[PATCH 0/9] Fixes for a distro with a merged /usr

[PATCH 0/9] Fixes for a distro with a merged /usr

[Joshua Lock]

I've been playing around with building a distro with a merged /usr[1] and
the following series fixes several issues I ran across in OE core metadata.

These have been build tested by building a core-image-sato image with
ipk packaging for qemux86 with both a standard/default configuration and
for my merged /usr test distro.

I used buildhistory to compare core-image-sato before and after this series
and the only change is:

Changes to images/qemux86/glibc/core-image-sato (files-in-image.txt):
  /bin/lsmod.kmod changed symlink target from ../bin/kmod to kmod

Regards,

Joshua

1. https://wiki.freedesktop.org/www/Software/systemd/TheCaseForTheUsrMerge/

The following changes since commit bdeb32b4cdbe316f17c2fd854d59e05e8f2e8ffc:

  rt-tests: drop unnecessary added-missing-dependencies.patch (2015-09-01 11:43:38 +0100)

are available in the git repository at:

  git://git.openembedded.org/openembedded-core-contrib joshuagl/merged-usr
  http://cgit.openembedded.org/cgit.cgi/openembedded-core-contrib/log/?h=joshuagl/merged-usr

Joshua Lock (9):
  kmod: fix link creation when base_bindir != /bin
  busybox: fixes for when base_bindir != /bin
  systemd: remove hard-coded paths in FILES entries
  alsa-utils: handle udev dir being 2 levels below /
  pulseaudio: handle udev dir being 2 levels below /
  bluez: handle udev dir being 2 levels below /
  pcmciautils: handle udev dir being 2 levels below /
  nfs-utils: don't force use of /sbin as sbindir
  procps: fix for base_sbindir == sbindir

 meta/recipes-bsp/pcmciautils/pcmciautils_018.bb    |  4 +-
 meta/recipes-connectivity/bluez5/bluez5.inc        |  1 +
 .../nfs-utils/nfs-utils_1.3.1.bb                   |  8 ++++
 meta/recipes-core/busybox/busybox.inc              | 49 ++++++++++++----------
 meta/recipes-core/systemd/systemd_219.bb           | 10 ++---
 meta/recipes-extended/procps/procps_3.3.10.bb      |  5 ++-
 meta/recipes-kernel/kmod/kmod_git.bb               |  4 +-
 meta/recipes-multimedia/alsa/alsa-utils_1.0.29.bb  |  2 +-
 meta/recipes-multimedia/pulseaudio/pulseaudio.inc  |  2 +-
 9 files changed, 49 insertions(+), 36 deletions(-)

-- 
2.1.4

[PATCH 1/9] kmod: fix link creation when base_bindir != /bin

[Joshua Lock]

If base_bindir is not a direct child of / the link creation in
do_install_append creates incorrect relative links.

Instead pass a full path to the link source and use the -r flag
to ln to ensure the link is relative.

Signed-off-by: Joshua Lock <joshua.lock@collabora.co.uk>
---
 meta/recipes-kernel/kmod/kmod_git.bb | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/meta/recipes-kernel/kmod/kmod_git.bb b/meta/recipes-kernel/kmod/kmod_git.bb
index ba4d85e..120fd87 100644
--- a/meta/recipes-kernel/kmod/kmod_git.bb
+++ b/meta/recipes-kernel/kmod/kmod_git.bb
@@ -21,9 +21,9 @@ do_install_append () {
         install -dm755 ${D}${base_bindir}
         install -dm755 ${D}${base_sbindir}
         # add symlinks to kmod
-        ln -s ..${base_bindir}/kmod ${D}${base_bindir}/lsmod
+        ln -sr ${D}${base_bindir}/kmod ${D}${base_bindir}/lsmod
         for tool in insmod rmmod depmod modinfo modprobe; do
-                ln -s ..${base_bindir}/kmod ${D}${base_sbindir}/${tool}
+                ln -sr ${D}${base_bindir}/kmod ${D}${base_sbindir}/${tool}
         done
         # configuration directories
         install -dm755 ${D}${base_libdir}/depmod.d
-- 
2.1.4

[PATCH 2/9] busybox: fixes for when base_bindir != /bin

[Joshua Lock]

* Replace all hard-coded paths with variables
* Run sed over busybox.links.* to replace /bin with ${base_bindir}

