[meta-raspberrypi][PATCH v2] qtbase: enable Raspberry Pi support

[meta-raspberrypi][PATCH v2] qtbase: enable Raspberry Pi support

[Jonathan Liu]

[Support #16]

Signed-off-by: Jonathan Liu <net147@gmail.com>
---
 recipes-qt/qt5/qtbase_%.bbappend | 6 ++++++
 1 file changed, 6 insertions(+)
 create mode 100644 recipes-qt/qt5/qtbase_%.bbappend

diff --git a/recipes-qt/qt5/qtbase_%.bbappend b/recipes-qt/qt5/qtbase_%.bbappend
new file mode 100644
index 0000000..3e5e667
--- /dev/null
+++ b/recipes-qt/qt5/qtbase_%.bbappend
@@ -0,0 +1,6 @@
+do_configure_prepend_rpi() {
+    cat > ${S}/mkspecs/oe-device-extra.pri << 'EOF'
+EGLFS_DEVICE_INTEGRATION = eglfs_brcm
+QMAKE_LIBS_EGL = -lEGL -lGLESv2
+EOF
+}
-- 
2.6.1

Re: [meta-raspberrypi][PATCH v2] qtbase: enable Raspberry Pi support

[Paul Eggleton]

On Monday 26 October 2015 09:43:08 Jonathan Liu wrote:
> [Support #16]
> 
> Signed-off-by: Jonathan Liu <net147@gmail.com>
> ---
>  recipes-qt/qt5/qtbase_%.bbappend | 6 ++++++
>  1 file changed, 6 insertions(+)
>  create mode 100644 recipes-qt/qt5/qtbase_%.bbappend
> 
> diff --git a/recipes-qt/qt5/qtbase_%.bbappend
> b/recipes-qt/qt5/qtbase_%.bbappend new file mode 100644
> index 0000000..3e5e667
> --- /dev/null
> +++ b/recipes-qt/qt5/qtbase_%.bbappend
> @@ -0,0 +1,6 @@
> +do_configure_prepend_rpi() {
> +    cat > ${S}/mkspecs/oe-device-extra.pri << 'EOF'
> +EGLFS_DEVICE_INTEGRATION = eglfs_brcm
> +QMAKE_LIBS_EGL = -lEGL -lGLESv2
> +EOF
> +}

I'm wondering if this is a good idea or not. I appreciate the intention, but 
does bbappending non-machine-specific recipes like this play well when you have 
other machines enabled in the same distro?

Cheers,
Paul

-- 

Paul Eggleton
Intel Open Source Technology Centre

Re: [meta-raspberrypi][PATCH v2] qtbase: enable Raspberry Pi support

[Jonathan Liu]

On 26/10/2015 7:08 PM, Paul Eggleton wrote:
> On Monday 26 October 2015 09:43:08 Jonathan Liu wrote:
>> [Support #16]
>>
>> Signed-off-by: Jonathan Liu <net147@gmail.com>
>> ---
>>   recipes-qt/qt5/qtbase_%.bbappend | 6 ++++++
>>   1 file changed, 6 insertions(+)
>>   create mode 100644 recipes-qt/qt5/qtbase_%.bbappend
>>
>> diff --git a/recipes-qt/qt5/qtbase_%.bbappend
>> b/recipes-qt/qt5/qtbase_%.bbappend new file mode 100644
>> index 0000000..3e5e667
>> --- /dev/null
>> +++ b/recipes-qt/qt5/qtbase_%.bbappend
>> @@ -0,0 +1,6 @@
>> +do_configure_prepend_rpi() {
>> +    cat > ${S}/mkspecs/oe-device-extra.pri << 'EOF'
>> +EGLFS_DEVICE_INTEGRATION = eglfs_brcm
>> +QMAKE_LIBS_EGL = -lEGL -lGLESv2
>> +EOF
>> +}
> I'm wondering if this is a good idea or not. I appreciate the intention, but
> does bbappending non-machine-specific recipes like this play well when you have
> other machines enabled in the same distro?
>
> Cheers,
> Paul
>
It would rebuild qtbase (and recipes depending on it) without Raspberry 
Pi support for other machines.
I see similar machine-specific do_configure_prepend elsewhere (e.g. 
https://github.com/Freescale/meta-fsl-arm/blob/master/qt5-layer/recipes-qt/qt5/qtbase_%25.bbappend).

