Re: Git migration docs

["Jason D. Clinton" <me jasonclinton com>]

On Thu, Jan 15, 2009 at 3:57 PM, Murray Cumming <murrayc murrayc com> wrote:
> On Thu, 2009-01-15 at 22:32 +0100, Mathias Hasselmann wrote:
>> So maybe we should just drop the old habit of creating large self
>> contained commits which require exhaustive ChangeLog entries: Micro
>> patches that focus on just one change are much easier to review.
>
> But shouldn't interdependent changes be reviewed together? I generally
> find that these many mini-commits make it hard for me to see the big
> picture. I tend to wish that the person had combined the many
> local-commits into one atomic commit when pushing it to the main
> repository.

If they are interdependent in the sense that they are required to make
the whole aggregate change build-able, they should be squashed down to
one commit before being pushed upstream/submitted. git-rebase in other
projects is used for this purpose. This is largely a continuation of
the "if it doesn't build, don't commit it" mentality.

On the kernel, for example, patches tend to be independent in the
sense that each change is atomic and would be justified on its own
accord; however, a patch series tends to lean toward achieving an
aggregate objective. This scenario tends to apply to implementing new
features that touches many pieces.

Obviously, good maintainer review of patches is the only way this kind
of quality culture can be achieved. It's a social problem, not a
technical one.