Patchwork [meta-qt5] fix do_rootfs for images with dev-pkgs in IMAGE_FEATURES

login
register
mail settings
Submitter Andreas Müller
Date Sept. 26, 2013, 9:45 a.m.
Message ID <1380188747-9731-1-git-send-email-schnitzeltony@googlemail.com>
Download mbox | patch
Permalink /patch/59015/
State Rejected
Headers show

Comments

Andreas Müller - Sept. 26, 2013, 9:45 a.m.
| Collected errors:
|  * satisfy_dependencies_for: Cannot satisfy the following dependencies for qtquickcontrols-dev:
|  *     qtquickcontrols (= 5.1.0-r0) *
|  * opkg_install_cmd: Cannot install package qtquickcontrols-dev.

Signed-off-by: Andreas Müller <schnitzeltony@googlemail.com>
---
 recipes-qt/qt5/qtquickcontrols.inc | 2 ++
 1 file changed, 2 insertions(+)
ml@communistcode.co.uk - Sept. 26, 2013, 9:51 a.m.
On 26/09/13 10:45, Andreas Müller wrote:
> | Collected errors:
> |  * satisfy_dependencies_for: Cannot satisfy the following dependencies for qtquickcontrols-dev:
> |  *     qtquickcontrols (= 5.1.0-r0) *
> |  * opkg_install_cmd: Cannot install package qtquickcontrols-dev.
> 
> Signed-off-by: Andreas Müller <schnitzeltony@googlemail.com>
> ---
>  recipes-qt/qt5/qtquickcontrols.inc | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/recipes-qt/qt5/qtquickcontrols.inc b/recipes-qt/qt5/qtquickcontrols.inc
> index 63e884e..06b5350 100644
> --- a/recipes-qt/qt5/qtquickcontrols.inc
> +++ b/recipes-qt/qt5/qtquickcontrols.inc
> @@ -1,3 +1,5 @@
>  require qt5.inc
>  
>  DEPENDS += "qtdeclarative"
> +
> +ALLOW_EMPTY_${PN} = "1"
> 

Hi Andreas,

I ran into this problem the other day, and after speaking to Ross
Burton, he suggested that the better way to fix this is with:

RDEPENDS_${PN}-dev = ""

I'm starting to get the feeling that the dev-pkgs feature is also a
little broken, as it seems as though it doesn't handle cases where
sub-packages are installed, but the main package is not.
Ross Burton - Sept. 26, 2013, 9:58 a.m.
On 26 September 2013 10:51, Jack Mitchell <ml@communistcode.co.uk> wrote:
> I ran into this problem the other day, and after speaking to Ross
> Burton, he suggested that the better way to fix this is with:
>
> RDEPENDS_${PN}-dev = ""
>
> I'm starting to get the feeling that the dev-pkgs feature is also a
> little broken, as it seems as though it doesn't handle cases where
> sub-packages are installed, but the main package is not.

Well, the problem is the default dependency on the main package in the
-dev package, which happens whether or not the main package even
exists.  That's why I prefer removing the dependency instead of
creating empty packages just for the bad dependency.

Ross
Andreas Müller - Sept. 26, 2013, 10:02 a.m.
On Thu, Sep 26, 2013 at 11:58 AM, Burton, Ross <ross.burton@intel.com> wrote:
> On 26 September 2013 10:51, Jack Mitchell <ml@communistcode.co.uk> wrote:
>> I ran into this problem the other day, and after speaking to Ross
>> Burton, he suggested that the better way to fix this is with:
>>
>> RDEPENDS_${PN}-dev = ""
>>
>> I'm starting to get the feeling that the dev-pkgs feature is also a
>> little broken, as it seems as though it doesn't handle cases where
>> sub-packages are installed, but the main package is not.
>
> Well, the problem is the default dependency on the main package in the
> -dev package, which happens whether or not the main package even
> exists.  That's why I prefer removing the dependency instead of
> creating empty packages just for the bad dependency.
>
agreed :)

Andreas
Enrico Scholz - Sept. 26, 2013, 10:45 a.m.
"Burton, Ross" <ross.burton@intel.com> writes:

>> I ran into this problem the other day, and after speaking to Ross
>> Burton, he suggested that the better way to fix this is with:
>>
>> RDEPENDS_${PN}-dev = ""
>>
>> I'm starting to get the feeling that the dev-pkgs feature is also a
>> little broken, as it seems as though it doesn't handle cases where
>> sub-packages are installed, but the main package is not.
>
> Well, the problem is the default dependency on the main package in the
> -dev package, which happens whether or not the main package even
> exists.  That's why I prefer removing the dependency instead of
> creating empty packages just for the bad dependency.

ack.  This causes also problems with populate_sdk as it can add
conflicting packages to the package set.  E.g. when having 'dropbear'
and 'openssh-sftp-server' in the image, the -dev RDEPENDS tries to add
the conflicting 'dropbear' and 'openssh' packages to the sdk.

Instead of (or better: additional to) changing 'RDEPENDS_${PN}-dev', I
would prefer the removal of 'ALLOW_EMPTY_${PN}-dev' from bitbake.conf.
There is not reason to create -dev packages for applications.

In 'meta/conf/bitbake.conf' the '${PN}' in

|:RDEPENDS_${PN}-dev = "${PN} (= ${EXTENDPKGV})"

should be replaced by something which means the "library" package(s).


Enrico

Patch

diff --git a/recipes-qt/qt5/qtquickcontrols.inc b/recipes-qt/qt5/qtquickcontrols.inc
index 63e884e..06b5350 100644
--- a/recipes-qt/qt5/qtquickcontrols.inc
+++ b/recipes-qt/qt5/qtquickcontrols.inc
@@ -1,3 +1,5 @@ 
 require qt5.inc
 
 DEPENDS += "qtdeclarative"
+
+ALLOW_EMPTY_${PN} = "1"