[PATCH 0/3] Patches pending on O.S. Systems tree

[PATCH 0/3] Patches pending on O.S. Systems tree

[Otavio Salvador]

The following changes since commit 0a496df0209c93fd00ea929b5f27faa1a6e600c0:

  clutter-1.6: Add patch to update gettext macro version (2011-06-01 18:32:30 +0100)

are available in the git repository at:
  git://github.com/OSSystems/oe-core master
  https://github.com/OSSystems/oe-core/tree/master

Otavio Salvador (3):
  busybox: enable mdev by default
  qmake_base.bbclass: fix lrelease/lupdate binary names
  gnutls: link against 'dl' library

 meta/classes/qmake_base.bbclass                    |    5 +++++
 meta/recipes-core/busybox/busybox-1.18.4/defconfig |    2 +-
 meta/recipes-core/busybox/busybox.inc              |    1 -
 meta/recipes-core/busybox/busybox_1.18.4.bb        |    2 +-
 meta/recipes-support/gnutls/gnutls.inc             |    4 +++-
 meta/recipes-support/gnutls/gnutls_2.12.5.bb       |    2 +-
 6 files changed, 11 insertions(+), 5 deletions(-)

-- 
1.7.2.5

[PATCH 1/3] busybox: enable mdev by default

[Otavio Salvador]

Since we do not require a configuration file by default we don't force
it's addition on the package. If the a layer wants to have it enabled
it should also set the CONFFILES for busybox-mdev package.

Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>
---
 meta/recipes-core/busybox/busybox-1.18.4/defconfig |    2 +-
 meta/recipes-core/busybox/busybox.inc              |    1 -
 meta/recipes-core/busybox/busybox_1.18.4.bb        |    2 +-
 3 files changed, 2 insertions(+), 3 deletions(-)

diff --git a/meta/recipes-core/busybox/busybox-1.18.4/defconfig b/meta/recipes-core/busybox/busybox-1.18.4/defconfig
index 5327a64..6b48cc1 100644
--- a/meta/recipes-core/busybox/busybox-1.18.4/defconfig
+++ b/meta/recipes-core/busybox/busybox-1.18.4/defconfig
@@ -518,7 +518,7 @@ CONFIG_FEATURE_HWCLOCK_ADJTIME_FHS=y
 CONFIG_LOSETUP=y
 # CONFIG_LSPCI is not set
 # CONFIG_LSUSB is not set
-# CONFIG_MDEV is not set
+CONFIG_MDEV=y
 # CONFIG_FEATURE_MDEV_CONF is not set
 # CONFIG_FEATURE_MDEV_RENAME is not set
 # CONFIG_FEATURE_MDEV_RENAME_REGEXP is not set
diff --git a/meta/recipes-core/busybox/busybox.inc b/meta/recipes-core/busybox/busybox.inc
index 86fbdae..9b874c5 100644
--- a/meta/recipes-core/busybox/busybox.inc
+++ b/meta/recipes-core/busybox/busybox.inc
@@ -29,7 +29,6 @@ INITSCRIPT_PARAMS_${PN}-mdev = "start 06 S ."
 INITSCRIPT_NAME_${PN}-udhcpd = "busybox-udhcpd" 
 INITSCRIPT_NAME_${PN}-udhcpc = "busybox-udhcpc" 
 CONFFILES_${PN}-syslog = "${sysconfdir}/syslog.conf.${PN}"
-CONFFILES_${PN}-mdev = "${sysconfdir}/mdev.conf"
 
 RRECOMMENDS_${PN} = "${PN}-syslog ${PN}-udhcpc"
 
diff --git a/meta/recipes-core/busybox/busybox_1.18.4.bb b/meta/recipes-core/busybox/busybox_1.18.4.bb
index a5080d5..a999490 100644
--- a/meta/recipes-core/busybox/busybox_1.18.4.bb
+++ b/meta/recipes-core/busybox/busybox_1.18.4.bb
@@ -1,5 +1,5 @@
 require busybox.inc
-PR = "r2"
+PR = "r3"
 
 SRC_URI = "http://www.busybox.net/downloads/busybox-${PV}.tar.bz2;name=tarball \
            file://udhcpscript.patch \
-- 
1.7.2.5

[PATCH 2/3] qmake_base.bbclass: fix lrelease/lupdate binary names

[Otavio Salvador]

To support translation, qmake based projects usually call lrelease and
lupdate however OE changes the binary names so this needs some mangle
to work out of box.

Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>
---
 meta/classes/qmake_base.bbclass |    5 +++++
 1 files changed, 5 insertions(+), 0 deletions(-)

diff --git a/meta/classes/qmake_base.bbclass b/meta/classes/qmake_base.bbclass
index a054efd..165d689 100644
--- a/meta/classes/qmake_base.bbclass
+++ b/meta/classes/qmake_base.bbclass
@@ -92,6 +92,11 @@ qmake_base_do_configure() {
 		bbnote "qmake prevar substitution: ${EXTRA_QMAKEVARS_PRE}"
 	fi
 
+	# Hack .pro files to use OE utilities
+	find -name '*.pro' \
+	     -exec sed -i -e 's,=\s*.*/lrelease,= ${OE_QMAKE_LRELEASE},g' \
+	                  -e 's,=\s*.*/lupdate,= ${OE_QMAKE_LUPDATE},g' '{}' ';'
+
 #bbnote "Calling '${OE_QMAKE_QMAKE} -makefile -spec ${QMAKESPEC} -o Makefile $QMAKE_VARSUBST_PRE $AFTER $PROFILES $QMAKE_VARSUBST_POST'"
 	unset QMAKESPEC || true
 	${OE_QMAKE_QMAKE} -makefile -spec ${QMAKESPEC} -o Makefile $QMAKE_VARSUBST_PRE $AFTER $PROFILES $QMAKE_VARSUBST_POST || die "Error calling ${OE_QMAKE_QMAKE} on $PROFILES"
-- 
1.7.2.5

[PATCH 3/3] gnutls: link against 'dl' library

[Otavio Salvador]

Fixes a build failure due missing references to 'dlopen' and
like. This started to happen after update to 2.12.5 at
ad52eacebdc413aa351de078ee878182a9c71e16. This also fixes the recipe
to use the INC_PR value otherwise we end not bumping PR if changing
only the '.inc' file.

Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>
---
 meta/recipes-support/gnutls/gnutls.inc       |    4 +++-
 meta/recipes-support/gnutls/gnutls_2.12.5.bb |    2 +-
 2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/meta/recipes-support/gnutls/gnutls.inc b/meta/recipes-support/gnutls/gnutls.inc
index 9f8d81b..9452190 100644
--- a/meta/recipes-support/gnutls/gnutls.inc
+++ b/meta/recipes-support/gnutls/gnutls.inc
@@ -3,7 +3,7 @@ HOMEPAGE = "http://www.gnu.org/software/gnutls/"
 BUGTRACKER = "https://savannah.gnu.org/support/?group=gnutls"
 DEPENDS = "zlib lzo libtasn1 libgcrypt (>= 1.4.2) libcap"
 
-INC_PR = "r1"
+INC_PR = "r2"
 
 LICENSE = "GPLv3+ & LGPLv2.1+"
 LICENSE_${PN} = "LGPLv2.1+"
@@ -30,6 +30,8 @@ do_configure_prepend() {
 	done
 }
 
+export LDFLAGS += "-ldl"
+
 PACKAGES =+ "${PN}-openssl ${PN}-extra ${PN}-bin ${PN}-xx"
 
 FILES_${PN} = "${libdir}/libgnutls.so.*"
diff --git a/meta/recipes-support/gnutls/gnutls_2.12.5.bb b/meta/recipes-support/gnutls/gnutls_2.12.5.bb
index b27a065..9d3f5ab 100644
--- a/meta/recipes-support/gnutls/gnutls_2.12.5.bb
+++ b/meta/recipes-support/gnutls/gnutls_2.12.5.bb
@@ -1,6 +1,6 @@
 require gnutls.inc
 
-PR = "r0"
+PR = "${INC_PR}.0"
 
 SRC_URI += "file://gnutls-openssl.patch \
             file://configure-fix.patch"
-- 
1.7.2.5

Re: [PATCH 3/3] gnutls: link against 'dl' library

[Phil Blundell]

On Wed, 2011-06-01 at 20:09 +0000, Otavio Salvador wrote:
> +export LDFLAGS += "-ldl"

The configure script ought to be figuring this out for itself.  Do you
know why that isn't working?

p.

Re: [PATCH 1/3] busybox: enable mdev by default

[Phil Blundell]

On Wed, 2011-06-01 at 20:09 +0000, Otavio Salvador wrote:
> -# CONFIG_MDEV is not set
> +CONFIG_MDEV=y

Per previous discussion, I am still uneasy about this change.  I think
we really need some sort of coherent policy for what exactly the default
busybox configuration in oe-core is meant to be doing, and then (if
necessary) a set of patches to make it match the policy.  Just flipping
random features on and off does not seem like a good way to proceed.

p.

Re: [PATCH 3/3] gnutls: link against 'dl' library

[Otavio Salvador]

On Wed, Jun 1, 2011 at 20:33, Phil Blundell <pb@pbcl.net> wrote:
> On Wed, 2011-06-01 at 20:09 +0000, Otavio Salvador wrote:
>> +export LDFLAGS += "-ldl"
>
> The configure script ought to be figuring this out for itself.  Do you
> know why that isn't working?

No I don't.

Are you able to reproduce it?

-- 
Otavio Salvador                             O.S. Systems
E-mail: otavio@ossystems.com.br  http://www.ossystems.com.br
Mobile: +55 53 9981-7854              http://projetos.ossystems.com.br

Re: [PATCH 1/3] busybox: enable mdev by default

[Otavio Salvador]

On Wed, Jun 1, 2011 at 20:37, Phil Blundell <pb@pbcl.net> wrote:
> On Wed, 2011-06-01 at 20:09 +0000, Otavio Salvador wrote:
>> -# CONFIG_MDEV is not set
>> +CONFIG_MDEV=y
>
> Per previous discussion, I am still uneasy about this change.  I think
> we really need some sort of coherent policy for what exactly the default
> busybox configuration in oe-core is meant to be doing, and then (if
> necessary) a set of patches to make it match the policy.  Just flipping
> random features on and off does not seem like a good way to proceed.

OE-core has support to mdev as device handling mechanism as such this
ought to be enabled by default IMO.

Personally it doesn't matter since I have already overriden it in my
internal layer.

-- 
Otavio Salvador                             O.S. Systems
E-mail: otavio@ossystems.com.br  http://www.ossystems.com.br
Mobile: +55 53 9981-7854              http://projetos.ossystems.com.br

Re: [PATCH 3/3] gnutls: link against 'dl' library

[Phil Blundell]

On Wed, 2011-06-01 at 20:39 +0000, Otavio Salvador wrote:
> On Wed, Jun 1, 2011 at 20:33, Phil Blundell <pb@pbcl.net> wrote:
> > On Wed, 2011-06-01 at 20:09 +0000, Otavio Salvador wrote:
> >> +export LDFLAGS += "-ldl"
> >
> > The configure script ought to be figuring this out for itself.  Do you
> > know why that isn't working?
> 
> No I don't.
> 
> Are you able to reproduce it?

No, it works fine for me.  I get:

checking how to link with libpthread... -lpthread
checking for libdl... yes
checking how to link with libdl... -ldl
checking whether we are using the GNU C++ compiler... yes

This is with the current head of oe-core on qemuarm.

p.

Re: [PATCH 3/3] gnutls: link against 'dl' library

[Otavio Salvador]

On Wed, Jun 1, 2011 at 20:53, Phil Blundell <pb@pbcl.net> wrote:
> On Wed, 2011-06-01 at 20:39 +0000, Otavio Salvador wrote:
>> Are you able to reproduce it?
>
> No, it works fine for me.  I get:
>
> checking how to link with libpthread... -lpthread
> checking for libdl... yes
> checking how to link with libdl... -ldl
> checking whether we are using the GNU C++ compiler... yes
>
> This is with the current head of oe-core on qemuarm.

I am building for i586 but I doubt it is related.

