That isn't the question.  The presumption is that you do not have a set of
patches to reorganize.  You have a set of files that should be replaced by
a different set of files and from that you want to organize the diff into a
logical sequence of patches on the assumption that a single patch that
effected the diff would be too big.

On Thu, Nov 20, 2014 at 3:43 AM, Taras Kondratiuk <
[email protected]> wrote:

> On 11/19/2014 04:15 PM, Bill Fischofer wrote:
> > Slightly off-topic, but relevant to ODP development and something we
> > just discussed in the arch hangout.  But this captures the problem I was
> > trying to articulate.  I'd appreciate thoughts and recommendations for
> > best practices as to how to deal with this.
> >
> > Say I have a repository with a master branch that consists of a set of
> > files with their own commit histories.
> >
> > Now assume that I'm given another set of files that differ from some
> > subset of the master files in varying ways.  How these files came to be
> > is not relevant to this question.
> >
> > If I do a diff between the master and the new files what I'd have is a
> > single patch that transforms the master files into new versions that
> > reflect the new files, however this patch may be too large to be
> > conveniently reviewed/digested.
> >
> > So the question is this.  How can I best reorganize this single large
> > patch into a series of orthogonal patches that can be submitted against
> > the master that combined produce the same results as the original giant
> > patch?
> >
> > I believe this scenario captures the essence of the problem Ola and I
> > were trying to articulate.
> >
> > Thanks for any insights and tool suggestions for this.
>
> This example missing important details like:
> - Are these changes made on top of the same commit you want to apply
>   them to?
> - Are these new changes committed as one big patch or they are still
>   in your working tree?
>
> Depending on these details first steps will be different, but anyway
> you need to come to a state when you have all your changes unstashed in
> a working directory.
> Then with 'git add -p' pick changes related to a first patch.
> Commit them.
> Pick changes for a next patch and commit them.
> Proceed like this until you commit everything.
>
> Important thing to mention: situation described here can be avoided by
> sorting patches at least partially during development. That takes just
> a bit more time, but a the end changes are split into major pieces, so
> converting them to a clean patch series is much easier.
>
_______________________________________________
lng-odp mailing list
[email protected]
http://lists.linaro.org/mailman/listinfo/lng-odp

Reply via email to