Re: Broken dependency behavior in master?

[Scott Garman <scott.a.garman@intel.com>]

On 04/01/2011 10:34 AM, Richard Purdie wrote:
> On Fri, 2011-04-01 at 08:23 -0700, Scott Garman wrote:
>> On 04/01/2011 04:31 AM, Richard Purdie wrote:
>>>> This makes me think something could be seriously broken in master when
>>>> it comes to handling build dependencies.
>>>>
>>>> I've verified that sgml-common-native is showing up as a build
>>>> dependency properly in the dependency explorer.
>>>
>>> I'm not convinced as its clear above that sgml-common-native is being
>>> built before docbook-dsssl-stylesheets-native do_configure runs.
>>
>> That was the case on the autobuilder. On my local development system, I
>> proved to myself without a doubt that sgml-common-native only got as far
>> as do_configure before the docbook-sgml-dtd-3.1-native error was
>> encountered.
>
> With or without sstate involvement?

I've reproduced this now, multiple times, using builds from scratch, 
including the fact that reverting that commit fixes the the problem for me.

> This looks like some kind of sstate issue but without logs this is near
> impossible to comment on. The logs you've pointed at don't show the
> problem you're describing.

Please grab what you need to from here:

http://www.zenlinux.com/dropbox/

> I don't think the problem is as simple as you're suggesting it is though
> as the core dependency resolution appears to be working (and has not
> changed in a long time). What perhaps isn't in something in the sstate
> task acceleration.

Sure, I'm not insisting that this is the cause - it was just my initial 
hunch. And your intuition is clearly better than mine when it comes to 
this, so I trust it.

Scott

-- 
Scott Garman
Embedded Linux Engineer - Yocto Project
Intel Open Source Technology Center