Re: [Archer] Re: [Archer] Archer git account inquiry

[Tom Tromey <tromey@redhat.com>]

Pierre>   I don't really care about where I put my modifications,
Pierre> if nobody is against it, archer repository is fine with me.

It is perfectly ok.

Pierre>   Related question: should I change the wiki page
Pierre> to add my branch? Do I need to logon somewhere for this?

Yes, please document new branches on the wiki.
You need an account on the wiki, you can make one there somehow.

Pierre> Coming back to the primary reason of my request for git account:
Pierre> I am mainly interested in an easier way to split 
Pierre> larges patches into several sub-patches like many of you
Pierre> are doing.

Pierre>   So my main question is:
Pierre> what do you use to create your email series with sub-patches?
Pierre> -- git alone?
Pierre> or stgit, guilt or Quilt?

I write my patches willy-nilly, in the craziest style imaginable.

This is usually what I push to the public repository as well.

For submissions, I make a new private branch and extract bits from the
development branch in various ways -- sometimes cherry-picking,
sometimes patches -- and use git rebase to make a patch series.  For
this, the "Perfect Patch Series" instructions in the git manual are
ideal, I just follow them exactly.

After this, if I need to rework the patches, I just do it on my local
branch, this time rebasing, etc, to keep the patches clean.

Pierre> How do you handle ChangeLog entries for instance?
Pierre> do you really put them into the commits, but then
Pierre> they would appear formatted as a diff in the patch also
Pierre> or do you have special scripts for this?

Usually I ignore them until the point of patch submission, because they
are such a pain to deal with.

On occasion I have written my ChangeLog entries during the rebase
process, then just used them as the commit messages in git.  Then if I
need to rework the patch series, I also update the commit messages as I
go.

I don't usually modify my branch ChangeLog files since it is just a
pain.  It makes random conflicts if I rebase (there are some tools to
help with this, but I haven't bothered yet), and on the development
branch I often rework things, so just accumulating the ChangeLog entries
would yield the wrong result for the upstream submission anyhow.

Tom