[for-daisy] Pseudo backport breaks build in some hosts

[for-daisy] Pseudo backport breaks build in some hosts

[Otavio Salvador]

Hello,

In clean hosts the pseudo build is broken. It seems to be lacking the
attr dependency which has been added during the Dizzy development
cycle but not included when doing the backport for Daisy; I suspect it
has not been detected in Yocto Project AB because it should have attr
headers available.

We seem to need a better way to test backports as this is the second
time backports has included regressions. I think we ought to have a
daisy-next (and likely dizzy-next) branches with intended fixes for
easy testing outside of Yocto Project AB and get wider exposure for
different host setups.

Best Regards,

-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750

Re: [for-daisy] Pseudo backport breaks build in some hosts

[Saul Wold]

On 11/05/2014 12:34 PM, Otavio Salvador wrote:
> Hello,
>
> In clean hosts the pseudo build is broken. It seems to be lacking the
> attr dependency which has been added during the Dizzy development
> cycle but not included when doing the backport for Daisy; I suspect it
> has not been detected in Yocto Project AB because it should have attr
> headers available.
>
We actually just completed a full pass QA cycle on what was to be 
1.6.2rc1, we found a couple of issues that are being address (like we 
forgot to bump the actual Poky DISTRO_VERSION and the build-appliance 
rev). Along with a couple more CVE patches that have come in.  As you 
point out it did go through a full AB, which has different Linux OS 
setups.  The fact that this was missed, I am not sure.

> We seem to need a better way to test backports as this is the second
> time backports has included regressions. I think we ought to have a
> daisy-next (and likely dizzy-next) branches with intended fixes for
> easy testing outside of Yocto Project AB and get wider exposure for
> different host setups.
>
So I have been pushing my changes to daisy-next prior to them being 
merged into daisy itself.  You are welcome to pull from it, test it and 
report back, just beware that it can be rebased.

Sau!


> Best Regards,
>

Re: [for-daisy] Pseudo backport breaks build in some hosts

[Otavio Salvador]

On Wed, Nov 5, 2014 at 8:52 PM, Saul Wold <sgw@linux.intel.com> wrote:
> On 11/05/2014 12:34 PM, Otavio Salvador wrote:
>> In clean hosts the pseudo build is broken. It seems to be lacking the
>> attr dependency which has been added during the Dizzy development
>> cycle but not included when doing the backport for Daisy; I suspect it
>> has not been detected in Yocto Project AB because it should have attr
>> headers available.
>
> We actually just completed a full pass QA cycle on what was to be 1.6.2rc1,
> we found a couple of issues that are being address (like we forgot to bump
> the actual Poky DISTRO_VERSION and the build-appliance rev). Along with a
> couple more CVE patches that have come in.  As you point out it did go
> through a full AB, which has different Linux OS setups.  The fact that this
> was missed, I am not sure.

Well, last two big updates to Daisy has had issues; so I think we need
to improve the process.

One possibility which came to my mind is if you announce you are
intending to merge daisy-next on mailing list and give some days for
'regressions report' before merging.

I am not opposed to minor things go in during the process (or even
direct to daisy) but version bumps and big pulls would be good.

>> We seem to need a better way to test backports as this is the second
>> time backports has included regressions. I think we ought to have a
>> daisy-next (and likely dizzy-next) branches with intended fixes for
>> easy testing outside of Yocto Project AB and get wider exposure for
>> different host setups.
>>
> So I have been pushing my changes to daisy-next prior to them being merged
> into daisy itself.  You are welcome to pull from it, test it and report
> back, just beware that it can be rebased.

Great.


-- 
Otavio Salvador                             O.S. Systems
http://www.ossystems.com.br        http://code.ossystems.com.br
Mobile: +55 (53) 9981-7854            Mobile: +1 (347) 903-9750