[OE-core] [PATCH 1/1] dbus: fix a hard dependency about dbus-ptest

[Paul Eggleton]

On Monday 17 March 2014 10:44:44 Paul Eggleton wrote:
> On Monday 17 March 2014 10:09:28 Chong Lu wrote:
> > On 03/14/2014 06:12 PM, Paul Eggleton wrote:
> > > On Friday 14 March 2014 17:14:27 Chong Lu wrote:
> > >> On 03/13/2014 06:01 PM, Paul Eggleton wrote:
> > >>> On Thursday 13 March 2014 17:54:03 Chong Lu wrote:
> > >>>> On 03/06/2014 06:04 PM, Burton, Ross wrote:
> > >>>>> On 6 March 2014 08:06, Chong Lu <Chong.Lu at windriver.com> wrote:
> > >>>>>> +PACKAGES += "dbus-ptest"
> > >>>>>> +ALLOW_EMPTY_${PN}-ptest = "1"
> > >>>>>> +RDEPENDS_${PN}-ptest = "${@base_contains('IMAGE_FEATURES',
> > >>>>>> 'ptest-pkgs',
> > >>>>>> 'dbus-test-ptest', 'dbus-test', d)}"
> > >>>>> 
> > >>>>> Clearly I wasn't thinking clearly last week.  You can't check
> > >>>>> IMAGE_FEATURES in a non-image recipe so this isn't a good solution.
> > >>>> 
> > >>>> Do you have any suggestion about this issue?
> > >>>> If we don't check IMAGE_FEATURES, I have no way to resolve this
> > >>>> problem
> > >>>> that only build but not install ptest. I'm very confused.
> > >>> 
> > >>> This should check DISTRO_FEATURES, not IMAGE_FEATURES. Assuming ptest
> > >>> is
> > >>> in
> > >>> DISTRO_FEATURES, even if ptest is not in IMAGE_FEATURES it doesn't
> > >>> matter
> > >>> if the dependency between dbus-ptest and dbus-ptest-ptest exists
> > >>> because
> > >>> dbus- ptest itself shouldn't be in the image, so dbus-ptest-ptest
> > >>> shouldn't be pulled in either. Does that make sense?
> > >>> 
> > >>> Cheers,
> > >>> Paul
> > >> 
> > >> Hi Paul,
> > >> 
> > >> Thanks for your reply!
> > >> But I still have a question.
> > >> 
> > >> I try to following way:
> > >> 
> > >> # git diff
> > >> diff --git a/meta/recipes-core/dbus/dbus-ptest_1.6.18.bb
> > >> b/meta/recipes-core/dbus/dbus-ptest_1.6.18.bb
> > >> index bc260de..50dfe54 100644
> > >> --- a/meta/recipes-core/dbus/dbus-ptest_1.6.18.bb
> > >> +++ b/meta/recipes-core/dbus/dbus-ptest_1.6.18.bb
> > >> @@ -9,6 +9,7 @@ DEPENDS = "python-pygobject dbus dbus-glib"
> > >> 
> > >>    RDEPENDS_${PN} += "make"
> > >>    RDEPENDS_${PN}-dev = ""
> > >> 
> > >> +ALLOW_EMPTY_${PN} = "1"
> > >> 
> > >>    SRC_URI =
> > >>    "http://dbus.freedesktop.org/releases/dbus/dbus-${PV}.tar.gz
> > >>    \
> > >>    
> > >>               file://tmpdir.patch \
> > >> 
> > >> diff --git a/meta/recipes-core/dbus/dbus.inc
> > >> b/meta/recipes-core/dbus/dbus.inc
> > >> index 677ff78..ba7f3cb 100644
> > >> --- a/meta/recipes-core/dbus/dbus.inc
> > >> +++ b/meta/recipes-core/dbus/dbus.inc
> > >> @@ -6,7 +6,7 @@ LICENSE = "AFL-2 | GPLv2+"
> > >> 
> > >>    LIC_FILES_CHKSUM =
> > >>    "file://COPYING;md5=10dded3b58148f3f1fd804b26354af3e \
> > >> 
> > >> file://dbus/dbus.h;beginline=6;endline=20;md5=7755c9d7abccd5dbd25a6a974
> > >> 53
> > >> 8bb 3c" DEPENDS = "expat virtual/libintl"
> > >> -RDEPENDS_dbus = "${@base_contains('DISTRO_FEATURES', 'ptest',
> > >> 'dbus-ptest-ptest', '', d)}"
> > >> +RDEPENDS_dbus = "dbus-ptest"
> > > 
> > > The line above is the problem. You *cannot* do this or dbus-ptest (and
> > > therefore dbus-ptest-ptest) will always be installed whether you want it
> > > or not. You need to remove this line.
> > 
> > Do you mean the patch we only need to remove RDEPENDS_dbus =
> > "${@base_contains('DISTRO_FEATURES', 'ptest', 'dbus-ptest-ptest', '',
> > d)}" in dbus.inc ?
> > If do this, we can't build and install dbus-ptest through `bitbake dbus'
> > command.
> > How to relate between dbus and dbus-ptest?
> 
> Assuming you rename the current dbus-ptest recipe to dbus-test:
> 
> 1) If the dbus recipe refers to dbus-test-ptest elsewhere (i.e.
> RDEPENDS_dbus-ptest = "dbus-test-ptest") then dbus-ptest will be built as
> needed.

Sorry that should have been "dbus-test will be built as needed".

> 2) If dbus is part of the image and ptest-pkgs is in IMAGE_FEATURES, the
> dbus- ptest package (and therefore the dbus-test-ptest) will be brought in
> automatically. The relationship between dbus and dbus-ptest is solely on
> the name, you do not need to add any RDEPENDS for that part.

I should also mention that we need a fix for this soon in order to make the 1.6 
release (and we *really* don't want to release 1.6 with this issue.)

Thanks,
Paul

-- 

Paul Eggleton
Intel Open Source Technology Centre