Otavio, any thoughts on a better way to handle this?

Regards,
Jonathan

Re: [meta-raspberrypi][PATCH v2] qtbase: enable Raspberry Pi support

[Andrei Gherzan]

Hi Paul,

On Mon, Oct 26, 2015 at 08:08:29AM +0000, Paul Eggleton wrote:
> On Monday 26 October 2015 09:43:08 Jonathan Liu wrote:
> > [Support #16]
> >
> > Signed-off-by: Jonathan Liu <net147@gmail.com>
> > ---
> >  recipes-qt/qt5/qtbase_%.bbappend | 6 ++++++
> >  1 file changed, 6 insertions(+)
> >  create mode 100644 recipes-qt/qt5/qtbase_%.bbappend
> >
> > diff --git a/recipes-qt/qt5/qtbase_%.bbappend
> > b/recipes-qt/qt5/qtbase_%.bbappend new file mode 100644
> > index 0000000..3e5e667
> > --- /dev/null
> > +++ b/recipes-qt/qt5/qtbase_%.bbappend
> > @@ -0,0 +1,6 @@
> > +do_configure_prepend_rpi() {
> > +    cat > ${S}/mkspecs/oe-device-extra.pri << 'EOF'
> > +EGLFS_DEVICE_INTEGRATION = eglfs_brcm
> > +QMAKE_LIBS_EGL = -lEGL -lGLESv2
> > +EOF
> > +}
>
> I'm wondering if this is a good idea or not. I appreciate the intention, but
> does bbappending non-machine-specific recipes like this play well when you have
> other machines enabled in the same distro?
>

But this change will be applied only for _rpi. How is this affecting other
machines?

--
Andrei Gherzan

Re: [meta-raspberrypi][PATCH v2] qtbase: enable Raspberry Pi support

[Paul Eggleton]

On Monday 26 October 2015 10:15:49 Andrei Gherzan wrote:
> Hi Paul,
> 
> On Mon, Oct 26, 2015 at 08:08:29AM +0000, Paul Eggleton wrote:
> > On Monday 26 October 2015 09:43:08 Jonathan Liu wrote:
> > > [Support #16]
> > > 
> > > Signed-off-by: Jonathan Liu <net147@gmail.com>
> > > ---
> > > 
> > >  recipes-qt/qt5/qtbase_%.bbappend | 6 ++++++
> > >  1 file changed, 6 insertions(+)
> > >  create mode 100644 recipes-qt/qt5/qtbase_%.bbappend
> > > 
> > > diff --git a/recipes-qt/qt5/qtbase_%.bbappend
> > > b/recipes-qt/qt5/qtbase_%.bbappend new file mode 100644
> > > index 0000000..3e5e667
> > > --- /dev/null
> > > +++ b/recipes-qt/qt5/qtbase_%.bbappend
> > > @@ -0,0 +1,6 @@
> > > +do_configure_prepend_rpi() {
> > > +    cat > ${S}/mkspecs/oe-device-extra.pri << 'EOF'
> > > +EGLFS_DEVICE_INTEGRATION = eglfs_brcm
> > > +QMAKE_LIBS_EGL = -lEGL -lGLESv2
> > > +EOF
> > > +}
> > 
> > I'm wondering if this is a good idea or not. I appreciate the intention,
> > but does bbappending non-machine-specific recipes like this play well
> > when you have other machines enabled in the same distro?
> 
> But this change will be applied only for _rpi. How is this affecting other
> machines?

If everything works such that the resulting packages get marked as machine-
specific and the package manager correctly installs them in preference to the 
generic architecture packages on the machines in question, then there won't be 
a massive problem (though the side-effect is probably that any recipes 
depending on qtbase will also become machine-specific, that is possibly more of 
an issue).