Signed-off-by: Joshua Lock <joshua.lock@collabora.co.uk>
---
 meta/recipes-core/busybox/busybox.inc | 49 +++++++++++++++++++----------------
 1 file changed, 26 insertions(+), 23 deletions(-)

diff --git a/meta/recipes-core/busybox/busybox.inc b/meta/recipes-core/busybox/busybox.inc
index 0769d92..92b4a29 100644
--- a/meta/recipes-core/busybox/busybox.inc
+++ b/meta/recipes-core/busybox/busybox.inc
@@ -178,6 +178,9 @@ do_install () {
 	if [ "${base_sbindir}" != "/sbin" ]; then
 		sed -i "s:^/sbin/:${base_sbindir}/:" busybox.links*
 	fi
+	if [ "${base_bindir}" != "/bin" ]; then
+		sed -i "s:^/bin/:${base_bindir}/:" busybox.links*
+	fi
 
 	install -d ${D}${sysconfdir}/init.d
 
@@ -338,10 +341,10 @@ python do_package_prepend () {
         return
 
     if os.path.exists('%s/etc/busybox.links' % (dvar)):
-        set_alternative_vars("/etc/busybox.links", "/bin/busybox")
+        set_alternative_vars("${sysconfdir}/busybox.links", "${base_bindir}/busybox")
     else:
-        set_alternative_vars("/etc/busybox.links.nosuid", "/bin/busybox.nosuid")
-        set_alternative_vars("/etc/busybox.links.suid", "/bin/busybox.suid")
+        set_alternative_vars("${sysconfdir}/busybox.links.nosuid", "${base_bindir}/busybox.nosuid")
+        set_alternative_vars("${sysconfdir}/busybox.links.suid", "${base_bindir}/busybox.suid")
 }
 
 pkg_postinst_${PN} () {
@@ -353,22 +356,22 @@ pkg_postinst_${PN} () {
 	if test "x$D" = "x"; then
 		# Remove busybox.nosuid if it's a symlink, because this situation indicates
 		# that we're installing or upgrading to a one-binary busybox.
-		if test -h /bin/busybox.nosuid; then
-			rm -f /bin/busybox.nosuid
+		if test -h ${bindir}/busybox.nosuid; then
+			rm -f ${bindir}/busybox.nosuid
 		fi
 		for suffix in "" ".nosuid" ".suid"; do
-			if test -e /etc/busybox.links$suffix; then
+			if test -e ${sysconfdir}/busybox.links$suffix; then
 				while read link; do
 					if test ! -e "$link"; then
 						case "$link" in
 							/*/*/*)
-								to="../../bin/busybox$suffix"
+								to="../..${base_bindir}/busybox$suffix"
 								;;
 							/bin/*)
 								to="busybox$suffix"
 								;;
 							/*/*)
-								to="../bin/busybox$suffix"
+								to="..${base_bindir}/busybox$suffix"
 								;;
 						esac
 						# we can use busybox here because even if we are using splitted busybox
@@ -376,7 +379,7 @@ pkg_postinst_${PN} () {
 						busybox rm -f $link
 						busybox ln -s $to $link
 					fi
-				done < /etc/busybox.links$suffix
+				done < ${sysconfdir}/busybox.links$suffix
 			fi
 		done
 	fi
@@ -387,19 +390,19 @@ pkg_prerm_${PN} () {
 	# providing its files, this will make update-alternatives work, but the update-rc.d part
 	# for syslog, httpd and/or udhcpd will fail if there is no other package providing sh
 	tmpdir=`mktemp -d /tmp/busyboxrm-XXXXXX`
-	ln -s /bin/busybox $tmpdir/[
-	ln -s /bin/busybox $tmpdir/test
-	ln -s /bin/busybox $tmpdir/head
-	ln -s /bin/busybox $tmpdir/sh
-	ln -s /bin/busybox $tmpdir/basename
-	ln -s /bin/busybox $tmpdir/echo
-	ln -s /bin/busybox $tmpdir/mv
-	ln -s /bin/busybox $tmpdir/ln
-	ln -s /bin/busybox $tmpdir/dirname
-	ln -s /bin/busybox $tmpdir/rm
-	ln -s /bin/busybox $tmpdir/sed
-	ln -s /bin/busybox $tmpdir/sort
-	ln -s /bin/busybox $tmpdir/grep
+	ln -s ${base_bindir}/busybox $tmpdir/[
+	ln -s ${base_bindir}/busybox $tmpdir/test
+	ln -s ${base_bindir}/busybox $tmpdir/head
+	ln -s ${base_bindir}/busybox $tmpdir/sh
+	ln -s ${base_bindir}/busybox $tmpdir/basename
+	ln -s ${base_bindir}/busybox $tmpdir/echo
+	ln -s ${base_bindir}/busybox $tmpdir/mv
+	ln -s ${base_bindir}/busybox $tmpdir/ln
+	ln -s ${base_bindir}/busybox $tmpdir/dirname
+	ln -s ${base_bindir}/busybox $tmpdir/rm
+	ln -s ${base_bindir}/busybox $tmpdir/sed
+	ln -s ${base_bindir}/busybox $tmpdir/sort
+	ln -s ${base_bindir}/busybox $tmpdir/grep
 	export PATH=$PATH:$tmpdir
 }
 
@@ -407,7 +410,7 @@ pkg_prerm_${PN}-syslog () {
 	# remove syslog
 	if test "x$D" = "x"; then
 		if test "$1" = "upgrade" -o "$1" = "remove"; then
-			/etc/init.d/syslog stop
+			${sysconfdir}/init.d/syslog stop
 		fi
 	fi
 }
-- 
2.1.4

