[OE-core] [PATCH] qmake2.bbclass: export OE_QMAKE_QT_CONFIG

[Jonathan Liu]

On 7/06/2013 9:31 AM, Martin Jansa wrote:
> On Fri, Jun 07, 2013 at 09:17:09AM +1000, Jonathan Liu wrote:
>> On 07/06/2013, at 6:07 AM, Saul Wold <sgw at linux.intel.com> wrote:
>>
>>> On 06/06/2013 12:18 PM, Felipe Tonello wrote:
>>>> On Thu, Jun 6, 2013 at 2:54 AM, Paul Eggleton
>>>> <paul.eggleton at linux.intel.com> wrote:
>>>>> On Wednesday 05 June 2013 10:01:09 Saul Wold wrote:
>>>>>> On 06/05/2013 09:57 AM, Paul Eggleton wrote:
>>>>>>> On Wednesday 05 June 2013 09:46:49 Saul Wold wrote:
>>>>>>>> On 06/05/2013 02:30 AM, Martin Jansa wrote:
>>>>>>>>> On Wed, Jun 05, 2013 at 07:01:50PM +1000, Jonathan Liu wrote:
>>>>>>>>>> qconfig.pri was not being loaded by qmake properly. This means Qt
>>>>>>>>>> qmake projects are unable to query QT_ARCH, QT_VERSION and other
>>>>>>>>>> variables defined in qconfig.pri.
>>>>>>>>>>
>>>>>>>>>> Export OE_QMAKE_QT_CONFIG, setting it to the location of qconfig.pri
>>>>>>>>>> so that it can be located by qmake.
>>>>>>>>> There is such patch already:
>>>>>>>>> http://lists.openembedded.org/pipermail/openembedded-core/2013-May/07831
>>>>>>>>> 4.
>>>>>>>>> html
>>>>>>>> Got lost in my queue, adding it for the next MUT.
>>>>>>> FWIW I am still concerned by the assertion in the commit message about it
>>>>>>> causing undesirable behaviour. I'd want that checked out and the note
>>>>>>> removed before we look at merging this.
>>>>>> OK, I will shelf this until I hear more about the testing of this change
>>>>>> regarding the original patch's note.
>>>>> So I tested a rebuild of qt4-x11-free with Felipe's patch and two of
>>>>> Jonathan's ("qt4: fix QMAKE_QT_CONFIG being overwritten with empty value" and
>>>>> "classes/qmake_base: allow parallel make"), and did not observe any unexpected
>>>>> behaviour and buildhistory looks clean for qt itself and fotowall/quicky. So
>>>>> I'd say feel free to throw these into MUT, but I think we need to drop the
>>>>> note.
>>>> Awesome.
>>> So can someone put a final updated patch series together.  I am still not sure about Jonathans's allow parallel make change, it seemed to have the -j option since it's added with a _prepend which is not overriden by the = assignment.
>>>
>>> Sau!
>> It has make -j 16 MAKEFLAGS=.
>> -j 16 adds to MAKEFLAGS, and MAKEFLAGS= clears it. It is trivial to
>> check this is the case and I have verified it both when building
>> fotowall and independently on command line for building Qt projects
>> outside of Yocto.
> disadvantage of this (at least in qt5) is that it starts independent
> make processes which all inherit MAKEFLAGS, so instead of 16 processes
> you can get a lot more - I've seen something like 40 when building
> qtwebkit with -j 8.
>
> But on other hand it's faster then -j1.
> https://github.com/meta-qt5/meta-qt5/commit/e9c61a9e6f51f500df349f8fe569ec095a1b8d2b
> https://github.com/meta-qt5/meta-qt5/commit/f76e45b5a8c59b1b8e4b806d2a2a09e6baf4d4e2
>
What about using make -j 16 -l 16?
This will prevent more make processes being created if the load average 
is 16 or higher.

Regards,
Jonathan