I get:

checking for libpthread... yes
checking how to link with libpthread... /usr/lib/libpthread.so
checking for libdl... no
checking whether we are using the GNU C++ compiler... yes

-- 
Otavio Salvador                             O.S. Systems
E-mail: otavio@ossystems.com.br  http://www.ossystems.com.br
Mobile: +55 53 9981-7854              http://projetos.ossystems.com.br

Re: [PATCH 3/3] gnutls: link against 'dl' library

[Saul Wold]

On 06/01/2011 02:00 PM, Otavio Salvador wrote:
> On Wed, Jun 1, 2011 at 20:53, Phil Blundell<pb@pbcl.net>  wrote:
>> On Wed, 2011-06-01 at 20:39 +0000, Otavio Salvador wrote:
>>> Are you able to reproduce it?
>>
>> No, it works fine for me.  I get:
>>
>> checking how to link with libpthread... -lpthread
>> checking for libdl... yes
>> checking how to link with libdl... -ldl
>> checking whether we are using the GNU C++ compiler... yes
>>
>> This is with the current head of oe-core on qemuarm.
>
> I am building for i586 but I doubt it is related.
>
> I get:
>
> checking for libpthread... yes
> checking how to link with libpthread... /usr/lib/libpthread.so
> checking for libdl... no
> checking whether we are using the GNU C++ compiler... yes
>
I also built for x86 and beagle, both have libdl, can you check that you 
do not have some kind of site configuration that is overriding, also 
maybe post your log from .../gnutls-2.12.5/lib/config.log

Thanks
	Sau!

Re: [PATCH 3/3] gnutls: link against 'dl' library

[Martin Jansa]

[-- Attachment #1: Type: text/plain, Size: 1626 bytes --]

On Wed, Jun 01, 2011 at 02:43:53PM -0700, Saul Wold wrote:
> On 06/01/2011 02:00 PM, Otavio Salvador wrote:
> > On Wed, Jun 1, 2011 at 20:53, Phil Blundell<pb@pbcl.net>  wrote:
> >> On Wed, 2011-06-01 at 20:39 +0000, Otavio Salvador wrote:
> >>> Are you able to reproduce it?
> >>
> >> No, it works fine for me.  I get:
> >>
> >> checking how to link with libpthread... -lpthread
> >> checking for libdl... yes
> >> checking how to link with libdl... -ldl
> >> checking whether we are using the GNU C++ compiler... yes
> >>
> >> This is with the current head of oe-core on qemuarm.
> >
> > I am building for i586 but I doubt it is related.
> >
> > I get:
> >
> > checking for libpthread... yes
> > checking how to link with libpthread... /usr/lib/libpthread.so
> > checking for libdl... no
> > checking whether we are using the GNU C++ compiler... yes
> >
> I also built for x86 and beagle, both have libdl, can you check that you 
> do not have some kind of site configuration that is overriding, also 
> maybe post your log from .../gnutls-2.12.5/lib/config.log

here it tries to link against host libdl

configure:26050: checking for libdl
configure:26069: arm-oe-linux-gnueabi-gcc -march=armv4t -mtune=arm920t
--sysroot=/OE/shr-core/tmp/sysroots/om-gta02 -o conftest -O2 -pipe -g
-feliminate-unused-debug-types
-I/OE/shr-core/tmp/sysroots/om-gta02/usr/include -Wl,-O1
-Wl,--as-needed conftest.c  /usr/lib/libdl.so >&5
/usr/lib/libdl.so: file not recognized: File format not recognized
collect2: ld returned 1 exit status

-- 
Martin 'JaMa' Jansa     jabber: Martin.Jansa@gmail.com

Re: [PATCH 3/3] gnutls: link against 'dl' library

[Saul Wold]

On 06/01/2011 02:52 PM, Martin Jansa wrote:
> On Wed, Jun 01, 2011 at 02:43:53PM -0700, Saul Wold wrote:
>> On 06/01/2011 02:00 PM, Otavio Salvador wrote:
>>> On Wed, Jun 1, 2011 at 20:53, Phil Blundell<pb@pbcl.net>   wrote:
>>>> On Wed, 2011-06-01 at 20:39 +0000, Otavio Salvador wrote:
>>>>> Are you able to reproduce it?
>>>>
>>>> No, it works fine for me.  I get:
>>>>
>>>> checking how to link with libpthread... -lpthread
>>>> checking for libdl... yes
>>>> checking how to link with libdl... -ldl
>>>> checking whether we are using the GNU C++ compiler... yes
>>>>
>>>> This is with the current head of oe-core on qemuarm.
>>>
>>> I am building for i586 but I doubt it is related.
>>>
>>> I get:
>>>
>>> checking for libpthread... yes
>>> checking how to link with libpthread... /usr/lib/libpthread.so
>>> checking for libdl... no
>>> checking whether we are using the GNU C++ compiler... yes
>>>
>> I also built for x86 and beagle, both have libdl, can you check that you
>> do not have some kind of site configuration that is overriding, also
>> maybe post your log from .../gnutls-2.12.5/lib/config.log
>
> here it tries to link against host libdl
>
> configure:26050: checking for libdl
> configure:26069: arm-oe-linux-gnueabi-gcc -march=armv4t -mtune=arm920t
> --sysroot=/OE/shr-core/tmp/sysroots/om-gta02 -o conftest -O2 -pipe -g
> -feliminate-unused-debug-types
> -I/OE/shr-core/tmp/sysroots/om-gta02/usr/include -Wl,-O1
> -Wl,--as-needed conftest.c  /usr/lib/libdl.so>&5
> /usr/lib/libdl.so: file not recognized: File format not recognized
> collect2: ld returned 1 exit status
>
Well that's interesting, since I see it linking using -ldl, there must 
be some other setting that's affecting this, I have been looking in 
gnutls to see if I can find something else that sets libdl.so vs -ldl,
I have not found it.

configure:26075: i586-poky-linux-gcc -march=i586 
--sysroot=/builds/world/tmp/sysroots/qemux86 -o conftest -O2 -pipe -g 
-feliminate-unused-debug-types 
-I/builds/world/tmp/sysroots/qemux86/usr/include -Wl,-O1 
-Wl,--as-needed conftest.c  -ldl >&5
configure:26075: $? = 0
configure:26085: result: yes
configure:26092: checking how to link with libdl
configure:26094: result: -ldl

Sau!