Cheers,
Paul

-- 

Paul Eggleton
Intel Open Source Technology Centre

Re: [meta-raspberrypi][PATCH v2] qtbase: enable Raspberry Pi support

[Jonathan Liu]

On 26/10/2015 8:41 PM, Paul Eggleton wrote:
> On Monday 26 October 2015 10:15:49 Andrei Gherzan wrote:
>> Hi Paul,
>>
>> On Mon, Oct 26, 2015 at 08:08:29AM +0000, Paul Eggleton wrote:
>>> On Monday 26 October 2015 09:43:08 Jonathan Liu wrote:
>>>> [Support #16]
>>>>
>>>> Signed-off-by: Jonathan Liu <net147@gmail.com>
>>>> ---
>>>>
>>>>   recipes-qt/qt5/qtbase_%.bbappend | 6 ++++++
>>>>   1 file changed, 6 insertions(+)
>>>>   create mode 100644 recipes-qt/qt5/qtbase_%.bbappend
>>>>
>>>> diff --git a/recipes-qt/qt5/qtbase_%.bbappend
>>>> b/recipes-qt/qt5/qtbase_%.bbappend new file mode 100644
>>>> index 0000000..3e5e667
>>>> --- /dev/null
>>>> +++ b/recipes-qt/qt5/qtbase_%.bbappend
>>>> @@ -0,0 +1,6 @@
>>>> +do_configure_prepend_rpi() {
>>>> +    cat > ${S}/mkspecs/oe-device-extra.pri << 'EOF'
>>>> +EGLFS_DEVICE_INTEGRATION = eglfs_brcm
>>>> +QMAKE_LIBS_EGL = -lEGL -lGLESv2
>>>> +EOF
>>>> +}
>>> I'm wondering if this is a good idea or not. I appreciate the intention,
>>> but does bbappending non-machine-specific recipes like this play well
>>> when you have other machines enabled in the same distro?
>> But this change will be applied only for _rpi. How is this affecting other
>> machines?
> If everything works such that the resulting packages get marked as machine-
> specific and the package manager correctly installs them in preference to the
> generic architecture packages on the machines in question, then there won't be
> a massive problem (though the side-effect is probably that any recipes
> depending on qtbase will also become machine-specific, that is possibly more of
> an issue).
>
> Cheers,
> Paul
>
Good point. The package arch needs to machine-specific. I will send a v3 
to fix this.

Regards,
Jonathan

Re: [meta-raspberrypi][PATCH v2] qtbase: enable Raspberry Pi support

[Paul Eggleton]

On Monday 26 October 2015 21:48:24 Jonathan Liu wrote:
> On 26/10/2015 8:41 PM, Paul Eggleton wrote:
> > On Monday 26 October 2015 10:15:49 Andrei Gherzan wrote:
> >> Hi Paul,
> >> 
> >> On Mon, Oct 26, 2015 at 08:08:29AM +0000, Paul Eggleton wrote:
> >>> On Monday 26 October 2015 09:43:08 Jonathan Liu wrote:
> >>>> [Support #16]
> >>>> 
> >>>> Signed-off-by: Jonathan Liu <net147@gmail.com>
> >>>> ---
> >>>> 
> >>>>   recipes-qt/qt5/qtbase_%.bbappend | 6 ++++++
> >>>>   1 file changed, 6 insertions(+)
> >>>>   create mode 100644 recipes-qt/qt5/qtbase_%.bbappend
> >>>> 
> >>>> diff --git a/recipes-qt/qt5/qtbase_%.bbappend
> >>>> b/recipes-qt/qt5/qtbase_%.bbappend new file mode 100644
> >>>> index 0000000..3e5e667
> >>>> --- /dev/null
> >>>> +++ b/recipes-qt/qt5/qtbase_%.bbappend
> >>>> @@ -0,0 +1,6 @@
> >>>> +do_configure_prepend_rpi() {
> >>>> +    cat > ${S}/mkspecs/oe-device-extra.pri << 'EOF'
> >>>> +EGLFS_DEVICE_INTEGRATION = eglfs_brcm
> >>>> +QMAKE_LIBS_EGL = -lEGL -lGLESv2
> >>>> +EOF
> >>>> +}
> >>> 
> >>> I'm wondering if this is a good idea or not. I appreciate the intention,
> >>> but does bbappending non-machine-specific recipes like this play well
> >>> when you have other machines enabled in the same distro?
> >> 
> >> But this change will be applied only for _rpi. How is this affecting
> >> other
> >> machines?
> > 
> > If everything works such that the resulting packages get marked as
> > machine-
> > specific and the package manager correctly installs them in preference to
> > the generic architecture packages on the machines in question, then there
> > won't be a massive problem (though the side-effect is probably that any
> > recipes depending on qtbase will also become machine-specific, that is
> > possibly more of an issue).
> > 
> > Cheers,
> > Paul
> 
> Good point. The package arch needs to machine-specific. I will send a v3
> to fix this.

