flink-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Kostas Tzoumas <ktzou...@apache.org>
Subject Re: Question about Commit Policy
Date Wed, 07 Jan 2015 15:13:31 GMT
+1

Let's encourage the use of component tags, I don't see the need for
enforcing it. For commits that affect one component, I expect people will
use it.

On Wed, Jan 7, 2015 at 3:40 PM, Fabian Hueske <fhueske@apache.org> wrote:

> +1 for the guide and JIRA references.
>
> I'd keep the component tags optional though.
> As Max said, there is less space to display a meaning message if a commit
> addresses several components. Separating changes into commits by components
> sounds not very practical to me.
> Also without a clear definition of when to add which component tag, we
> cannot rely on them anyway.
>
> Git should also have better tools than browsing commit messages when
> looking for a commit that changed specific code.
>
> 2015-01-07 15:24 GMT+01:00 Stephan Ewen <sewen@apache.org>:
>
> > I personally like the tags very much. I think the streaming component was
> > the first to introduce it and it stuck me as a very good idea.
> >
> > +1 to stick with them
> >
> > On Wed, Jan 7, 2015 at 3:03 PM, Márton Balassi <balassi.marton@gmail.com
> >
> > wrote:
> >
> > > I prefer component declarations, the current best practice comes in
> handy
> > > when searching through commits. Answering a "when did key selection
> > change
> > > for streaming?" type question I just had to answer would have been a
> bit
> > > more difficult without it - manageable though.
> > >
> > > In case of streaming it does not yield much to omit the component
> > > declaration, most of the time then we would need to add it to the
> commit
> > > message itself, e.g. :
> > > "[streaming] Join API rework", could be e.g. rewritten as "Join API
> > rework
> > > for streaming". I do prefer the former one, because it is not only more
> > > straight-forward to understand, but a bit shorter as well.
> > > Of course there are counter-examples, like "[streaming] DataStream
> > > refactor" -> "DataStream refactor".
> > >
> > > I can accept optional, but would like to keep it strongly recommended
> for
> > > streaming. I also find the [docs] tag helpful.
> > >
> > > On Wed, Jan 7, 2015 at 2:43 PM, Stephan Ewen <sewen@apache.org> wrote:
> > >
> > > > Should we put that to an official vote, or wait for people to comment
> > and
> > > > (if nobody objects) consider it as agreed on through lazy consensus?
> > > >
> > > > On Wed, Jan 7, 2015 at 2:34 PM, Márton Balassi <
> > balassi.marton@gmail.com
> > > >
> > > > wrote:
> > > >
> > > > > +1 for the guide, thanks for clarifying the issue
> > > > >
> > > > > On Wed, Jan 7, 2015 at 2:30 PM, Till Rohrmann <
> trohrmann@apache.org>
> > > > > wrote:
> > > > >
> > > > > > +1
> > > > > >
> > > > > > On Wed, Jan 7, 2015 at 12:41 PM, Aljoscha Krettek <
> > > aljoscha@apache.org
> > > > >
> > > > > > wrote:
> > > > > >
> > > > > > > Yes, we should have a guide like that somewhere.
> > > > > > >
> > > > > > >
> > > > > > > On Wed, Jan 7, 2015 at 12:33 PM, Stephan Ewen <
> sewen@apache.org>
> > > > > wrote:
> > > > > > >
> > > > > > > > We have not exactly defined this so far, but it is
a good
> point
> > > to
> > > > do
> > > > > > so.
> > > > > > > >
> > > > > > > > I personally find it good to have changes associated
with an
> > > issue,
> > > > > > > because
> > > > > > > > it allows you to trace back why the change was done.
> > > > > > > > To make sure we do not overdo this and impose totally
> > unnecessary
> > > > > > > overhead,
> > > > > > > > I would suggest the following:
> > > > > > > >
> > > > > > > > *No issue is required for*
> > > > > > > >
> > > > > > > >   - Small fixes like typos, simple warnings,
> adding/improving a
> > > > > comment
> > > > > > > >
> > > > > > > >   - Adding and improving existing pages of the documentation
> > > > > > > >
> > > > > > > >   - Simple improvements of style / elegance / efficiency
> > (simple
> > > > > > > rewriting
> > > > > > > > a loop / condition / method interaction) if no behavior
