Re: package_write_ipk somehow looks into the wrong path

[Mike Looijmans <mike.looijmans@topic.nl>]

On 05/17/2013 01:13 PM, Mike Looijmans wrote:
> On 05/17/2013 01:11 PM, Mike Looijmans wrote:
>> On 05/17/2013 12:15 PM, Martin Jansa wrote:
>>> On Fri, May 17, 2013 at 11:43:22AM +0200, Mike Looijmans wrote:
>>>>
>>>> I got this very weird build failure. It builds fine for one machine
>>>> (zedboard), but it craps out in a really weird way on the "zynq-zc702"
>>>> machine.
>>>>
>>>> Somehow the "package_write_ipk" task wants to find files in a directory
>>>> with "1-r0" (a version that has once existed a long long time ago) in
>>>> the version part, instead of the proper git hashed name.
>>>>
>>>> I've tried a "cleansstate" but that did not help. The "compile" step in
>>>> this recipe takes about 1,5 hour on my i7 system, so debugging this
>>>> recipe is painfully slow...
>>>>
>>>> I've attached the recipes, pasted the log below.
>>>
>>> try it with 2/2 patch from
>>> https://bugzilla.yoctoproject.org/show_bug.cgi?id=4102
>>
>> Okay, I just manually applied those changes. Then I get:
>>
>> ERROR: Recipe fpga-image is trying to change PV from
>> '2.AUTOINC-bfd37feb034919cc11b3879745f2770810a4e796' to '1'. This will
>> cause do_package_write_ipk failing to find right workdir.
>>
>>
>> That helps, but I still don't have the faintest idea what is going on
>> here. What's wrong with my recipe?
>
> As a last resort, I tried just putting PV="3" into the recipe. WTF?
>
>
> ERROR: Recipe fpga-image is trying to change PV from '3' to '1'. This
> will cause do_package_write_ipk failing to find right workdir.
> ERROR: Function failed: read_subpackage_metadata
> ERROR: Logfile of failure stored in:
> /home/mike/zynq/build/tmp-eglibc/work/zynq-zc702-oe-linux-gnueabi/fpga-image/3-r0/temp/log.do_package_write_ipk.13519
>
> ERROR: Task 12
> (/home/mike/zynq/meta-zynq/recipes-bsp/fpga/fpga-image.bb,
> do_package_write_ipk) failed with exit code '1'
>
>
> Who's touching my PV?

And found the problem at last. As a long time user of "classic" OE, I added:
PACKAGE_ARCH = "${MACHINE}"
this was causing the failure, replacing it with
PACKAGE_ARCH = "${MACHINE_ARCH}"
as other recipes appear to do fixed it.

Mike.