[PATCH 3/9] systemd: remove hard-coded paths in FILES entries

[Joshua Lock]

Signed-off-by: Joshua Lock <joshua.lock@collabora.co.uk>
---
 meta/recipes-core/systemd/systemd_219.bb | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/meta/recipes-core/systemd/systemd_219.bb b/meta/recipes-core/systemd/systemd_219.bb
index 4d32fd5..fbd86e8 100644
--- a/meta/recipes-core/systemd/systemd_219.bb
+++ b/meta/recipes-core/systemd/systemd_219.bb
@@ -306,10 +306,10 @@ FILES_${PN} = " ${base_bindir}/* \
                 ${exec_prefix}/lib/sysctl.d \
                 ${exec_prefix}/lib/sysusers.d \
                 ${localstatedir} \
-                /lib/udev/rules.d/70-uaccess.rules \
-                /lib/udev/rules.d/71-seat.rules \
-                /lib/udev/rules.d/73-seat-late.rules \
-                /lib/udev/rules.d/99-systemd.rules \
+                ${base_libdir}/udev/rules.d/70-uaccess.rules \
+                ${base_libdir}/udev/rules.d/71-seat.rules \
+                ${base_libdir}/udev/rules.d/73-seat-late.rules \
+                ${base_libdir}/udev/rules.d/99-systemd.rules \
                "
 
 FILES_${PN}-dbg += "${rootlibdir}/.debug ${systemd_unitdir}/.debug ${systemd_unitdir}/*/.debug ${base_libdir}/security/.debug/"
@@ -327,7 +327,7 @@ RRECOMMENDS_${PN} += "systemd-serialgetty systemd-vconsole-setup \
 
 PACKAGES =+ "udev-dbg udev udev-hwdb"
 
-FILES_udev-dbg += "/lib/udev/.debug"
+FILES_udev-dbg += "${base_libdir}/udev/.debug"
 
 RPROVIDES_udev = "hotplug"
 
-- 
2.1.4

Re: [PATCH 3/9] systemd: remove hard-coded paths in FILES entries

[Richard Purdie]

On Thu, 2015-09-03 at 15:58 +0100, Joshua Lock wrote:
> Signed-off-by: Joshua Lock <joshua.lock@collabora.co.uk>
> ---
>  meta/recipes-core/systemd/systemd_219.bb | 10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)
> 
> diff --git a/meta/recipes-core/systemd/systemd_219.bb b/meta/recipes-core/systemd/systemd_219.bb
> index 4d32fd5..fbd86e8 100644
> --- a/meta/recipes-core/systemd/systemd_219.bb
> +++ b/meta/recipes-core/systemd/systemd_219.bb
> @@ -306,10 +306,10 @@ FILES_${PN} = " ${base_bindir}/* \
>                  ${exec_prefix}/lib/sysctl.d \
>                  ${exec_prefix}/lib/sysusers.d \
>                  ${localstatedir} \
> -                /lib/udev/rules.d/70-uaccess.rules \
> -                /lib/udev/rules.d/71-seat.rules \
> -                /lib/udev/rules.d/73-seat-late.rules \
> -                /lib/udev/rules.d/99-systemd.rules \
> +                ${base_libdir}/udev/rules.d/70-uaccess.rules \
> +                ${base_libdir}/udev/rules.d/71-seat.rules \
> +                ${base_libdir}/udev/rules.d/73-seat-late.rules \
> +                ${base_libdir}/udev/rules.d/99-systemd.rules \
>                 "
>  
>  FILES_${PN}-dbg += "${rootlibdir}/.debug ${systemd_unitdir}/.debug ${systemd_unitdir}/*/.debug ${base_libdir}/security/.debug/"
> @@ -327,7 +327,7 @@ RRECOMMENDS_${PN} += "systemd-serialgetty systemd-vconsole-setup \
>  
>  PACKAGES =+ "udev-dbg udev udev-hwdb"
>  
> -FILES_udev-dbg += "/lib/udev/.debug"
> +FILES_udev-dbg += "${base_libdir}/udev/.debug"

From what I remember, this is intentional and this change likely breaks
multilib. You don't want two sets of rules on a multilib system.

Cheers,

Richard

Re: [PATCH 3/9] systemd: remove hard-coded paths in FILES entries

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 369 bytes --]

On 3 September 2015 at 16:12, Richard Purdie <
richard.purdie@linuxfoundation.org> wrote:

> From what I remember, this is intentional and this change likely breaks
> multilib. You don't want two sets of rules on a multilib system.
>

Indeed - you need nonarch_base_libdir or whatever it's called here I think,
as that doesn't get renamed by multilib.