>
>
>
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core

Re: [PATCH 3/3] gnutls: link against 'dl' library

[Martin Jansa]

[-- Attachment #1: Type: text/plain, Size: 3307 bytes --]

On Wed, Jun 01, 2011 at 03:19:18PM -0700, Saul Wold wrote:
> On 06/01/2011 02:52 PM, Martin Jansa wrote:
> > On Wed, Jun 01, 2011 at 02:43:53PM -0700, Saul Wold wrote:
> >> On 06/01/2011 02:00 PM, Otavio Salvador wrote:
> >>> On Wed, Jun 1, 2011 at 20:53, Phil Blundell<pb@pbcl.net>   wrote:
> >>>> On Wed, 2011-06-01 at 20:39 +0000, Otavio Salvador wrote:
> >>>>> Are you able to reproduce it?
> >>>>
> >>>> No, it works fine for me.  I get:
> >>>>
> >>>> checking how to link with libpthread... -lpthread
> >>>> checking for libdl... yes
> >>>> checking how to link with libdl... -ldl
> >>>> checking whether we are using the GNU C++ compiler... yes
> >>>>
> >>>> This is with the current head of oe-core on qemuarm.
> >>>
> >>> I am building for i586 but I doubt it is related.
> >>>
> >>> I get:
> >>>
> >>> checking for libpthread... yes
> >>> checking how to link with libpthread... /usr/lib/libpthread.so
> >>> checking for libdl... no
> >>> checking whether we are using the GNU C++ compiler... yes
> >>>
> >> I also built for x86 and beagle, both have libdl, can you check that you
> >> do not have some kind of site configuration that is overriding, also
> >> maybe post your log from .../gnutls-2.12.5/lib/config.log
> >
> > here it tries to link against host libdl
> >
> > configure:26050: checking for libdl
> > configure:26069: arm-oe-linux-gnueabi-gcc -march=armv4t -mtune=arm920t
> > --sysroot=/OE/shr-core/tmp/sysroots/om-gta02 -o conftest -O2 -pipe -g
> > -feliminate-unused-debug-types
> > -I/OE/shr-core/tmp/sysroots/om-gta02/usr/include -Wl,-O1
> > -Wl,--as-needed conftest.c  /usr/lib/libdl.so>&5
> > /usr/lib/libdl.so: file not recognized: File format not recognized
> > collect2: ld returned 1 exit status
> >
> Well that's interesting, since I see it linking using -ldl, there must 
> be some other setting that's affecting this, I have been looking in 
> gnutls to see if I can find something else that sets libdl.so vs -ldl,
> I have not found it.
> 
> configure:26075: i586-poky-linux-gcc -march=i586 
> --sysroot=/builds/world/tmp/sysroots/qemux86 -o conftest -O2 -pipe -g 
> -feliminate-unused-debug-types 
> -I/builds/world/tmp/sysroots/qemux86/usr/include -Wl,-O1 
> -Wl,--as-needed conftest.c  -ldl >&5
> configure:26075: $? = 0
> configure:26085: result: yes
> configure:26092: checking how to link with libdl
> configure:26094: result: -ldl

what do you have in configlog for LIBDL?

here it's all empty
LIBDL=''
LIBDL_PREFIX=''
LTLIBDL=''

if I add 
+             --with-libdl-prefix=${STAGING_DIR_HOST}${prefix} \
+             --with-libpthread-prefix=${STAGING_DIR_HOST}${prefix} \
to EXTRA_OE_CONF (we already have couple of --with-*-prefix there) it's
found correctly (and this is at least better workarround then forcing
-ldl directly to LDFLAGS):
LIBDL='/OE/shr-core/tmp/sysroots/om-gta02/usr/lib/libdl.so'
LIBDL_PREFIX='/OE/shr-core/tmp/sysroots/om-gta02/usr'
LTLIBDL='-L/OE/shr-core/tmp/sysroots/om-gta02/usr/lib -ldl'

same problem is with libpthread..

maybe their m4/lib-link.m4 does something wrong in
AC_DEFUN([AC_LIB_HAVE_LINKFLAGS],
[
  AC_REQUIRE([AC_LIB_PREPARE_PREFIX])
  AC_REQUIRE([AC_LIB_RPATH])

Regards,
-- 
Martin 'JaMa' Jansa     jabber: Martin.Jansa@gmail.com

Re: [PATCH 3/3] gnutls: link against 'dl' library

[Koen Kooi]

Op 2 jun 2011, om 08:31 heeft Martin Jansa het volgende geschreven:

> On Wed, Jun 01, 2011 at 03:19:18PM -0700, Saul Wold wrote:
>> On 06/01/2011 02:52 PM, Martin Jansa wrote:
>>> On Wed, Jun 01, 2011 at 02:43:53PM -0700, Saul Wold wrote:
>>>> On 06/01/2011 02:00 PM, Otavio Salvador wrote:
>>>>> On Wed, Jun 1, 2011 at 20:53, Phil Blundell<pb@pbcl.net>   wrote:
>>>>>> On Wed, 2011-06-01 at 20:39 +0000, Otavio Salvador wrote:
>>>>>>> Are you able to reproduce it?
>>>>>> 
>>>>>> No, it works fine for me.  I get:
>>>>>> 
>>>>>> checking how to link with libpthread... -lpthread
>>>>>> checking for libdl... yes
>>>>>> checking how to link with libdl... -ldl
>>>>>> checking whether we are using the GNU C++ compiler... yes
>>>>>> 
>>>>>> This is with the current head of oe-core on qemuarm.
>>>>> 
>>>>> I am building for i586 but I doubt it is related.
>>>>> 
>>>>> I get:
>>>>> 
>>>>> checking for libpthread... yes
>>>>> checking how to link with libpthread... /usr/lib/libpthread.so
>>>>> checking for libdl... no
>>>>> checking whether we are using the GNU C++ compiler... yes
>>>>> 
>>>> I also built for x86 and beagle, both have libdl, can you check that you
>>>> do not have some kind of site configuration that is overriding, also
>>>> maybe post your log from .../gnutls-2.12.5/lib/config.log
>>> 
>>> here it tries to link against host libdl
>>> 
>>> configure:26050: checking for libdl
>>> configure:26069: arm-oe-linux-gnueabi-gcc -march=armv4t -mtune=arm920t
>>> --sysroot=/OE/shr-core/tmp/sysroots/om-gta02 -o conftest -O2 -pipe -g
>>> -feliminate-unused-debug-types
>>> -I/OE/shr-core/tmp/sysroots/om-gta02/usr/include -Wl,-O1
>>> -Wl,--as-needed conftest.c  /usr/lib/libdl.so>&5
>>> /usr/lib/libdl.so: file not recognized: File format not recognized
>>> collect2: ld returned 1 exit status
>>> 
>> Well that's interesting, since I see it linking using -ldl, there must 
>> be some other setting that's affecting this, I have been looking in 
>> gnutls to see if I can find something else that sets libdl.so vs -ldl,
>> I have not found it.
>> 
>> configure:26075: i586-poky-linux-gcc -march=i586 
>> --sysroot=/builds/world/tmp/sysroots/qemux86 -o conftest -O2 -pipe -g 
>> -feliminate-unused-debug-types 
>> -I/builds/world/tmp/sysroots/qemux86/usr/include -Wl,-O1 
>> -Wl,--as-needed conftest.c  -ldl >&5
>> configure:26075: $? = 0
>> configure:26085: result: yes
>> configure:26092: checking how to link with libdl
>> configure:26094: result: -ldl
> 
> what do you have in configlog for LIBDL?
> 
> here it's all empty
> LIBDL=''
> LIBDL_PREFIX=''
> LTLIBDL=''

