[meta-intel] New valleyisland support

[Bruce Ashfield]

On 2014-03-17, 6:31 PM, Darren Hart wrote:
> On 3/13/14, 6:53, "Bruce Ashfield" <bruce.ashfield at windriver.com> wrote:
>
>> On 14-03-13 09:43 AM, Tom Zanussi wrote:
>>> On Thu, 2014-03-13 at 06:40 +0000, Keskinarkaus, Teemu wrote:
>>>> Hi,
>>>>
>>>> Okay, here is a little background and some new issues I've encountered.
>>>>
>>>> We have been using Yocto 1.5.1 earlier to create image for our HWs.
>>>> Now the new HW uses E38xx so I started testing the new Yocto/meta-intel
>>>> - branches to get (better) support for it.
>>>>
>>>> Since the current ValleyIsland support for meta-intel was for Dylan -
>>>> branch of Yocto and also because I wasn't able to create image with
>>>> working Intel HD Graphics driver I started looking for newer versions.
>>>> Graphics worked with VESA-driver, but not with actual Intel HD Graphics
>>>> driver.  Also the kernel was quite old being 3.8.13.
>>>>
>>>> I did some digging and noticed that dvhart/bsp-ng had support for the
>>>> intel-corei7 so that's why I ended up there. I never intended to use it
>>>> other than testing if I can create working image that has working Intel
>>>> HD Graphics for Valleyisland.  Then I tried latest poky/meta-intel
>>>> combination since dvhart/bsp-ng was merged there.
>>>>
>>>> When trying to create core-image-sato using the intel-corei7-64 bsp I
>>>> ended up with this:
>>>>
>>>> ERROR: Fetcher failure: Unable to find revision
>>>> 29594404d7fe73cd80eaa4ee8c43dcc53970c60e in branch meta even from
>>>> upstream
>>>> ERROR: Function failed: Fetcher failure for URL:
>>>> 'git://git.pokylinux.org/linux-yocto-dev.git;nocheckout=1;branch=standar
>>>> d/base,meta;name=machine,meta'. Unable to fetch URL from any source.
>>>> ERROR: Logfile of failure stored in:
>>>> /opt/build/tmp/work/corei7-64-intel-common-poky-linux/linux-yocto-dev/3.
>>>> 14++gitAUTOINC+29594404d7_29594404d7-r0/temp/log.do_fetch.25380
>>>> ERROR: Task 73
>>>> (/opt/poky/../poky/meta/recipes-kernel/linux/linux-yocto-dev.bb,
>>>> do_fetch) failed with exit code '1'
>>>>
>>>> I'm not sure if that is a Yocto - bug, Meta-intel bug or my bug.
>>>>
>>>
>>> I ran into this too and got around it by the hack below.  Not sure why
>>> it's not working, how it's actually supposed to work or if there's
>>> something we need to be doing in meta-intel layers, adding Bruce..
>>
>> ok, this is report #2 of this. I'm firing up some tests here to see
>> if I can reproduce it.
>>
>> You shouldn't need your mod to make it work .. so I'll dig out what
>> is happening.
>>
>> Bruce
>>
>>>
>>> diff --git a/meta/recipes-kernel/linux/linux-yocto-dev.bb
>>> b/meta/recipes-kernel/linux/linux-
>>> index 5e09720..917714d 100644
>>> --- a/meta/recipes-kernel/linux/linux-yocto-dev.bb
>>> +++ b/meta/recipes-kernel/linux/linux-yocto-dev.bb
>>> @@ -28,8 +28,8 @@ SRC_URI =
>>> "git://git.pokylinux.org/linux-yocto-dev.git;nocheckout=1;branch
>>>    # linux-yocto-dev is the preferred provider, they will be overridden
>>> to
>>>    # AUTOREV in following anonymous python routine and resolved when the
>>>    # variables are finalized.
>>> -SRCREV_machine ?= "29594404d7fe73cd80eaa4ee8c43dcc53970c60e"
>>> -SRCREV_meta ?= "29594404d7fe73cd80eaa4ee8c43dcc53970c60e"
>
>
> Well that is obviously wrong, machine and meta can't possibly be the same
> commit ID. Bad dvhart... Oh wait... This is oe-core/poky. Bad Zedd.

:) guilty!

>
> These should be:
>
> $ git rev-parse origin/standard/base
> c4ee7a071d120118c4a5472222f7fb93b9c6f3b5
>
> $ git rev-parse origin/meta
> cdf9fb795b8e848cd3ddf3c5e0d98905fac27685
>
>
> Bruce, is this already queued or do you want me to prep a patch? Or... Are
> these the bogus ones that are meant to trigger failure .... Is the AUTOREV
> mechanism for linux-yocto-dev just not working?

It is supposed to be the latter, and Tom has been seeing issues, but I
can't reproduce the failure here. The AUTOREV overwrite is triggered in
anonymous python, all that I can think is that it isn't running early
enough.

Bruce

>
>>> +SRCREV_machine ?= "${AUTOREV}"
>>> +SRCREV_meta ?= "${AUTOREV}"
>
> FYI, you can do this in local.conf and avoid modifying the recipe sources
> (I prefer this method):
>
> SRCREV_meta_pn-linux-yocto-dev="${AUTOREV}"
> SRCREV_machine_pn-linux-yocto-dev="${AUTOREV}"
>
>
>
>>>
>>>    python () {
>>>        if d.getVar("PREFERRED_PROVIDER_virtual/kernel", True) !=
>>> "linux-yocto-dev":
>>>
>>>
>>>> Good to hear that linut-rt - support was not dropped, but merely just
>>>> changed.
>>>>
>>>> Btw. Is it intentional that in intel-corei7-64.conf the
>>>> PREFERRED_PROVIDER_virtual/kernel is set using = and not ?= which makes
>>>> it impossible to override it from later scripts? Or am I doing
>>>> something wrong again? I had to comment out that line to be able to use
>>>> my own kernel version on later tests.
>
>
> This is also my mistake. It should also be ?=. It was at = during
> development and I forgot to update it to ?= before pushing to meta-intel.
> Thank you for catching that.
>
>