[OE-core] Glib-2.0 cross-compile issue Was: Master Failures

Saul Wold sgw at linux.intel.com
Wed Jul 18 00:57:29 UTC 2012


On 07/17/2012 05:52 PM, Martin Jansa wrote:
> On Wed, Jul 18, 2012 at 02:38:50AM +0200, Martin Jansa wrote:
>> On Tue, Jul 17, 2012 at 05:26:48PM -0700, Saul Wold wrote:
>>> On 07/17/2012 02:58 PM, Saul Wold wrote:
>>>>
>>>> Folks,
>>>>
>>>> I am seeing a crop of failures on Master, this is not any failure I have
>>>> seen recently on any of my MUT builds. It appears to be 2 failures,
>>>> possibly related.
>>>
>>> Not related, I just confirmed this.
>>>
>>> The "ls: cannot access" are coming from license.bbclass, seems to be a
>>> timing issue related to the last patch here, not sure what triggered it
>>> here (I am not seeing it on other builds!)
>>>
>>> Paul's Change:
>>>
>>> http://git.openembedded.org/openembedded-core-contrib/log/?h=sgw%2Fstage&qt=grep&q=license
>>>
>>>>
>>>> First seems to be missing pkgdata/runtime files, followed by a failed
>>>> dependency by libglib-2.0-utils on the native python (from the native
>>>> sysroot).
>>>>
>>> This seems to be related to the glib update and package split out of the
>>> gdbus-codegen, looks like an auto tools substitution is happening for
>>> @PYTHON@ in the script and getting the native sysroot python.
>>
>> gdbus-codegen was in separate package before I've just moved package
>> definition from .bb to .inc.
>>
>> And buildhistory diff doesn't show anything weird:
>> http://git.shr-project.org/git/?p=buildhistory.git;a=commitdiff;h=00de88a123842b159de1d206e3d0d5775ed90a1e#patch2
>>
>> I don't see it here, but I'm not using RPM packaging with this check..
>>
>> Will try more tomorrow..
>
> It's probably caused by this
> http://permalink.gmane.org/gmane.comp.gnome.svn/604029
>
> which isn't very cross-compile friendly.. maybe we can just add patch
> reverting this change in 2.32.4?
>
Yup, that the change, I was just digging deeper, maybe just doing a sed 
in a configure_prepend to change it the $(PYTHON) -> "/usr/bin/env 
python" in the Makefile.am would be better

We have seen that kind of code before.

Sau!


> Cheers,
>
>>
>> Cheers,
>>
>>>
>>> Martin's 2 changes may need so work:
>>>
>>> http://git.openembedded.org/openembedded-core/commit/?id=85a930188300152bc9a1bbf640a2441f3177e26f
>>>
>>> and
>>> http://git.openembedded.org/openembedded-core/commit/?id=b070eb22372ad1ef5b21162d832772e560184cf8
>>>
>>> Please look at the autobuild failures and see if they can help you fix
>>> these issues.
>>>
>>> Thanks
>>> 	Sau!
>>>
>>>> I am just starting to dig into it, but if anyone has better ideas:
>>>>
>>>> Thanks
>>>>       Sau!
>>>>
>>>>
>>>>> | ls: cannot access
>>>>> /srv/home/pokybuild/yocto-autobuilder/yocto-slave/nightly-arm/build/build/tmp/pkgdata/*/runtime/dbus-1:
>>>>> No such file or directory

<snip>

>> --
>> Martin 'JaMa' Jansa     jabber: Martin.Jansa at gmail.com
>
>
>





More information about the Openembedded-core mailing list