_2_7_2_final_0.so -L.
-L/opt/tfx/arch/arm/usr/arm-unknown-linux-uclibcgnueabi/sysroot/usr/lib
-lrt
/usr/bin/ld: skipping incompatible
/opt/tfx/arch/arm/usr/arm-unknown-linux-uclibcgnueabi/sysroot/usr/lib/librt.so
when searching for -lrt
/usr/bin/ld: skipping incompatible
/opt/tfx/arch/arm/usr/arm-unknown-linux-uclibcgnueabi/sysroot/usr/lib/librt.a
when searching for -lrt
/usr/bin/ld: cannot find /lib/libpthread.so.0 inside
/usr/bin/ld: cannot find /usr/lib/libpthread_nonshared.a inside
collect2: ld returned 1 exit status
scons: *** [gps/clienthelpers.so] Error 1
scons: building terminated because of errors.

> @@ -223,10 +224,9 @@ endef
>  define GPSD_INSTALL_STAGING_CMDS
>         (cd $(@D); \
>                 $(GPSD_SCONS_ENV) \
> +               DESTDIR=$(TARGET_DIR) \
>                 $(SCONS) \
>                 $(GPSD_SCONS_OPTS) \
> -               destdir=$(STAGING_DIR) \
> -               includedir="$(STAGING_DIR)/usr/include" \
>                 install)
>  endef

Should DESTDIR be set to $(STAGING_DIR), rather than $(TARGET_DIR), here?

Simon.

[Buildroot] [PATCH 1/3] gpsd: proper usage of prefix and DESTDIR

[Thomas Petazzoni]

Le Wed, 25 Jul 2012 08:53:54 +0100,
Simon Dawson <spdawson@gmail.com> a ?crit :

> Hi Thomas. The build is now falling over for me:

This is the Python support of gpsd, correct? Was it building before my
patches?

Because if you look closely:

> /opt/tfx/arch/arm/usr/bin/ccache /usr/lib/ccache/gcc -pthread -o

It is building with the native compiler gcc and not the cross-compiler,
so obviously it cannot work. 

> > @@ -223,10 +224,9 @@ endef
> >  define GPSD_INSTALL_STAGING_CMDS
> >         (cd $(@D); \
> >                 $(GPSD_SCONS_ENV) \
> > +               DESTDIR=$(TARGET_DIR) \
> >                 $(SCONS) \
> >                 $(GPSD_SCONS_OPTS) \
> > -               destdir=$(STAGING_DIR) \
> > -               includedir="$(STAGING_DIR)/usr/include" \
> >                 install)
> >  endef
> 
> Should DESTDIR be set to $(STAGING_DIR), rather than $(TARGET_DIR), here?

Good point, will fix.

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] [PATCH 1/3] gpsd: proper usage of prefix and DESTDIR

[Simon Dawson]

On 25 July 2012 09:22, Thomas Petazzoni
<thomas.petazzoni@free-electrons.com> wrote:
> This is the Python support of gpsd, correct? Was it building before my
> patches?
>
> Because if you look closely:
>
>> /opt/tfx/arch/arm/usr/bin/ccache /usr/lib/ccache/gcc -pthread -o
>
> It is building with the native compiler gcc and not the cross-compiler,
> so obviously it cannot work.

Yes, you're right. It was using the native compiler before your
patches, but somehow managed to get through the build without falling
over. Another problem then.

Simon.

[Buildroot] [PATCH 1/3] gpsd: proper usage of prefix and DESTDIR

[Simon Dawson]

Acked-by: Simon Dawson <spdawson@gmail.com>

[Buildroot] [PATCH 1/3] gpsd: proper usage of prefix and DESTDIR

[Thomas Petazzoni]

Le Wed, 25 Jul 2012 09:49:35 +0100,
Simon Dawson <spdawson@gmail.com> a ?crit :

> Yes, you're right. It was using the native compiler before your
> patches, but somehow managed to get through the build without falling
> over. Another problem then.

Can you share a defconfig that shows the problem? I've tested the build
with BR2_PACKAGE_PYTHON + BR2_PACKAGE_GPSD, and it just builds (even
though the output is completely incorrect: the gpsd Python modules are
installed in the wrong location and are built for the host rather than
the target).