I get for cross gnutls in lib/config.log:

grep LIBDL lib/config.log 
HAVE_LIBDL='no'
LIBDL=''
LIBDL_PREFIX=''
LTLIBDL=''

regards,

Koen

Re: [PATCH 3/3] gnutls: link against 'dl' library

[Richard Purdie]

On Thu, 2011-06-02 at 08:31 +0200, Martin Jansa wrote:
> what do you have in configlog for LIBDL?
> 
> here it's all empty
> LIBDL=''
> LIBDL_PREFIX=''
> LTLIBDL=''
> 
> if I add 
> +             --with-libdl-prefix=${STAGING_DIR_HOST}${prefix} \
> +             --with-libpthread-prefix=${STAGING_DIR_HOST}${prefix} \
> to EXTRA_OE_CONF (we already have couple of --with-*-prefix there) it's
> found correctly (and this is at least better workarround then forcing
> -ldl directly to LDFLAGS):
> LIBDL='/OE/shr-core/tmp/sysroots/om-gta02/usr/lib/libdl.so'
> LIBDL_PREFIX='/OE/shr-core/tmp/sysroots/om-gta02/usr'
> LTLIBDL='-L/OE/shr-core/tmp/sysroots/om-gta02/usr/lib -ldl'
> 
> same problem is with libpthread..
> 
> maybe their m4/lib-link.m4 does something wrong in
> AC_DEFUN([AC_LIB_HAVE_LINKFLAGS],
> [
>   AC_REQUIRE([AC_LIB_PREPARE_PREFIX])
>   AC_REQUIRE([AC_LIB_RPATH])

I just looked at the code in lib/configure and concluded that its just
luck whether it figures out the correct values or does something really
nasty. Its poking about ${libdir} for this stuff so you end up with
something like the following combinations:

a) 64 bit system with 64 bit libs in /usr/lib64 and no 32 bit libs:
     Finds no -ldl

   Something else may or may not link it indirectly so the build may or 
   may not work.

b) 64 bit system with 64 bit libs in /usr/lib64 and 32 bit libs in /usr/lib, building for 32 bit:
     Finds a -ldl which "works".

   Things appear to work.

c) Other variations


The correct fix would appear to be to set the paths specifically using
the options Martin mentions above.

Cheers,

Richard

[PATCH] gnutls: add --with-libdl-prefix and --with-libpthread-prefix

[Martin Jansa]

Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
---
 meta/recipes-support/gnutls/gnutls.inc |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/meta/recipes-support/gnutls/gnutls.inc b/meta/recipes-support/gnutls/gnutls.inc
index 9f8d81b..03aed6a 100644
--- a/meta/recipes-support/gnutls/gnutls.inc
+++ b/meta/recipes-support/gnutls/gnutls.inc
@@ -22,6 +22,8 @@ inherit autotools binconfig pkgconfig gettext
 EXTRA_OECONF="--with-included-opencdk --with-included-libcfg --disable-rpath \
               --with-libtasn1-prefix=${STAGING_DIR_HOST}${prefix} \
               --with-libgcrypt --with-libgcrypt-prefix=${STAGING_DIR_HOST}${prefix} \
+              --with-libdl-prefix=${STAGING_DIR_HOST}${prefix} \
+              --with-libpthread-prefix=${STAGING_DIR_HOST}${prefix} \
               --with-lzo --disable-guile \
               "
 do_configure_prepend() {
-- 
1.7.5.3

Re: [PATCH] gnutls: add --with-libdl-prefix and --with-libpthread-prefix

[Richard Purdie]

On Thu, 2011-06-02 at 13:00 +0200, Martin Jansa wrote:
> Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
> ---
>  meta/recipes-support/gnutls/gnutls.inc |    2 ++
>  1 files changed, 2 insertions(+), 0 deletions(-)
> 
> diff --git a/meta/recipes-support/gnutls/gnutls.inc b/meta/recipes-support/gnutls/gnutls.inc
> index 9f8d81b..03aed6a 100644
> --- a/meta/recipes-support/gnutls/gnutls.inc
> +++ b/meta/recipes-support/gnutls/gnutls.inc
> @@ -22,6 +22,8 @@ inherit autotools binconfig pkgconfig gettext
>  EXTRA_OECONF="--with-included-opencdk --with-included-libcfg --disable-rpath \
>                --with-libtasn1-prefix=${STAGING_DIR_HOST}${prefix} \
>                --with-libgcrypt --with-libgcrypt-prefix=${STAGING_DIR_HOST}${prefix} \
> +              --with-libdl-prefix=${STAGING_DIR_HOST}${prefix} \
> +              --with-libpthread-prefix=${STAGING_DIR_HOST}${prefix} \
>                --with-lzo --disable-guile \
>                "
>  do_configure_prepend() {

Merged to master, thanks.

Richard

Re: [PATCH 1/3] busybox: enable mdev by default

[Richard Purdie]

On Wed, 2011-06-01 at 20:40 +0000, Otavio Salvador wrote:
> On Wed, Jun 1, 2011 at 20:37, Phil Blundell <pb@pbcl.net> wrote:
> > On Wed, 2011-06-01 at 20:09 +0000, Otavio Salvador wrote:
> >> -# CONFIG_MDEV is not set
> >> +CONFIG_MDEV=y
> >
> > Per previous discussion, I am still uneasy about this change.  I think
> > we really need some sort of coherent policy for what exactly the default
> > busybox configuration in oe-core is meant to be doing, and then (if
> > necessary) a set of patches to make it match the policy.  Just flipping
> > random features on and off does not seem like a good way to proceed.
> 
> OE-core has support to mdev as device handling mechanism as such this
> ought to be enabled by default IMO.
> 
> Personally it doesn't matter since I have already overriden it in my
> internal layer.

I'm afraid I'm with Phil on this. I don't like the idea of enabling
something we don't actually use. This really needs to become some kind
of configure option which would at the same time disable/replace udev so
the patch in its currently form isn't acceptable.