Ross

[-- Attachment #2: Type: text/html, Size: 801 bytes --]

Re: [PATCH 3/9] systemd: remove hard-coded paths in FILES entries

[Joshua Lock]

On Thu, 2015-09-03 at 16:22 +0100, Burton, Ross wrote:
> 
> On 3 September 2015 at 16:12, Richard Purdie <
> richard.purdie@linuxfoundation.org> wrote:
> > From what I remember, this is intentional and this change likely
> > breaks
> > multilib. You don't want two sets of rules on a multilib system.
> > 
> Indeed - you need nonarch_base_libdir or whatever it's called here I
> think, as that doesn't get renamed by multilib.

Got it, thanks - I'm testing a v2 series with a multilib build thrown
in to the mix.

Thanks both for the feedback,

Joshua

Re: [PATCH 3/9] systemd: remove hard-coded paths in FILES entries

[Mark Hatle]

On 9/3/15 10:12 AM, Richard Purdie wrote:
> On Thu, 2015-09-03 at 15:58 +0100, Joshua Lock wrote:
>> Signed-off-by: Joshua Lock <joshua.lock@collabora.co.uk>
>> ---
>>  meta/recipes-core/systemd/systemd_219.bb | 10 +++++-----
>>  1 file changed, 5 insertions(+), 5 deletions(-)
>>
>> diff --git a/meta/recipes-core/systemd/systemd_219.bb b/meta/recipes-core/systemd/systemd_219.bb
>> index 4d32fd5..fbd86e8 100644
>> --- a/meta/recipes-core/systemd/systemd_219.bb
>> +++ b/meta/recipes-core/systemd/systemd_219.bb
>> @@ -306,10 +306,10 @@ FILES_${PN} = " ${base_bindir}/* \
>>                  ${exec_prefix}/lib/sysctl.d \
>>                  ${exec_prefix}/lib/sysusers.d \
>>                  ${localstatedir} \
>> -                /lib/udev/rules.d/70-uaccess.rules \
>> -                /lib/udev/rules.d/71-seat.rules \
>> -                /lib/udev/rules.d/73-seat-late.rules \
>> -                /lib/udev/rules.d/99-systemd.rules \
>> +                ${base_libdir}/udev/rules.d/70-uaccess.rules \
>> +                ${base_libdir}/udev/rules.d/71-seat.rules \
>> +                ${base_libdir}/udev/rules.d/73-seat-late.rules \
>> +                ${base_libdir}/udev/rules.d/99-systemd.rules \
>>                 "
>>  
>>  FILES_${PN}-dbg += "${rootlibdir}/.debug ${systemd_unitdir}/.debug ${systemd_unitdir}/*/.debug ${base_libdir}/security/.debug/"
>> @@ -327,7 +327,7 @@ RRECOMMENDS_${PN} += "systemd-serialgetty systemd-vconsole-setup \
>>  
>>  PACKAGES =+ "udev-dbg udev udev-hwdb"
>>  
>> -FILES_udev-dbg += "/lib/udev/.debug"
>> +FILES_udev-dbg += "${base_libdir}/udev/.debug"
> 
> From what I remember, this is intentional and this change likely breaks
> multilib. You don't want two sets of rules on a multilib system.

Correct.. it has to go into /lib or if everything is move to /usr, then /usr/lib.

--Mark

> Cheers,
> 
> Richard
> 
> 
> 

Re: [PATCH 3/9] systemd: remove hard-coded paths in FILES entries

[Joshua Lock]

On Thu, 2015-09-03 at 12:10 -0500, Mark Hatle wrote:
> On 9/3/15 10:12 AM, Richard Purdie wrote:
> > On Thu, 2015-09-03 at 15:58 +0100, Joshua Lock wrote:
> > > Signed-off-by: Joshua Lock <joshua.lock@collabora.co.uk>
> > > ---
> > >  meta/recipes-core/systemd/systemd_219.bb | 10 +++++-----
> > >  1 file changed, 5 insertions(+), 5 deletions(-)
> > > 
> > > diff --git a/meta/recipes-core/systemd/systemd_219.bb
> > > b/meta/recipes-core/systemd/systemd_219.bb
> > > index 4d32fd5..fbd86e8 100644
> > > --- a/meta/recipes-core/systemd/systemd_219.bb
> > > +++ b/meta/recipes-core/systemd/systemd_219.bb
> > > @@ -306,10 +306,10 @@ FILES_${PN} = " ${base_bindir}/* \
> > >                  ${exec_prefix}/lib/sysctl.d \
> > >                  ${exec_prefix}/lib/sysusers.d \
> > >                  ${localstatedir} \
> > > -                /lib/udev/rules.d/70-uaccess.rules \
> > > -                /lib/udev/rules.d/71-seat.rules \
> > > -                /lib/udev/rules.d/73-seat-late.rules \
> > > -                /lib/udev/rules.d/99-systemd.rules \
> > > +                ${base_libdir}/udev/rules.d/70-uaccess.rules \
> > > +                ${base_libdir}/udev/rules.d/71-seat.rules \
> > > +                ${base_libdir}/udev/rules.d/73-seat-late.rules \
> > > +                ${base_libdir}/udev/rules.d/99-systemd.rules \
> > >                 "
> > >  
> > >  FILES_${PN}-dbg += "${rootlibdir}/.debug
> > > ${systemd_unitdir}/.debug ${systemd_unitdir}/*/.debug
> > > ${base_libdir}/security/.debug/"
> > > @@ -327,7 +327,7 @@ RRECOMMENDS_${PN} += "systemd-serialgetty
> > > systemd-vconsole-setup \
> > >  
> > >  PACKAGES =+ "udev-dbg udev udev-hwdb"
> > >  
> > > -FILES_udev-dbg += "/lib/udev/.debug"
> > > +FILES_udev-dbg += "${base_libdir}/udev/.debug"
> > 
> > From what I remember, this is intentional and this change likely
> > breaks
> > multilib. You don't want two sets of rules on a multilib system.
> 
> Correct.. it has to go into /lib or if everything is move to /usr,
> then /usr/lib.