Do you actually need the Python support in gpsd? :-)

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] [PATCH 1/3] gpsd: proper usage of prefix and DESTDIR

[Simon Dawson]

Hi Thomas

On 25 July 2012 09:55, Thomas Petazzoni
<thomas.petazzoni@free-electrons.com> wrote:
> Can you share a defconfig that shows the problem? I've tested the build
> with BR2_PACKAGE_PYTHON + BR2_PACKAGE_GPSD, and it just builds (even
> though the output is completely incorrect: the gpsd Python modules are
> installed in the wrong location and are built for the host rather than
> the target).

BR2_arm=y
BR2_arm926t=y
BR2_TOOLCHAIN_BUILDROOT_LARGEFILE=y
BR2_TOOLCHAIN_BUILDROOT_INET_IPV6=y
BR2_TOOLCHAIN_BUILDROOT_WCHAR=y
BR2_TOOLCHAIN_BUILDROOT_CXX=y
BR2_PACKAGE_MPFR=y
BR2_PACKAGE_GPSD=y
BR2_PACKAGE_GPSD_NTP_SHM=y
BR2_PACKAGE_PYTHON=y
BR2_PACKAGE_PYTHON_PY_PYC=y
BR2_PACKAGE_PYTHON_NFC=y

> Do you actually need the Python support in gpsd? :-)

I think the Python support is needed for gpsfake, gpscat, and a few
other things. For my particular use case, it's not required.

Simon.

[Buildroot] [PATCH 1/3] gpsd: proper usage of prefix and DESTDIR

[Thomas Petazzoni]

Le Wed, 25 Jul 2012 11:35:44 +0100,
Simon Dawson <spdawson@gmail.com> a ?crit :

> On 25 July 2012 09:55, Thomas Petazzoni
> <thomas.petazzoni@free-electrons.com> wrote:
> > Can you share a defconfig that shows the problem? I've tested the
> > build with BR2_PACKAGE_PYTHON + BR2_PACKAGE_GPSD, and it just
> > builds (even though the output is completely incorrect: the gpsd
> > Python modules are installed in the wrong location and are built
> > for the host rather than the target).
> 
> BR2_arm=y
> BR2_arm926t=y
> BR2_TOOLCHAIN_BUILDROOT_LARGEFILE=y
> BR2_TOOLCHAIN_BUILDROOT_INET_IPV6=y
> BR2_TOOLCHAIN_BUILDROOT_WCHAR=y
> BR2_TOOLCHAIN_BUILDROOT_CXX=y
> BR2_PACKAGE_MPFR=y
> BR2_PACKAGE_GPSD=y
> BR2_PACKAGE_GPSD_NTP_SHM=y
> BR2_PACKAGE_PYTHON=y
> BR2_PACKAGE_PYTHON_PY_PYC=y
> BR2_PACKAGE_PYTHON_NFC=y

This configuration builds fine here. So I guess the difference is that
I build inside a chroot in which I have almost no development package
installed in the host, and you build in your normal development
machine, on which you have many development packages installed, and
those are influencing the build somehow.

For now, I would suggest that we leave the Python support in gpsd on
the side, until someone who actually needs it steps up to fix the
issues. So I'll force python=no unconditionally. Do you agree with this?

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com

[Buildroot] [PATCH 1/3] gpsd: proper usage of prefix and DESTDIR

[Simon Dawson]

On 25 July 2012 15:11, Thomas Petazzoni
<thomas.petazzoni@free-electrons.com> wrote:
> For now, I would suggest that we leave the Python support in gpsd on
> the side, until someone who actually needs it steps up to fix the
> issues. So I'll force python=no unconditionally. Do you agree with this?

Yes, absolutely.

Simon.

[Buildroot] [PATCH 1/3] gpsd: proper usage of prefix and DESTDIR

[Thomas Petazzoni]

Le Wed, 25 Jul 2012 00:10:29 +0200,
Thomas Petazzoni <thomas.petazzoni@free-electrons.com> a ?crit :

> prefix should always be /usr, and destdir must be passed as DESTDIR,
> and in the environment, not as a scons argument. Finally, we pass the
> sysroot= argument to scons so that it doesn't add -L/usr/lib
> parameters when compiling.
> 
> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

Applied all 3 patches, with the TARGET_DIR -> STAGING_DIR fix
mentioned by Simon.

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com