Cheers,

Richard

Re: [PATCH 1/3] busybox: enable mdev by default

[Otavio Salvador]

On Thu, Jun 2, 2011 at 16:37, Richard Purdie
<richard.purdie@linuxfoundation.org> wrote:
> I'm afraid I'm with Phil on this. I don't like the idea of enabling
> something we don't actually use. This really needs to become some kind
> of configure option which would at the same time disable/replace udev so
> the patch in its currently form isn't acceptable.

I'll drop it from my tree since I am using an bbappend on our local layer.

Cheers,

-- 
Otavio Salvador                             O.S. Systems
E-mail: otavio@ossystems.com.br  http://www.ossystems.com.br
Mobile: +55 53 9981-7854              http://projetos.ossystems.com.br

Re: [PATCH 2/3] qmake_base.bbclass: fix lrelease/lupdate binary names

[Paul Eggleton]

On Wednesday 01 June 2011 21:09:54 Otavio Salvador wrote:
> To support translation, qmake based projects usually call lrelease and
> lupdate however OE changes the binary names so this needs some mangle
> to work out of box.
> 
> Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>
> ---
>  meta/classes/qmake_base.bbclass |    5 +++++
>  1 files changed, 5 insertions(+), 0 deletions(-)
> 
> diff --git a/meta/classes/qmake_base.bbclass
> b/meta/classes/qmake_base.bbclass index a054efd..165d689 100644
> --- a/meta/classes/qmake_base.bbclass
> +++ b/meta/classes/qmake_base.bbclass
> @@ -92,6 +92,11 @@ qmake_base_do_configure() {
>  		bbnote "qmake prevar substitution: ${EXTRA_QMAKEVARS_PRE}"
>  	fi
> 
> +	# Hack .pro files to use OE utilities
> +	find -name '*.pro' \
> +	     -exec sed -i -e 's,=\s*.*/lrelease,= ${OE_QMAKE_LRELEASE},g' \
> +	                  -e 's,=\s*.*/lupdate,= ${OE_QMAKE_LUPDATE},g' '{}' ';'
> +
>  #bbnote "Calling '${OE_QMAKE_QMAKE} -makefile -spec ${QMAKESPEC} -o
> Makefile $QMAKE_VARSUBST_PRE $AFTER $PROFILES $QMAKE_VARSUBST_POST'" unset
> QMAKESPEC || true
>  	${OE_QMAKE_QMAKE} -makefile -spec ${QMAKESPEC} -o Makefile
> $QMAKE_VARSUBST_PRE $AFTER $PROFILES $QMAKE_VARSUBST_POST || die "Error
> calling ${OE_QMAKE_QMAKE} on $PROFILES"

Acked-by: Paul Eggleton <paul.eggleton@linux.intel.com>

-- 

Paul Eggleton
Intel Open Source Technology Centre

Re: [PATCH 2/3] qmake_base.bbclass: fix lrelease/lupdate binary names

[Richard Purdie]

On Thu, 2011-06-02 at 17:51 +0100, Paul Eggleton wrote:
> On Wednesday 01 June 2011 21:09:54 Otavio Salvador wrote:
> > To support translation, qmake based projects usually call lrelease and
> > lupdate however OE changes the binary names so this needs some mangle
> > to work out of box.
> > 
> > Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>
> > ---
> >  meta/classes/qmake_base.bbclass |    5 +++++
> >  1 files changed, 5 insertions(+), 0 deletions(-)
> > 
> > diff --git a/meta/classes/qmake_base.bbclass
> > b/meta/classes/qmake_base.bbclass index a054efd..165d689 100644
> > --- a/meta/classes/qmake_base.bbclass
> > +++ b/meta/classes/qmake_base.bbclass
> > @@ -92,6 +92,11 @@ qmake_base_do_configure() {
> >  		bbnote "qmake prevar substitution: ${EXTRA_QMAKEVARS_PRE}"
> >  	fi
> > 
> > +	# Hack .pro files to use OE utilities
> > +	find -name '*.pro' \
> > +	     -exec sed -i -e 's,=\s*.*/lrelease,= ${OE_QMAKE_LRELEASE},g' \
> > +	                  -e 's,=\s*.*/lupdate,= ${OE_QMAKE_LUPDATE},g' '{}' ';'
> > +
> >  #bbnote "Calling '${OE_QMAKE_QMAKE} -makefile -spec ${QMAKESPEC} -o
> > Makefile $QMAKE_VARSUBST_PRE $AFTER $PROFILES $QMAKE_VARSUBST_POST'" unset
> > QMAKESPEC || true
> >  	${OE_QMAKE_QMAKE} -makefile -spec ${QMAKESPEC} -o Makefile
> > $QMAKE_VARSUBST_PRE $AFTER $PROFILES $QMAKE_VARSUBST_POST || die "Error
> > calling ${OE_QMAKE_QMAKE} on $PROFILES"
> 
> Acked-by: Paul Eggleton <paul.eggleton@linux.intel.com>

Merged to master, thanks.

Richard

Re: [PATCH 1/3] busybox: enable mdev by default

[Khem Raj]

On Thursday, June 02, 2011 09:37:41 AM Richard Purdie wrote:
> On Wed, 2011-06-01 at 20:40 +0000, Otavio Salvador wrote:
> > On Wed, Jun 1, 2011 at 20:37, Phil Blundell <pb@pbcl.net> wrote:
> > > On Wed, 2011-06-01 at 20:09 +0000, Otavio Salvador wrote:
> > >> -# CONFIG_MDEV is not set
> > >> +CONFIG_MDEV=y
> > > 
> > > Per previous discussion, I am still uneasy about this change.  I think
> > > we really need some sort of coherent policy for what exactly the
> > > default busybox configuration in oe-core is meant to be doing, and
> > > then (if necessary) a set of patches to make it match the policy. 
> > > Just flipping random features on and off does not seem like a good way
> > > to proceed.
> > 
> > OE-core has support to mdev as device handling mechanism as such this
> > ought to be enabled by default IMO.
> > 
> > Personally it doesn't matter since I have already overriden it in my
> > internal layer.
> 
> I'm afraid I'm with Phil on this. I don't like the idea of enabling
> something we don't actually use. This really needs to become some kind
> of configure option which would at the same time disable/replace udev so
> the patch in its currently form isn't acceptable.
> 

mdev or udev are image features so probably should be controlled by 
IMAGE_FEATURES or some such
> Cheers,
> 
> Richard
> 
> 
> 
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core

Re: [PATCH 1/3] busybox: enable mdev by default

[Koen Kooi]

Op 3 jun 2011, om 03:06 heeft Khem Raj het volgende geschreven:

> On Thursday, June 02, 2011 09:37:41 AM Richard Purdie wrote:
>> On Wed, 2011-06-01 at 20:40 +0000, Otavio Salvador wrote:
>>> On Wed, Jun 1, 2011 at 20:37, Phil Blundell <pb@pbcl.net> wrote:
>>>> On Wed, 2011-06-01 at 20:09 +0000, Otavio Salvador wrote:
>>>>> -# CONFIG_MDEV is not set
>>>>> +CONFIG_MDEV=y
>>>> 
>>>> Per previous discussion, I am still uneasy about this change.  I think
>>>> we really need some sort of coherent policy for what exactly the
>>>> default busybox configuration in oe-core is meant to be doing, and
>>>> then (if necessary) a set of patches to make it match the policy. 
>>>> Just flipping random features on and off does not seem like a good way
>>>> to proceed.
>>> 
>>> OE-core has support to mdev as device handling mechanism as such this
>>> ought to be enabled by default IMO.
>>> 
>>> Personally it doesn't matter since I have already overriden it in my
>>> internal layer.
>> 
>> I'm afraid I'm with Phil on this. I don't like the idea of enabling
>> something we don't actually use. This really needs to become some kind
>> of configure option which would at the same time disable/replace udev so
>> the patch in its currently form isn't acceptable.
>> 
> 
> mdev or udev are image features so probably should be controlled by 
> IMAGE_FEATURES or some such

You can't put IMAGE_FEATURES in the equivalent of EXTRA_OECONF since the packages in the feeds don't magically change when being installed in a different image.

Re: [PATCH 1/3] busybox: enable mdev by default

[Richard Purdie]

On Fri, 2011-06-03 at 08:37 +0200, Koen Kooi wrote:
> Op 3 jun 2011, om 03:06 heeft Khem Raj het volgende geschreven:
> 
> > On Thursday, June 02, 2011 09:37:41 AM Richard Purdie wrote:
> >> On Wed, 2011-06-01 at 20:40 +0000, Otavio Salvador wrote:
> >>> On Wed, Jun 1, 2011 at 20:37, Phil Blundell <pb@pbcl.net> wrote:
> >>>> On Wed, 2011-06-01 at 20:09 +0000, Otavio Salvador wrote:
> >>>>> -# CONFIG_MDEV is not set
> >>>>> +CONFIG_MDEV=y
> >>>> 
> >>>> Per previous discussion, I am still uneasy about this change.  I think
> >>>> we really need some sort of coherent policy for what exactly the
> >>>> default busybox configuration in oe-core is meant to be doing, and
> >>>> then (if necessary) a set of patches to make it match the policy. 
> >>>> Just flipping random features on and off does not seem like a good way
> >>>> to proceed.
> >>> 
> >>> OE-core has support to mdev as device handling mechanism as such this
> >>> ought to be enabled by default IMO.
> >>> 
> >>> Personally it doesn't matter since I have already overriden it in my
> >>> internal layer.
> >> 
> >> I'm afraid I'm with Phil on this. I don't like the idea of enabling
> >> something we don't actually use. This really needs to become some kind
> >> of configure option which would at the same time disable/replace udev so
> >> the patch in its currently form isn't acceptable.
> >> 
> > 
> > mdev or udev are image features so probably should be controlled by 
> > IMAGE_FEATURES or some such
> 
> You can't put IMAGE_FEATURES in the equivalent of EXTRA_OECONF since
> the packages in the feeds don't magically change when being installed
> in a different image.

Right, it would have to be a DISTRO_FEATURE for that reason...

Cheers,

Richard

Re: [PATCH 1/3] busybox: enable mdev by default

[Koen Kooi]

Op 3 jun 2011, om 10:24 heeft Richard Purdie het volgende geschreven:

> On Fri, 2011-06-03 at 08:37 +0200, Koen Kooi wrote:
>> Op 3 jun 2011, om 03:06 heeft Khem Raj het volgende geschreven:
>> 
>>> On Thursday, June 02, 2011 09:37:41 AM Richard Purdie wrote:
>>>> On Wed, 2011-06-01 at 20:40 +0000, Otavio Salvador wrote:
>>>>> On Wed, Jun 1, 2011 at 20:37, Phil Blundell <pb@pbcl.net> wrote:
>>>>>> On Wed, 2011-06-01 at 20:09 +0000, Otavio Salvador wrote:
>>>>>>> -# CONFIG_MDEV is not set
>>>>>>> +CONFIG_MDEV=y
>>>>>> 
>>>>>> Per previous discussion, I am still uneasy about this change.  I think
>>>>>> we really need some sort of coherent policy for what exactly the
>>>>>> default busybox configuration in oe-core is meant to be doing, and
>>>>>> then (if necessary) a set of patches to make it match the policy. 
>>>>>> Just flipping random features on and off does not seem like a good way
>>>>>> to proceed.
>>>>> 
>>>>> OE-core has support to mdev as device handling mechanism as such this
>>>>> ought to be enabled by default IMO.
>>>>> 
>>>>> Personally it doesn't matter since I have already overriden it in my
>>>>> internal layer.
>>>> 
>>>> I'm afraid I'm with Phil on this. I don't like the idea of enabling
>>>> something we don't actually use. This really needs to become some kind
>>>> of configure option which would at the same time disable/replace udev so
>>>> the patch in its currently form isn't acceptable.
>>>> 
>>> 
>>> mdev or udev are image features so probably should be controlled by 
>>> IMAGE_FEATURES or some such
>> 
>> You can't put IMAGE_FEATURES in the equivalent of EXTRA_OECONF since
>> the packages in the feeds don't magically change when being installed
>> in a different image.
> 
> Right, it would have to be a DISTRO_FEATURE for that reason...

