bluez-dtl1-workaround not provided?

bluez-dtl1-workaround not provided?

[Andrea Adami]

Hello,

build from scratch of console image for c7x0/armv5te leads to:

...
Configuring sysvinit.
Collected errors:
 * satisfy_dependencies_for: Cannot satisfy the following dependencies
for task-base-extended:
 * <--->bluez-dtl1-workaround *.
 * opkg_install_cmd: Cannot install package task-base-extended.
 * resolve_conffiles: Existing conffile
/oe/build/tmp/rootfs/console-image/etc/device_table is different from
the conffile in the new package. The new conffile will be placed at
/oe/build/tmp/rootfs/console-image/etc/device_table-opkg.

A quick grep reveals:

andrea@mizar /oe/openembedded $ grep -R bluez-dtl1-workaround .
Binary file ./.git/index matches
./classes/kernel.bbclass:DEPENDS_kernel-module-dtl1-cs = "bluez-dtl1-workaround"
./classes/kernel.bbclass:RDEPENDS_kernel-module-dtl1-cs =
"bluez-dtl1-workaround"
./conf/machine/jornada7xx.conf:MACHINE_EXTRA_RDEPENDS =
"kernel-modules bluez-dtl1-workaround"
./recipes/tasks/task-base.bb:    ${@base_contains('COMBINED_FEATURES',
'pcmcia', 'bluez-dtl1-workaround', '',d)} \
./recipes/tasks/task-base.bb:# bluez-dtl1-workaround above is
workaround for bitbake not handling DEPENDS on it in
andrea@mizar /oe/openembedded $


So, first doubt is: why make this mandatory for everyone having pcmcia?
Second, why is it failing now?

Any idea?

Andrea

Re: bluez-dtl1-workaround not provided?

[Frans Meulenbroeks]

2010/11/25 Andrea Adami <andrea.adami@gmail.com>:
> Hello,
>
> build from scratch of console image for c7x0/armv5te leads to:
>
> ...
> Configuring sysvinit.
> Collected errors:
>  * satisfy_dependencies_for: Cannot satisfy the following dependencies
> for task-base-extended:
>  * <--->bluez-dtl1-workaround *.
>  * opkg_install_cmd: Cannot install package task-base-extended.
>  * resolve_conffiles: Existing conffile
> /oe/build/tmp/rootfs/console-image/etc/device_table is different from
> the conffile in the new package. The new conffile will be placed at
> /oe/build/tmp/rootfs/console-image/etc/device_table-opkg.
>
> A quick grep reveals:
>
> andrea@mizar /oe/openembedded $ grep -R bluez-dtl1-workaround .
> Binary file ./.git/index matches
> ./classes/kernel.bbclass:DEPENDS_kernel-module-dtl1-cs = "bluez-dtl1-workaround"
> ./classes/kernel.bbclass:RDEPENDS_kernel-module-dtl1-cs =
> "bluez-dtl1-workaround"
> ./conf/machine/jornada7xx.conf:MACHINE_EXTRA_RDEPENDS =
> "kernel-modules bluez-dtl1-workaround"
> ./recipes/tasks/task-base.bb:    ${@base_contains('COMBINED_FEATURES',
> 'pcmcia', 'bluez-dtl1-workaround', '',d)} \
> ./recipes/tasks/task-base.bb:# bluez-dtl1-workaround above is
> workaround for bitbake not handling DEPENDS on it in
> andrea@mizar /oe/openembedded $
>
>
> So, first doubt is: why make this mandatory for everyone having pcmcia?
> Second, why is it failing now?
>
> Any idea?
>
> Andrea
>

And the strange thing is if you build the package
bluez-dtl1-workaround manually first and after that build the image it
works.
Guess something is wrong with depends

Frans

Re: bluez-dtl1-workaround not provided?

[Martin Jansa]

[-- Attachment #1: Type: text/plain, Size: 356 bytes --]

On Thu, Nov 25, 2010 at 06:24:36PM +0100, Andrea Adami wrote:
> Hello,
> 
> build from scratch of console image for c7x0/armv5te leads to:

With new kernel I had this some time ago.. I'm using this patch since
then, but never send it here as I didn't look why this was needed
before..