I think that will get taken care of automatically by virtue of the machine-
specific override. My question is more about:

1) Will that mechanism entirely work, including desired behaviours in both the 
build system and the package manager, and

2) Are any side-effects worth it?

Cheers,
Paul

-- 

Paul Eggleton
Intel Open Source Technology Centre

Re: [meta-raspberrypi][PATCH v2] qtbase: enable Raspberry Pi support

[Jonathan Liu]

On 26/10/2015 9:54 PM, Paul Eggleton wrote:
> On Monday 26 October 2015 21:48:24 Jonathan Liu wrote:
>> On 26/10/2015 8:41 PM, Paul Eggleton wrote:
>>> On Monday 26 October 2015 10:15:49 Andrei Gherzan wrote:
>>>> Hi Paul,
>>>>
>>>> On Mon, Oct 26, 2015 at 08:08:29AM +0000, Paul Eggleton wrote:
>>>>> On Monday 26 October 2015 09:43:08 Jonathan Liu wrote:
>>>>>> [Support #16]
>>>>>>
>>>>>> Signed-off-by: Jonathan Liu <net147@gmail.com>
>>>>>> ---
>>>>>>
>>>>>>    recipes-qt/qt5/qtbase_%.bbappend | 6 ++++++
>>>>>>    1 file changed, 6 insertions(+)
>>>>>>    create mode 100644 recipes-qt/qt5/qtbase_%.bbappend
>>>>>>
>>>>>> diff --git a/recipes-qt/qt5/qtbase_%.bbappend
>>>>>> b/recipes-qt/qt5/qtbase_%.bbappend new file mode 100644
>>>>>> index 0000000..3e5e667
>>>>>> --- /dev/null
>>>>>> +++ b/recipes-qt/qt5/qtbase_%.bbappend
>>>>>> @@ -0,0 +1,6 @@
>>>>>> +do_configure_prepend_rpi() {
>>>>>> +    cat > ${S}/mkspecs/oe-device-extra.pri << 'EOF'
>>>>>> +EGLFS_DEVICE_INTEGRATION = eglfs_brcm
>>>>>> +QMAKE_LIBS_EGL = -lEGL -lGLESv2
>>>>>> +EOF
>>>>>> +}
>>>>> I'm wondering if this is a good idea or not. I appreciate the intention,
>>>>> but does bbappending non-machine-specific recipes like this play well
>>>>> when you have other machines enabled in the same distro?
>>>> But this change will be applied only for _rpi. How is this affecting
>>>> other
>>>> machines?
>>> If everything works such that the resulting packages get marked as
>>> machine-
>>> specific and the package manager correctly installs them in preference to
>>> the generic architecture packages on the machines in question, then there
>>> won't be a massive problem (though the side-effect is probably that any
>>> recipes depending on qtbase will also become machine-specific, that is
>>> possibly more of an issue).
>>>
>>> Cheers,
>>> Paul
>> Good point. The package arch needs to machine-specific. I will send a v3
>> to fix this.
> I think that will get taken care of automatically by virtue of the machine-
> specific override. My question is more about:
>
> 1) Will that mechanism entirely work, including desired behaviours in both the
> build system and the package manager, and
>
> 2) Are any side-effects worth it?
>
> Cheers,
> Paul
>
It wasn't automatically taken care of. I checked the package 
architecture and it was incorrect with the v2 patch.
The machine-specific architecture has a higher priority (from my reading 
of the opkg handling) so opkg at least should prefer it if a more 
generic architecture package is also present. I haven't tested it 
thoroughly though.

