> On 9 Nov 2016, at 12:51 pm, René J.V. Bertin <[email protected]> wrote:
> 
> Hi,
> 
> I've just managed (I think) to fork macports-ports via github.com, add it as 
> an additional remote to my working copy of the original, created a topic 
> branch in my fork and made a pull request from there.
> 
> Questions raised during that process:
> 
> 1- Initially I followed github's suggestions as usual and added a README.md 
> (in a first commit), thinking I'd be able to avoid that file easily enough. 
> Instead it appears that pull requests can not be made for a specific commit 
> or file/directory; README.md was an unintended part of my request. When I 
> removed the file the pull request showed the addition and removal and I'm 
> quite sure the same would have been true for a reverse commit. I managed to 
> back out and recreate the branch, but it's highly annoying you only find out 
> such things after doing a commit.
> Is there a way around this, or are pull requests always a reflection of the 
> difference between the original master/head and the fork's head? IOW, is it 
> going to be necessary to use 1 branch per pull request? That'd make it very 
> impractical to use a single ports tree as both a local source for installed 
> ports and a source for pull requests …

In my view, no it is not practical. Pull requests are to pull one branch, all 
diffs, from one to another. This is why I maintain the sooner people get use to 
the idea of making a separate branch for each piece of work, and pull request, 
the faster they will make progress with working with git. This is actually the 
power of git, not a hindrance. But it takes time for newcomers to git to 
realise this ;)

> 
> 2- Suppose it *is* possible to have all local changes and custom ports in a 
> single personal branch and then create pull requests when time and port are 
> ripe. IIRC there's a magic incantation to rebase a topic branch on the remote 
> origin/master (i.e. merge in remote changes on top of the changes in your 
> topic branch without losing its history). Of course I haven't been able to 
> retain that formula. It would be useful to describe the procedure on the 
> working-with wiki.
> 
> Thanks,
> R.

Attachment: smime.p7s
Description: S/MIME cryptographic signature

Reply via email to