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

[-- Attachment #2: 0004-kernel-kill-bluez-dtl1-workaround.patch --]
[-- Type: text/plain, Size: 1771 bytes --]

From df57c608af9e0a98ff1a1b3dc8259d80f86f09b9 Mon Sep 17 00:00:00 2001
From: Martin Jansa <Martin.Jansa@gmail.com>
Date: Tue, 10 Aug 2010 21:37:29 +0200
Subject: [PATCH 04/18] kernel: kill bluez-dtl1-workaround

Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
---
 classes/kernel.bbclass     |    5 -----
 recipes/tasks/task-base.bb |    1 -
 2 files changed, 0 insertions(+), 6 deletions(-)

diff --git a/classes/kernel.bbclass b/classes/kernel.bbclass
index 7039d27..5874d75 100644
--- a/classes/kernel.bbclass
+++ b/classes/kernel.bbclass
@@ -266,11 +266,6 @@ ALLOW_EMPTY_kernel = "1"
 ALLOW_EMPTY_kernel-base = "1"
 ALLOW_EMPTY_kernel-image = "1"
 
-# Userspace workarounds for kernel modules issues
-# This is shame, fix the kernel instead!
-DEPENDS_kernel-module-dtl1-cs = "bluez-dtl1-workaround"
-RDEPENDS_kernel-module-dtl1-cs = "bluez-dtl1-workaround"
-
 # renamed modules
 RPROVIDES_kernel-module-aes-generic = "kernel-module-aes"
 
diff --git a/recipes/tasks/task-base.bb b/recipes/tasks/task-base.bb
index ecc5278..7fc7cee 100644
--- a/recipes/tasks/task-base.bb
+++ b/recipes/tasks/task-base.bb
@@ -266,7 +266,6 @@ RRECOMMENDS_task-base-bluetooth = "\
     ${@base_contains('COMBINED_FEATURES', 'pcmcia', 'kernel-module-bluecard-cs', '',d)} \
     ${@base_contains('COMBINED_FEATURES', 'pcmcia', 'kernel-module-bluetoothuart-cs', '',d)} \
     ${@base_contains('COMBINED_FEATURES', 'pcmcia', 'kernel-module-dtl1-cs', '',d)} \
-    ${@base_contains('COMBINED_FEATURES', 'pcmcia', 'bluez-dtl1-workaround', '',d)} \
     "
 # bluez-dtl1-workaround above is workaround for bitbake not handling DEPENDS on it in
 # kernel.bbclass. It should be there as long as kernel-module-dtl1-cs is, but not longer.
-- 
1.7.3.2

Re: bluez-dtl1-workaround not provided?

[Khem Raj]

On Thu, Nov 25, 2010 at 11:46 PM, Martin Jansa <martin.jansa@gmail.com> wrote:
> On Thu, Nov 25, 2010 at 06:24:36PM +0100, Andrea Adami wrote:
>> Hello,
>>
>> build from scratch of console image for c7x0/armv5te leads to:
>
> With new kernel I had this some time ago.. I'm using this patch since
> then, but never send it here as I didn't look why this was needed
> before..
>

patch looks ok to me. Please install into master and release branch

Thx
-Khem

> --
> Martin 'JaMa' Jansa     jabber: Martin.Jansa@gmail.com
>
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
>
>

[PATCHv2] kernel.bbclass: put bluez-dtl1-workaround to DEPENDS directly

[Martin Jansa]

* DEPENDS_kernel-module-dtl1-cs aren't probably included in run queue
* RDEPENDS_kernel-module-dtl1-cs probably also doesn't put it to run queue,
  because kernel-module-* are created by
  PACKAGES_DYNAMIC += "kernel-module-*"
  so not even known before even building kernel, but does work ok in runtime
* bluez-dtl1-workaround is small and fast to build so won't hurt to
  build even when kernel doesn't have this bluez module enabled in
  defconfig

Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
---
 classes/kernel.bbclass     |    3 +--
 recipes/tasks/task-base.bb |    4 ----
 2 files changed, 1 insertions(+), 6 deletions(-)

diff --git a/classes/kernel.bbclass b/classes/kernel.bbclass
index 7039d27..422bcd7 100644
--- a/classes/kernel.bbclass
+++ b/classes/kernel.bbclass
@@ -1,7 +1,7 @@
 inherit linux-kernel-base module_strip
 