Right, but you can build mdev and choose not to use it, which is what .dev currently does. So while the distro could in theory set a DISTRO_FEATURE, we shouldn't make other decisions based on that, like deciding that if mdev is in DISTRO_FEATURES every image will use it.

regards,

Koen

Re: [PATCH 1/3] busybox: enable mdev by default

[Richard Purdie]

On Fri, 2011-06-03 at 10:59 +0200, Koen Kooi wrote:
> Op 3 jun 2011, om 10:24 heeft Richard Purdie het volgende geschreven:
> 
> > On Fri, 2011-06-03 at 08:37 +0200, Koen Kooi wrote:
> >> Op 3 jun 2011, om 03:06 heeft Khem Raj het volgende geschreven:
> >> 
> >>> On Thursday, June 02, 2011 09:37:41 AM Richard Purdie wrote:
> >>>> On Wed, 2011-06-01 at 20:40 +0000, Otavio Salvador wrote:
> >>>>> On Wed, Jun 1, 2011 at 20:37, Phil Blundell <pb@pbcl.net> wrote:
> >>>>>> On Wed, 2011-06-01 at 20:09 +0000, Otavio Salvador wrote:
> >>>>>>> -# CONFIG_MDEV is not set
> >>>>>>> +CONFIG_MDEV=y
> >>>>>> 
> >>>>>> Per previous discussion, I am still uneasy about this change.  I think
> >>>>>> we really need some sort of coherent policy for what exactly the
> >>>>>> default busybox configuration in oe-core is meant to be doing, and
> >>>>>> then (if necessary) a set of patches to make it match the policy. 
> >>>>>> Just flipping random features on and off does not seem like a good way
> >>>>>> to proceed.
> >>>>> 
> >>>>> OE-core has support to mdev as device handling mechanism as such this
> >>>>> ought to be enabled by default IMO.
> >>>>> 
> >>>>> Personally it doesn't matter since I have already overriden it in my
> >>>>> internal layer.
> >>>> 
> >>>> I'm afraid I'm with Phil on this. I don't like the idea of enabling
> >>>> something we don't actually use. This really needs to become some kind
> >>>> of configure option which would at the same time disable/replace udev so
> >>>> the patch in its currently form isn't acceptable.
> >>>> 
> >>> 
> >>> mdev or udev are image features so probably should be controlled by 
> >>> IMAGE_FEATURES or some such
> >> 
> >> You can't put IMAGE_FEATURES in the equivalent of EXTRA_OECONF since
> >> the packages in the feeds don't magically change when being installed
> >> in a different image.
> > 
> > Right, it would have to be a DISTRO_FEATURE for that reason...
> 
> Right, but you can build mdev and choose not to use it, which is
> what .dev currently does. So while the distro could in theory set a
> DISTRO_FEATURE, we shouldn't make other decisions based on that, like
> deciding that if mdev is in DISTRO_FEATURES every image will use it.

Why we wouldn't enable that functionality based on the option?

Cheers,

Richard

Re: [PATCH 1/3] busybox: enable mdev by default

[Phil Blundell]

On Thu, 2011-06-02 at 18:06 -0700, Khem Raj wrote:
> mdev or udev are image features so probably should be controlled by 
> IMAGE_FEATURES or some such

Well, yeah, but this is the thing with busybox: it is a monolithic
executable, so you need to decide at configure time what features it's
going to include.  Although there is a separate busybox-mdev package, it
just contains the auxiliary scripting and the bulk of the mdev code goes
into busybox itself.

This means that making mdev a pure IMAGE_FEATURE isn't practical, since
it would basically require mdev to be always compiled in on the
offchance that some image might want it.  That would lead to code bloat
for those images (probably the majority) which don't.

Equally, having a single straightforward DISTRO_FEATURE to control mdev
is not terribly satisfactory, since some distros might want to configure
busybox with mdev enabled but actually use udev (or nothing) for some or
all of their images.

So, it seems that we really need both: a way for distros to configure
whether busybox is built with mdev on or not, and then an IMAGE_FEATURE
to control whether any given image gets mdev, or udev, or nothing.  The
former of those could be done as a DISTRO_FEATURE but, since it's only
going to affect busybox, I'm not sure that there is much benefit in
doing that; it seems like the distros could just as easily take care of
that by direct manipulation of the busybox configuration.

p.

Re: [PATCH 1/3] busybox: enable mdev by default

[Koen Kooi]

Op 3 jun 2011, om 11:50 heeft Richard Purdie het volgende geschreven:

> On Fri, 2011-06-03 at 10:59 +0200, Koen Kooi wrote:
>> Op 3 jun 2011, om 10:24 heeft Richard Purdie het volgende geschreven:
>> 
>>> On Fri, 2011-06-03 at 08:37 +0200, Koen Kooi wrote:
>>>> Op 3 jun 2011, om 03:06 heeft Khem Raj het volgende geschreven:
>>>> 
>>>>> On Thursday, June 02, 2011 09:37:41 AM Richard Purdie wrote:
>>>>>> On Wed, 2011-06-01 at 20:40 +0000, Otavio Salvador wrote:
>>>>>>> On Wed, Jun 1, 2011 at 20:37, Phil Blundell <pb@pbcl.net> wrote:
>>>>>>>> On Wed, 2011-06-01 at 20:09 +0000, Otavio Salvador wrote:
>>>>>>>>> -# CONFIG_MDEV is not set
>>>>>>>>> +CONFIG_MDEV=y
>>>>>>>> 
>>>>>>>> Per previous discussion, I am still uneasy about this change.  I think
>>>>>>>> we really need some sort of coherent policy for what exactly the
>>>>>>>> default busybox configuration in oe-core is meant to be doing, and
>>>>>>>> then (if necessary) a set of patches to make it match the policy. 
>>>>>>>> Just flipping random features on and off does not seem like a good way
>>>>>>>> to proceed.
>>>>>>> 
>>>>>>> OE-core has support to mdev as device handling mechanism as such this
>>>>>>> ought to be enabled by default IMO.
>>>>>>> 
>>>>>>> Personally it doesn't matter since I have already overriden it in my
>>>>>>> internal layer.
>>>>>> 
>>>>>> I'm afraid I'm with Phil on this. I don't like the idea of enabling
>>>>>> something we don't actually use. This really needs to become some kind
>>>>>> of configure option which would at the same time disable/replace udev so
>>>>>> the patch in its currently form isn't acceptable.
>>>>>> 
>>>>> 
>>>>> mdev or udev are image features so probably should be controlled by 
>>>>> IMAGE_FEATURES or some such
>>>> 
>>>> You can't put IMAGE_FEATURES in the equivalent of EXTRA_OECONF since
>>>> the packages in the feeds don't magically change when being installed
>>>> in a different image.
>>> 
>>> Right, it would have to be a DISTRO_FEATURE for that reason...
>> 
>> Right, but you can build mdev and choose not to use it, which is
>> what .dev currently does. So while the distro could in theory set a
>> DISTRO_FEATURE, we shouldn't make other decisions based on that, like
>> deciding that if mdev is in DISTRO_FEATURES every image will use it.
> 
> Why we wouldn't enable that functionality based on the option?

I was going to write a lengthy response, but Phil summed it up way too well in http://lists.linuxtogo.org/pipermail/openembedded-core/2011-June/003702.html :)