It seems worth it to have OpenGL ES graphics acceleration.

Regards,
Jonathan

Re: [meta-raspberrypi][PATCH v2] qtbase: enable Raspberry Pi support

[Otavio Salvador]

On Mon, Oct 26, 2015 at 6:24 AM, Jonathan Liu <net147@gmail.com> wrote:
> On 26/10/2015 7:08 PM, Paul Eggleton wrote:
> It would rebuild qtbase (and recipes depending on it) without Raspberry Pi
> support for other machines.
> I see similar machine-specific do_configure_prepend elsewhere (e.g.
> https://github.com/Freescale/meta-fsl-arm/blob/master/qt5-layer/recipes-qt/qt5/qtbase_%25.bbappend).
>
> Otavio, any thoughts on a better way to handle this?

We handle this moving it to a SoC subarch and using a dynamic package
arch handler to put every package depending on it as a subarch as
well. Ideally we could split the QPA module and build it separated. No
clue if it is possible or not.

-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750

Re: [meta-raspberrypi][PATCH v2] qtbase: enable Raspberry Pi support

[Jonathan Liu]

On 26/10/2015 10:28 PM, Otavio Salvador wrote:
> On Mon, Oct 26, 2015 at 6:24 AM, Jonathan Liu <net147@gmail.com> wrote:
>> On 26/10/2015 7:08 PM, Paul Eggleton wrote:
>> It would rebuild qtbase (and recipes depending on it) without Raspberry Pi
>> support for other machines.
>> I see similar machine-specific do_configure_prepend elsewhere (e.g.
>> https://github.com/Freescale/meta-fsl-arm/blob/master/qt5-layer/recipes-qt/qt5/qtbase_%25.bbappend).
>>
>> Otavio, any thoughts on a better way to handle this?
> We handle this moving it to a SoC subarch and using a dynamic package
> arch handler to put every package depending on it as a subarch as
> well. Ideally we could split the QPA module and build it separated. No
> clue if it is possible or not.
>
So for meta-fsl-arm, would only the qtbase-plugins package be using 
machine-specific arch and the other qtbase packages be using 
non-machine-specific arch?
Or is it done on a per-recipe rather than per-package granularity?

Regards,
Jonathan

Re: [meta-raspberrypi][PATCH v2] qtbase: enable Raspberry Pi support

[Martin Jansa]

[-- Attachment #1: Type: text/plain, Size: 2956 bytes --]

