On Wed, Mar 23 2016, Ian Jackson wrote:
> Marco d'Itri writes ("Re: a poll for Dgit workflows"):
>> I like the general idea of dgit, but I will never use it as long as it
>> requires committing patched trees.
>
> Obviously, for dgit to be useful, it has to define a standard
> interchange format. That format has to be patches-applied because
> otherwise naive users can't work with the source code properly.
It’s not just naive users that prefer working with git and
branches natively. My preferred form of modification is pull requests
for my git repositories (my packaging repositories are mirrored on
alioth and github). I can build and test the tip of any of my feature
branches or upstream, or the fully merged debian branch, so I can make
sure that the features I submit upstream actually work. I think that is
harder to do if the repository has just a jumble of patches in a
directory, and is missing the development history of the features.
While git-debcherry does automate the pain
point of serializing commits on different feature branches, the
commits are out of order for each feature, and it is a lot of busy work
getting the series of patches properly ordered, and there are always
merge artifacts in the patch series as slightly overlapping branches
are brought back into conformance.
I prefer not to get patches to the source predicated on how the
feature branches have been serialized; most people in this era of
github can and do work with remote git repositories, and others can
still send me patches by mail that I’ll break down into changes to a
fix branch that I shall send upstream.
Given the number of times that most projects get sent changes
versus the maintainer having to work with version control themselves, I
think that debian-single-patch is acceptable (well, in my opinion, of
course).
manoj
--
Today is the first day of the rest of your life.
Manoj Srivastava <srivasta@debian.org> <http://www.debian.org/~srivasta/>
4096R/C5779A1C E37E 5EC5 2A01 DA25 AD20 05B6 CF48 9438 C577 9A1C
Attachment:
smime.p7s
Description: S/MIME cryptographic signature