[yocto] Autobuilder migration status.

Flanagan, Elizabeth elizabeth.flanagan at intel.com
Mon Mar 25 15:29:36 PDT 2013


On Mon, Mar 25, 2013 at 3:18 PM, Trevor Woerner <twoerner at gmail.com> wrote:
> Thanks for all those answers.
>
> On Mon, Mar 25, 2013 at 5:45 PM, Flanagan, Elizabeth
> <elizabeth.flanagan at intel.com> wrote:
>> On Thu, Mar 21, 2013 at 8:31 AM, Trevor Woerner <twoerner at gmail.com> wrote:
>>> Not all builds, but these three seem to follow similar steps:
>>> 1. prepare
>>> 2. configure+build core-image-sato core-image-sato-dev
>>> core-image-sato-sdk core-image-minimal core-image-minimal-dev
>>> 3. configure+build core-image-sato core-image-sato-dev
>>> core-image-sato-sdk core-image-minimal core-image-minimal-dev
>>> 4. configure+build meta-toolchain-gmae
>>> 5. configure+build meta-toolchain-gmae
>>> 6. finish up
>
> ...
>
>>> Also, for me I think it would be better if #2 and #3 split
>>> out each of those build targets individually.
>>
>> Building them together provides a slight performance increase.
>
> Would there be even more of a performance increase if you just simply
> did a "bitbake world" instead of doing all of those targets
> sequentially? (I assume "world" builds all recipes and doesn't create
> images?)
>
>
> It appears that all build configurations (on a given slave) use the
> same location for DL_DIR and SSTATE_DIR, but the TMPDIR location is
> not provided in the configuration file. This would imply each build
> uses a fresh TMPDIR directory (since by default it would be set to
> ${TOPDIR}/tmp which comes from the local.conf)?

That is kind of correct. All build slaves on all machines utilize the
same DL_DIR and STATE_DIR exported via NFS. We've found putting the
DL_DIR and SSTATE_DIR on our nas (RAID 10 iirc) sped things up quite a
bit (as well as prefetching universe, splitting poky-lsb from poky,
symlinking when we could, etc).

> I'm still trying to
> figure out the best/optimal way to be using Yocto at home. When I
> first started, I would use a separate directory for each "project"
> (i.e. machine, board, architecture, etc). But lately I've realized
> that most people just do all their projects in one build location.
> However, when switching between different "projects", interspersed
> with various "git pulls" in between, I seem to get in a state when I'm
> forced to delete my entire TMPDIR location in order for anything to
> build cleanly again.
> Is this the general consensus?

I've not seen that but my workflow is different (one arch/distro
combination per directory with a shared DL_DIR and SSTATE_DIR, similar
to the AB). I'd be interested in others experiences.

-- 
Elizabeth Flanagan
Yocto Project
Build and Release



More information about the yocto mailing list