I've prepared a v2 that changes this to nonarch_base_libdir but
systemd+multilib appears to be broken, see:

https://bugzilla.yoctoproject.org/show_bug.cgi?id=8256

Regards,

Joshua 

Re: [PATCH 3/9] systemd: remove hard-coded paths in FILES entries

[Burton, Ross]

[-- Attachment #1: Type: text/plain, Size: 1140 bytes --]

On 4 September 2015 at 12:07, Joshua Lock <joshua.lock@collabora.co.uk>
wrote:

> I've prepared a v2 that changes this to nonarch_base_libdir but
> systemd+multilib appears to be broken, see:
>
> https://bugzilla.yoctoproject.org/show_bug.cgi?id=8256
>

As far as I understand it (and that's only a bit, I'll delegate to Mark)
this is because the sanity check is verifying that the files that get
overwritten between different subarches are actually identical, but
lib32-connman and connman have service files in /lib/systemd that refer to
binaries that are in $(libdir), which changes, so the files change.
Arguably connman should be using libexecdir and not libdir, but that also
would mean fixing libexecdir not to change (for which I have a patch series
that failed to re-send in this cycle).

I've commented on the bug that the example should be changed to be a
library such as lib32-glib-2.0, which is possibly a more useful real world
situation ("I have a 32-bit pre-compiled binary that needs glib" is more
likely than "I want a 64-bit distro with a 32-bit connman") and should
actually work.

Ross

Ross

[-- Attachment #2: Type: text/html, Size: 1711 bytes --]

[PATCH 4/9] alsa-utils: handle udev dir being 2 levels below /

[Joshua Lock]

When building with a merged /usr dir the udev directory
lives at /usr/lib/udev - update the FILES pattern to also
pick up udev files installed two levels below the / to
ensure a merged /usr works.

Signed-off-by: Joshua Lock <joshua.lock@collabora.co.uk>
---
 meta/recipes-multimedia/alsa/alsa-utils_1.0.29.bb | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta/recipes-multimedia/alsa/alsa-utils_1.0.29.bb b/meta/recipes-multimedia/alsa/alsa-utils_1.0.29.bb
index 263f38a..97fe4b2 100644
--- a/meta/recipes-multimedia/alsa/alsa-utils_1.0.29.bb
+++ b/meta/recipes-multimedia/alsa/alsa-utils_1.0.29.bb
@@ -58,7 +58,7 @@ FILES_alsa-utils-midi        = "${bindir}/aplaymidi ${bindir}/arecordmidi ${bind
 FILES_alsa-utils-aconnect    = "${bindir}/aconnect"
 FILES_alsa-utils-aseqnet     = "${bindir}/aseqnet"
 FILES_alsa-utils-iecset      = "${bindir}/iecset"
-FILES_alsa-utils-alsactl     = "${sbindir}/alsactl */udev/rules.d ${systemd_unitdir} ${localstatedir}/lib/alsa ${datadir}/alsa/init/"
+FILES_alsa-utils-alsactl     = "${sbindir}/alsactl */udev/rules.d */*/udev/rules.d ${systemd_unitdir} ${localstatedir}/lib/alsa ${datadir}/alsa/init/"
 FILES_alsa-utils-aseqdump    = "${bindir}/aseqdump"
 FILES_alsa-utils-alsaloop    = "${bindir}/alsaloop"
 FILES_alsa-utils-alsaucm     = "${bindir}/alsaucm"
-- 
2.1.4

[PATCH 5/9] pulseaudio: handle udev dir being 2 levels below /

[Joshua Lock]

When building with a merged /usr dir the udev directory
lives at /usr/lib/udev - update the FILES pattern to also
pick up udev files installed two levels below the / to
ensure a merged /usr works.