On Mon, Oct 26, 2015 at 10:54:43AM +0000, Paul Eggleton wrote:
> On Monday 26 October 2015 21:48:24 Jonathan Liu wrote:
> > On 26/10/2015 8:41 PM, Paul Eggleton wrote:
> > > On Monday 26 October 2015 10:15:49 Andrei Gherzan wrote:
> > >> Hi Paul,
> > >> 
> > >> On Mon, Oct 26, 2015 at 08:08:29AM +0000, Paul Eggleton wrote:
> > >>> On Monday 26 October 2015 09:43:08 Jonathan Liu wrote:
> > >>>> [Support #16]
> > >>>> 
> > >>>> Signed-off-by: Jonathan Liu <net147@gmail.com>
> > >>>> ---
> > >>>> 
> > >>>>   recipes-qt/qt5/qtbase_%.bbappend | 6 ++++++
> > >>>>   1 file changed, 6 insertions(+)
> > >>>>   create mode 100644 recipes-qt/qt5/qtbase_%.bbappend
> > >>>> 
> > >>>> diff --git a/recipes-qt/qt5/qtbase_%.bbappend
> > >>>> b/recipes-qt/qt5/qtbase_%.bbappend new file mode 100644
> > >>>> index 0000000..3e5e667
> > >>>> --- /dev/null
> > >>>> +++ b/recipes-qt/qt5/qtbase_%.bbappend
> > >>>> @@ -0,0 +1,6 @@
> > >>>> +do_configure_prepend_rpi() {
> > >>>> +    cat > ${S}/mkspecs/oe-device-extra.pri << 'EOF'
> > >>>> +EGLFS_DEVICE_INTEGRATION = eglfs_brcm
> > >>>> +QMAKE_LIBS_EGL = -lEGL -lGLESv2
> > >>>> +EOF
> > >>>> +}
> > >>> 
> > >>> I'm wondering if this is a good idea or not. I appreciate the intention,
> > >>> but does bbappending non-machine-specific recipes like this play well
> > >>> when you have other machines enabled in the same distro?
> > >> 
> > >> But this change will be applied only for _rpi. How is this affecting
> > >> other
> > >> machines?
> > > 
> > > If everything works such that the resulting packages get marked as
> > > machine-
> > > specific and the package manager correctly installs them in preference to
> > > the generic architecture packages on the machines in question, then there
> > > won't be a massive problem (though the side-effect is probably that any
> > > recipes depending on qtbase will also become machine-specific, that is
> > > possibly more of an issue).
> > > 
> > > Cheers,
> > > Paul
> > 
> > Good point. The package arch needs to machine-specific. I will send a v3
> > to fix this.
> 
> I think that will get taken care of automatically by virtue of the machine-
> specific override. My question is more about:

machine-specific override to set PACKAGE_ARCH works only when it's used as
subdirectory in SRC_URI, doesn't it? And it sets it only for qtbase, not
all other recipes which depend on qtbase (see my other reply in this
thread).

> 1) Will that mechanism entirely work, including desired behaviours in both the 
> build system and the package manager, and
> 
> 2) Are any side-effects worth it?
> 
> Cheers,
> Paul
> 
> -- 
> 
> Paul Eggleton
> Intel Open Source Technology Centre
> -- 
> _______________________________________________
> yocto mailing list
> yocto@yoctoproject.org
> https://lists.yoctoproject.org/listinfo/yocto

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

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 188 bytes --]

Re: [meta-raspberrypi][PATCH v2] qtbase: enable Raspberry Pi support

[Otavio Salvador]

On Mon, Oct 26, 2015 at 9:31 AM, Jonathan Liu <net147@gmail.com> wrote:
> On 26/10/2015 10:28 PM, Otavio Salvador wrote:
>> On Mon, Oct 26, 2015 at 6:24 AM, Jonathan Liu <net147@gmail.com> wrote:
>>> On 26/10/2015 7:08 PM, Paul Eggleton wrote:
>>> It would rebuild qtbase (and recipes depending on it) without Raspberry
>>> Pi
>>> support for other machines.
>>> I see similar machine-specific do_configure_prepend elsewhere (e.g.
>>>
>>> https://github.com/Freescale/meta-fsl-arm/blob/master/qt5-layer/recipes-qt/qt5/qtbase_%25.bbappend).
>>>
>>> Otavio, any thoughts on a better way to handle this?
>>
>> We handle this moving it to a SoC subarch and using a dynamic package
>> arch handler to put every package depending on it as a subarch as
>> well. Ideally we could split the QPA module and build it separated. No
>> clue if it is possible or not.
>
> So for meta-fsl-arm, would only the qtbase-plugins package be using
> machine-specific arch and the other qtbase packages be using
> non-machine-specific arch?
> Or is it done on a per-recipe rather than per-package granularity?

We have the handler, which allows for it to be done properly[1].

1. http://git.yoctoproject.org/cgit/cgit.cgi/meta-fsl-arm/tree/conf/machine/include/imx-base.inc#n68

Including 'qtbase' here will make it to move all packages depending on
it for the subarch.

The real long term solution is to try to split it and make the QPA
package machine specific. I don't know if it is possible.

-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750