Re: [OE-core] Patchwork & patch handling improvements


Barros Pena, Belen <belen.barros.pena@...>
 

On 02/12/2015 08:17, "openembedded-core-bounces@... on
behalf of Martin Jansa" <openembedded-core-bounces@...
on behalf of martin.jansa@...> wrote:

On Wed, Dec 02, 2015 at 04:01:40PM +1300, Paul Eggleton wrote:
On Tue, 01 Dec 2015 11:47:20 Martin Jansa wrote:
On Tue, Dec 01, 2015 at 07:49:50AM +1300, Paul Eggleton wrote:
Hi Trevor,

On Mon, 30 Nov 2015 10:19:35 Trevor Woerner wrote:
On 11/26/15 16:00, Paul Eggleton wrote:
I'm also
trying to ensure that the patch validation is generic enough so
it can
live in OE-Core, and thus we can easily update and refine it
over time
in
line with the code itself as well as encourage submitters to
use the
script on their own changes before sending.
This all sounds like an improvement and is therefore a step in
the right
direction :-)

A while back I had the idea of "porting" the kernel's
"checkpatch.pl" to
The Yocto Project (it was around the same time that I was trying
to
float the whole "Maintainers File" idea too, since I was also
trying to
re-purpose "get-maintainer.pl" as well). About one minute into
that
effort I realized the existing *.bb files were all over the place
in
terms of the order of statements and the order of the blocks of
statements. At that time I found one recipe style guide from OE,
and
another one from The Yocto Project, each of which described a
slightly
different preference. So I asked on the mailing list and quickly
discovered that both groups prefer a different style.

I'm not saying this job isn't worth doing, but I am pointing out
there's
the potential for feathers to be ruffled on both sides if someone
tries
to produce a definitive style guide for recipe files and then
enforces
it in an automated way. Since it is the OpenEmbedded Project's
job to
provide the recipes for The Yocto Project, I'm guessing this
question
needs to be decided by them? If that sounds reasonable, then
maybe The
Yocto Project needs to acquiesce to OE's decision?
I don't think there's that much of a division. I don't recall if it
was
you
that raised it at the time but the issue of having two style guides
did
get
rectified - I changed the one on the Yocto Project wiki to simply
be a
link to the OE style guide in June last year. It certainly didn't
come
about through a conscious decision to have a different style.

However there is a minor disagreement over indentation for shell
functions
between OE-Core and other layers - this persists because of the
backporting
pain a blanket replacement would potentially lead to. As I recall
this did
get discussed at the OE TSC level. I think that's one thing we
could just
not evaluate (or make an option) until such time as we resolve the
difference - and I do mean to see it resolved at some point in the
future.
Using consistent indentation (4 spaces) at least for new metadata
would
be step in right direction.

With the amount of changes which are backported to older releases I
still don't see this "backporting pain" argument. Doing it just before
the release is of course useful, because e.g. now more changes will be
backported to Jethro than to Fido or Dizzy. So having consistent
indentation in Jethro and master would prevent 95% of "backporting
pain". Maybe some Yocto 10.0 will finally get the meaning of
"consistent" indentation.
I agree it's not ideal. I said above, I do want to see it resolved.

Leaving indentation aside for a moment do you have any comments on my
proposal?
I'm not familiar with FDO fork, so I don't know how it looks and
behaves,
This is how it looks like currently

http://patchwork.freedesktop.org/project/intel-gfx/patches/

but any improvement on patchwork side is definitely welcome and
I appreciate it.

Does it support e.g. moving the patches to given bundle based on some
substring in subject? To sort e.g. meta-networking, meta-java,
meta-browser, .. patches automatically?
Mmm, you might not like this, but we are thinking of getting rid of
bundles. Basically, we assumed bundles were a manual way of creating patch
series. The new Patchwork can identify series, so we thought: great!
Bundles no longer needed.

There are other features been considered that might be an alternative to
bundles, like tagging



I don't expect FDO fork to provide other features I'm used to from
gerrit like cherry-picking to selected branch from the UI or doing the
review there. But still if we're stuck with patchwork forever (because
some people hate gerrit), then any improvement is really appreciated,
thanks for looking into it.

My only concern is about migrating current database, do you know if the
migration will keep the database including bundles as they are or do you
plan to set FDO version in parallel at least for some transition period?
Currently I have many patches in flight, because jenkins is running full
test-dependencies job for last 11 and based on progress it will take
14-21 more days to finish.

Regards,

--
Martin 'JaMa' Jansa jabber: Martin.Jansa@...

Join openembedded-architecture@lists.openembedded.org to automatically receive all group messages.