Signed-off-by: Joshua Lock <joshua.lock@collabora.co.uk>
---
 meta/recipes-multimedia/pulseaudio/pulseaudio.inc | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta/recipes-multimedia/pulseaudio/pulseaudio.inc b/meta/recipes-multimedia/pulseaudio/pulseaudio.inc
index 47f5f39..5454787 100644
--- a/meta/recipes-multimedia/pulseaudio/pulseaudio.inc
+++ b/meta/recipes-multimedia/pulseaudio/pulseaudio.inc
@@ -99,7 +99,7 @@ FILES_${PN}-dbg += "${libexecdir}/pulse/.debug \
 FILES_${PN}-dev += "${libdir}/pulse-${PV}/modules/*.la ${datadir}/vala ${libdir}/cmake"   
 FILES_${PN}-conf = "${sysconfdir}"
 FILES_${PN}-bin += "${sysconfdir}/default/volatiles/volatiles.04_pulse"
-FILES_${PN}-server = "${bindir}/pulseaudio ${bindir}/start-* ${sysconfdir} ${bindir}/pactl */udev/rules.d/*.rules ${systemd_user_unitdir}/*"
+FILES_${PN}-server = "${bindir}/pulseaudio ${bindir}/start-* ${sysconfdir} ${bindir}/pactl */udev/rules.d/*.rules */*/udev/rules.d/*.rules ${systemd_user_unitdir}/*"
 
 #SYSTEMD_PACKAGES = "${PN}-server"
 SYSTEMD_SERVICE_${PN}-server = "pulseaudio.service"
-- 
2.1.4

[PATCH 6/9] bluez: handle udev dir being 2 levels below /

[Joshua Lock]

When building with a merged /usr dir the udev directory
lives at /usr/lib/udev - update the FILES pattern to also
pick up udev files installed two levels below the / to
ensure a merged /usr works.

Signed-off-by: Joshua Lock <joshua.lock@collabora.co.uk>
---
 meta/recipes-connectivity/bluez5/bluez5.inc | 1 +
 1 file changed, 1 insertion(+)

diff --git a/meta/recipes-connectivity/bluez5/bluez5.inc b/meta/recipes-connectivity/bluez5/bluez5.inc
index 0fe5be5..039c443 100644
--- a/meta/recipes-connectivity/bluez5/bluez5.inc
+++ b/meta/recipes-connectivity/bluez5/bluez5.inc
@@ -95,6 +95,7 @@ FILES_${PN}-dbg += "\
   ${libdir}/bluetooth/plugins/.debug \
   ${libdir}/*/.debug \
   */udev/.debug \
+  */*/udev/.debug \
   "
 
 RDEPENDS_${PN}-testtools += "python python-dbus python-pygobject"
-- 
2.1.4

[PATCH 7/9] pcmciautils: handle udev dir being 2 levels below /

[Joshua Lock]

When building with a merged /usr dir the udev directory
lives at /usr/lib/udev - update the FILES pattern to also
pick up udev files installed two levels below the / to
ensure a merged /usr works.

Signed-off-by: Joshua Lock <joshua.lock@collabora.co.uk>
---
 meta/recipes-bsp/pcmciautils/pcmciautils_018.bb | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/meta/recipes-bsp/pcmciautils/pcmciautils_018.bb b/meta/recipes-bsp/pcmciautils/pcmciautils_018.bb
index 24ceed8..857bd07 100644
--- a/meta/recipes-bsp/pcmciautils/pcmciautils_018.bb
+++ b/meta/recipes-bsp/pcmciautils/pcmciautils_018.bb
@@ -9,5 +9,5 @@ SRC_URI[sha256sum] = "79e6ae441278e178c07501d492394ed2c0326fdb66894f6d040ec811b0
 
 PR = "r1"
 
-FILES_${PN}-dbg += "*/udev/.debug"
-FILES_${PN} += "*/udev"
+FILES_${PN}-dbg += "*/udev/.debug */*/udev/.debug"
+FILES_${PN} += "*/udev */*/udev"
-- 
2.1.4

[PATCH 8/9] nfs-utils: don't force use of /sbin as sbindir

[Joshua Lock]

In a system with a merged /usr /sbin isn't packaged.

Signed-off-by: Joshua Lock <joshua.lock@collabora.co.uk>
---
 meta/recipes-connectivity/nfs-utils/nfs-utils_1.3.1.bb | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/meta/recipes-connectivity/nfs-utils/nfs-utils_1.3.1.bb b/meta/recipes-connectivity/nfs-utils/nfs-utils_1.3.1.bb
index 6da8509..42101de 100644
--- a/meta/recipes-connectivity/nfs-utils/nfs-utils_1.3.1.bb
+++ b/meta/recipes-connectivity/nfs-utils/nfs-utils_1.3.1.bb
@@ -90,6 +90,14 @@ RDEPENDS_${PN}-stats = "python"
 
 FILES_${PN} += "${systemd_unitdir}"
 
+do_configure_prepend() {
+        sed -i -e 's,sbindir = /sbin,sbindir = ${base_sbindir},g' \
+            ${S}/utils/mount/Makefile.am
+
+        sed -i -e 's,sbindir = /sbin,sbindir = ${base_sbindir},g' \
+            ${S}/utils/osd_login/Makefile.am
+}
+
 # Make clean needed because the package comes with
 # precompiled 64-bit objects that break the build
 do_compile_prepend() {
-- 
2.1.4

Re: [PATCH 8/9] nfs-utils: don't force use of /sbin as sbindir

[Mark Hatle]

On 9/3/15 9:58 AM, Joshua Lock wrote:
> In a system with a merged /usr /sbin isn't packaged.

