Re: OMAP baseline test results for v3.10-rc6

[Tom Rini <trini@ti.com>]

On 06/25/2013 02:20 PM, Paul Walmsley wrote:
> + Vaibhav and Kevin
> 
> Hi,
> 
> On Tue, 25 Jun 2013, Felipe Balbi wrote:
> 
>> On Mon, Jun 17, 2013 at 05:23:17AM +0000, Paul Walmsley wrote:
>>> Boot to userspace:
>>>     FAIL ( 3/12): 37xxevm, am335xbone, am335xbonelt
>>
>> Paul, we have at least 2 different folks who can't reproduce your bone
>> and bone black boot to userspace failures. I wonder how you're trying to
>> boot them.
>>
>> Care to share your test scripts ?
> 
> Sure... the methodology is completely open and has been posted in the 
> online logs since the first test cycle.  (For some reason, almost no one 
> clicks through the test directory trees that I post online.  Is this a 
> documentation issue?  What can we do to make it easier for people to 
> explore this?)

Well, another link never hurts the search results :)

[snip]
> Am certainly open to the idea that there's something wrong with the way 
> that I'm booting either of these.  But AFAIK no one's been able to 
> identify exactly what it could be.  I haven't had the time recently to 
> spend hours going through the various permutations, given all the other 
> breakage :-(  BeagleBone-white has the additional complication that it is 
> not easy to automate, due to the way that power is delivered to the board, 
> so there is an extra dimension of difficulty there.

Ah-ha, I reproduced your failure.  If I make up a concat uImage + DTB,
rather than pass them separately, it fails to boot.  If you switch to
mainline U-Boot (v2012.10 or later) you get support for separate image +
dtb (v2013.04 gives you bootz and zImage support).  v2013.04 will also
work out of the box for BeagleBone-Black.

And yeah, I feel your pain about power cycling BeagleBone-White.  The QA
folks here sent me one of the relay controllers they use, and I think
Felipe is partial to one from phidgets.

>> Also, if you could share the entire thing, we will add your scripts to
>> our nightly tests as to try and avoid future regressions.
> 
> It would be great to have TI folks running those tests, or something 
> similar to them!  An early version of the test system has been shared with 
> a handful of folks, but it needs to be cleaned up further before broader 
> release.

We've got "something similar", at least wrt boot tests.  But since we
use separate kernel + DT, we hadn't seen this problem.

-- 
Tom