 PROVIDES += "virtual/kernel"
-DEPENDS += "virtual/${TARGET_PREFIX}gcc virtual/${TARGET_PREFIX}depmod-${@get_kernelmajorversion('${PV}')} virtual/${TARGET_PREFIX}gcc${KERNEL_CCSUFFIX} update-modules"
+DEPENDS += "virtual/${TARGET_PREFIX}gcc virtual/${TARGET_PREFIX}depmod-${@get_kernelmajorversion('${PV}')} virtual/${TARGET_PREFIX}gcc${KERNEL_CCSUFFIX} update-modules bluez-dtl1-workaround"
 
 # we include gcc above, we dont need virtual/libc
 INHIBIT_DEFAULT_DEPS = "1"
@@ -268,7 +268,6 @@ ALLOW_EMPTY_kernel-image = "1"
 
 # Userspace workarounds for kernel modules issues
 # This is shame, fix the kernel instead!
-DEPENDS_kernel-module-dtl1-cs = "bluez-dtl1-workaround"
 RDEPENDS_kernel-module-dtl1-cs = "bluez-dtl1-workaround"
 
 # renamed modules
diff --git a/recipes/tasks/task-base.bb b/recipes/tasks/task-base.bb
index ecc5278..6e2d9f0 100644
--- a/recipes/tasks/task-base.bb
+++ b/recipes/tasks/task-base.bb
@@ -266,11 +266,7 @@ RRECOMMENDS_task-base-bluetooth = "\
     ${@base_contains('COMBINED_FEATURES', 'pcmcia', 'kernel-module-bluecard-cs', '',d)} \
     ${@base_contains('COMBINED_FEATURES', 'pcmcia', 'kernel-module-bluetoothuart-cs', '',d)} \
     ${@base_contains('COMBINED_FEATURES', 'pcmcia', 'kernel-module-dtl1-cs', '',d)} \
-    ${@base_contains('COMBINED_FEATURES', 'pcmcia', 'bluez-dtl1-workaround', '',d)} \
     "
-# bluez-dtl1-workaround above is workaround for bitbake not handling DEPENDS on it in
-# kernel.bbclass. It should be there as long as kernel-module-dtl1-cs is, but not longer.
-
 RDEPENDS_task-base-irda = "\
     irda-utils"
 
-- 
1.7.3.2

Re: bluez-dtl1-workaround not provided?

[Marcin Juszkiewicz]

Dnia czwartek, 25 listopada 2010 o 18:24:36 Andrea Adami napisał(a):

> So, first doubt is: why make this mandatory for everyone having pcmcia?

This is for those rare people who still use bluetooth cf cards based on crappy 
chipset made by nokia. Card does not survive suspend/resume, gives very slow 
transfers etc. I had such one and rarely used.

Regards, 
-- 
JID:      hrw@jabber.org
Website:  http://marcin.juszkiewicz.com.pl/
LinkedIn: http://www.linkedin.com/in/marcinjuszkiewicz

Re: bluez-dtl1-workaround not provided?

[Phil Blundell]

On Thu, 2010-11-25 at 18:24 +0100, Andrea Adami wrote:
> So, first doubt is: why make this mandatory for everyone having pcmcia?

It shouldn't be.  The dependency in kernel-module-dtl1-cs should be
sufficient to pull it in for (only) those people who are installing the
driver that needs those bits.

p.

Re: bluez-dtl1-workaround not provided?

[Andrea Adami]

On Fri, Nov 26, 2010 at 11:59 AM, Phil Blundell <philb@gnu.org> wrote:
> On Thu, 2010-11-25 at 18:24 +0100, Andrea Adami wrote:
>> So, first doubt is: why make this mandatory for everyone having pcmcia?
>
> It shouldn't be.  The dependency in kernel-module-dtl1-cs should be
> sufficient to pull it in for (only) those people who are installing the
> driver that needs those bits.
>
> p.

The issue appeared again with today's metadata building
glibc/uclibc/eglibc images.
Can we clear the sensible task / classes before release?

Andrea

Re: bluez-dtl1-workaround not provided?

[Martin Jansa]

On Fri, Nov 26, 2010 at 10:59:31AM +0000, Phil Blundell wrote:
> On Thu, 2010-11-25 at 18:24 +0100, Andrea Adami wrote:
> > So, first doubt is: why make this mandatory for everyone having pcmcia?
> 
> It shouldn't be.  The dependency in kernel-module-dtl1-cs should be
> sufficient to pull it in for (only) those people who are installing the
> driver that needs those bits.

but many machine configs have
MACHINE_EXTRA_RRECOMMENDS = " kernel-modules"
which pulls kernel-module-dtl1-cs to every image

Regards,

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