This sounds like an error to me if /sbin doesn't exist at all. It should exist
either as a link to /usr/sbin (or the reverse).. but it should be there as lots
of things expect the directory (or link) to exist.

(It's also required by the FHS, even though that probably isn't enough of a
concern.)

> Signed-off-by: Joshua Lock <joshua.lock@collabora.co.uk>
> ---
>  meta/recipes-connectivity/nfs-utils/nfs-utils_1.3.1.bb | 8 ++++++++
>  1 file changed, 8 insertions(+)
> 
> diff --git a/meta/recipes-connectivity/nfs-utils/nfs-utils_1.3.1.bb b/meta/recipes-connectivity/nfs-utils/nfs-utils_1.3.1.bb
> index 6da8509..42101de 100644
> --- a/meta/recipes-connectivity/nfs-utils/nfs-utils_1.3.1.bb
> +++ b/meta/recipes-connectivity/nfs-utils/nfs-utils_1.3.1.bb
> @@ -90,6 +90,14 @@ RDEPENDS_${PN}-stats = "python"
>  
>  FILES_${PN} += "${systemd_unitdir}"
>  
> +do_configure_prepend() {
> +        sed -i -e 's,sbindir = /sbin,sbindir = ${base_sbindir},g' \
> +            ${S}/utils/mount/Makefile.am
> +
> +        sed -i -e 's,sbindir = /sbin,sbindir = ${base_sbindir},g' \
> +            ${S}/utils/osd_login/Makefile.am
> +}
> +
>  # Make clean needed because the package comes with
>  # precompiled 64-bit objects that break the build
>  do_compile_prepend() {
> 

So my concern is simply with the commit msg -- not the actual implementation.
What you have above looks correct to me.

Re: [PATCH 8/9] nfs-utils: don't force use of /sbin as sbindir

[Joshua Lock]

On Thu, 2015-09-03 at 12:14 -0500, Mark Hatle wrote:
> On 9/3/15 9:58 AM, Joshua Lock wrote:
> > In a system with a merged /usr /sbin isn't packaged.
> 
> This sounds like an error to me if /sbin doesn't exist at all. It
> should exist
> either as a link to /usr/sbin (or the reverse).. but it should be
> there as lots
> of things expect the directory (or link) to exist.

Sure, it exists in my images but not during packaging. In my testing I
have base_bindir, base_libdir and base_sbindir set to children of /usr
in my distro.conf and a base-files_%.bbappend that creates the
symlinks. Therefore at packaging time the /sbin directory (and symlink)
 doesn't exist.

I admit this commit message is pretty poor and I'll try and remember to
fix it in a v2.

Regards,

Joshua

> (It's also required by the FHS, even though that probably isn't
> enough of a
> concern.)
> 
> > Signed-off-by: Joshua Lock <joshua.lock@collabora.co.uk>
> > ---
> >  meta/recipes-connectivity/nfs-utils/nfs-utils_1.3.1.bb | 8
> > ++++++++
> >  1 file changed, 8 insertions(+)
> > 
> > diff --git a/meta/recipes-connectivity/nfs-utils/nfs-utils_1.3.1.bb
> > b/meta/recipes-connectivity/nfs-utils/nfs-utils_1.3.1.bb
> > index 6da8509..42101de 100644
> > --- a/meta/recipes-connectivity/nfs-utils/nfs-utils_1.3.1.bb
> > +++ b/meta/recipes-connectivity/nfs-utils/nfs-utils_1.3.1.bb
> > @@ -90,6 +90,14 @@ RDEPENDS_${PN}-stats = "python"
> >  
> >  FILES_${PN} += "${systemd_unitdir}"
> >  
> > +do_configure_prepend() {
> > +        sed -i -e 's,sbindir = /sbin,sbindir = ${base_sbindir},g'
> > \
> > +            ${S}/utils/mount/Makefile.am
> > +
> > +        sed -i -e 's,sbindir = /sbin,sbindir = ${base_sbindir},g'
> > \
> > +            ${S}/utils/osd_login/Makefile.am
> > +}
> > +
> >  # Make clean needed because the package comes with
> >  # precompiled 64-bit objects that break the build
> >  do_compile_prepend() {
> > 
> 
> So my concern is simply with the commit msg -- not the actual
> implementation.
> What you have above looks correct to me.

[PATCH 9/9] procps: fix for base_sbindir == sbindir

[Joshua Lock]

An rmdir call in do_install_append was trying to remove the
sbindir, however in a system with a merged usr this directory
is not empty and therefore failing to rm it causes an ERROR.

Instead check that sbindir != base_sbindir before trying to
remove the directory.