is
> > > changed
> > > > > > > >
> > > > > > > > ==> Basically anything that does not change or
add
> > functionality
> > > > > > > >
> > > > > > > >
> > > > > > > > *An issue is required for*
> > > > > > > >
> > > > > > > > Everything else, in particular:
> > > > > > > >
> > > > > > > >   - Anything that changes functionality or behavior
relevant
> to
> > > > users
> > > > > > > >
> > > > > > > >   - Anything that changes functionality or behavior
relevant
> to
> > > > other
> > > > > > > > components
> > > > > > > >
> > > > > > > >   - Anything that adds a feature
> > > > > > > >
> > > > > > > >
> > > > > > > > I would vote to allow coarse issues and have multiple
commits
> > > that
> > > > > > > > reference it
> > > > > > > >
> > > > > > > > [FLINK-1234] [runtime] Runtime support some cool new
thing
> > > > > > > > [FLINK-1234] [java api] Add hook for cool thing to
java api
> > > > > > > > [FLINK-1234] [scala api] Add hook for that thing to
scala api
> > > > > > > > [FLINK-1234] [optimizer] Make optimizer aware that
it can
> > exploit
> > > > > this
> > > > > > > > thing
> > > > > > > >
> > > > > > > >
> > > > > > > > -------------------------
> > > > > > > >
> > > > > > > > The guide lines for pull-requests for committers are
as
> > follows:
> > > > > > > >
> > > > > > > > *A pull request with comments/additional signoff is
required
> > for
> > > > > > anything
> > > > > > > > that*
> > > > > > > >
> > > > > > > >   - breaks the public APIs
> > > > > > > >
> > > > > > > >   - adds methods to the public APIs (that will need
to be
> kept
> > > > stable
> > > > > > > from
> > > > > > > > them on)
> > > > > > > >
> > > > > > > >   - alters user-facing behavior (e.g., mutability
of types,
> > null
> > > > > value
> > > > > > > > handling, window semantics, ...)
> > > > > > > >
> > > > > > > >   - adds user-facing knobs (switches, config parameters,
> > > execution
> > > > > > option
> > > > > > > > on the execution environment)
> > > > > > > >
> > > > > > > >   - adds additional maven dependencies
> > > > > > > >
> > > > > > > >   - changes the way components interact
> > > > > > > >
> > > > > > > >   - touches highly sensitive and performance critical
parts,
> > such
> > > > > > memory
> > > > > > > > management or network stack
> > > > > > > >
> > > > > > > > ==> Changes that come with a pull request should
have one or
> > more
> > > > > > issues
> > > > > > > > associated with them.
> > > > > > > >
> > > > > > > >
> > > > > > > > Anyone that wants to have comments or some additional
pairs
> of
> > > eyes
> > > > > in
> > > > > > > the
> > > > > > > > code should make a pull request as well.
> > > > > > > >
> > > > > > > > -------------------------
> > > > > > > >
> > > > > > > > *Naming scheme for commits*
> > > > > > > >
> > > > > > > > [issue] [component] Message
> > > > > > > >
> > > > > > > > For fixes without an issue, the issue can be dropped.
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > > What do you think? Should we put this into the Wiki?
> > > > > > > >
> > > > > > > >
> > > > > > > > Greetings,
> > > > > > > > Stephan
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > > On Wed, Jan 7, 2015 at 11:48 AM, Aljoscha Krettek
<
> > > > > aljoscha@apache.org
> > > > > > >
> > > > > > > > wrote:
> > > > > > > >
> > > > > > > > > Hi,
> > > > > > > > > I feel we never really talked about this. So,
should we
> open
> > > Jira
> > > > > > > issues
> > > > > > > > > even for very small fixes and then add the ticket
number to
> > the
> > > > > > commit?
> > > > > > > > Or
> > > > > > > > > should we just commit those small fixes. Right
now, I have
> > two
> > > > > small
> > > > > > > > fixes
> > > > > > > > > (one is 4 lines, the other one is two lines)
for the
> > > > ValueTypeInfo
> > > > > > and
> > > > > > > > > TextValueInputFormat. Very obscure stuff, I know.
:D
> > > > > > > > >
> > > > > > > > > Cheers,
> > > > > > > > > Aljoscha
> > > > > > > > >
> > > > > > > >
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message