Signed-off-by: Joshua Lock <joshua.lock@collabora.co.uk>
---
 meta/recipes-extended/procps/procps_3.3.10.bb | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/meta/recipes-extended/procps/procps_3.3.10.bb b/meta/recipes-extended/procps/procps_3.3.10.bb
index 65d64ec..dcfaba7 100644
--- a/meta/recipes-extended/procps/procps_3.3.10.bb
+++ b/meta/recipes-extended/procps/procps_3.3.10.bb
@@ -31,8 +31,9 @@ do_install_append () {
 	[ "${bindir}" != "${base_bindir}" ] && for i in ${base_bindir_progs}; do mv ${D}${bindir}/$i ${D}${base_bindir}/$i; done
 	install -d ${D}${base_sbindir}
 	[ "${sbindir}" != "${base_sbindir}" ] && for i in ${base_sbindir_progs}; do mv ${D}${sbindir}/$i ${D}${base_sbindir}/$i; done
-        # Remove now empty dir
-	rmdir ${D}/${sbindir}
+        if [ "${base_sbindir}" != "${sbindir}" ]; then
+                rmdir ${D}${sbindir}
+        fi
 
         install -d ${D}${sysconfdir}
         install -m 0644 ${WORKDIR}/sysctl.conf ${D}${sysconfdir}/sysctl.conf
-- 
2.1.4

Re: [PATCH 0/9] Fixes for a distro with a merged /usr

[Mark Hatle]

On 9/3/15 9:58 AM, Joshua Lock wrote:
> I've been playing around with building a distro with a merged /usr[1] and
> the following series fixes several issues I ran across in OE core metadata.

everything in '/' linked to '/usr/' or everything in '/usr/' linked down to '/'?

(I've seen both implementations... just curious which one you are trying.)

--Mark

> These have been build tested by building a core-image-sato image with
> ipk packaging for qemux86 with both a standard/default configuration and
> for my merged /usr test distro.
> 
> I used buildhistory to compare core-image-sato before and after this series
> and the only change is:
> 
> Changes to images/qemux86/glibc/core-image-sato (files-in-image.txt):
>   /bin/lsmod.kmod changed symlink target from ../bin/kmod to kmod
> 
> Regards,
> 
> Joshua
> 
> 1. https://wiki.freedesktop.org/www/Software/systemd/TheCaseForTheUsrMerge/
> 
> The following changes since commit bdeb32b4cdbe316f17c2fd854d59e05e8f2e8ffc:
> 
>   rt-tests: drop unnecessary added-missing-dependencies.patch (2015-09-01 11:43:38 +0100)
> 
> are available in the git repository at:
> 
>   git://git.openembedded.org/openembedded-core-contrib joshuagl/merged-usr
>   http://cgit.openembedded.org/cgit.cgi/openembedded-core-contrib/log/?h=joshuagl/merged-usr
> 
> Joshua Lock (9):
>   kmod: fix link creation when base_bindir != /bin
>   busybox: fixes for when base_bindir != /bin
>   systemd: remove hard-coded paths in FILES entries
>   alsa-utils: handle udev dir being 2 levels below /
>   pulseaudio: handle udev dir being 2 levels below /
>   bluez: handle udev dir being 2 levels below /
>   pcmciautils: handle udev dir being 2 levels below /
>   nfs-utils: don't force use of /sbin as sbindir
>   procps: fix for base_sbindir == sbindir
> 
>  meta/recipes-bsp/pcmciautils/pcmciautils_018.bb    |  4 +-
>  meta/recipes-connectivity/bluez5/bluez5.inc        |  1 +
>  .../nfs-utils/nfs-utils_1.3.1.bb                   |  8 ++++
>  meta/recipes-core/busybox/busybox.inc              | 49 ++++++++++++----------
>  meta/recipes-core/systemd/systemd_219.bb           | 10 ++---
>  meta/recipes-extended/procps/procps_3.3.10.bb      |  5 ++-
>  meta/recipes-kernel/kmod/kmod_git.bb               |  4 +-
>  meta/recipes-multimedia/alsa/alsa-utils_1.0.29.bb  |  2 +-
>  meta/recipes-multimedia/pulseaudio/pulseaudio.inc  |  2 +-
>  9 files changed, 49 insertions(+), 36 deletions(-)
> 

Re: [PATCH 0/9] Fixes for a distro with a merged /usr

[Joshua Lock]

On Thu, 2015-09-03 at 12:11 -0500, Mark Hatle wrote:
> On 9/3/15 9:58 AM, Joshua Lock wrote:
> > I've been playing around with building a distro with a merged
> > /usr[1] and
> > the following series fixes several issues I ran across in OE core
> > metadata.
> 
> everything in '/' linked to '/usr/' or everything in '/usr/' linked
> down to '/'?
> 
> (I've seen both implementations... just curious which one you are
> trying.)

The latter, as per [1]. That is /bin, /lib and /sbin are symlinked into
/usr.

Cheers,

Joshua

1. https://wiki.freedesktop.org/www/Software/systemd/TheCaseForTheUsrMe
rge/