Re: we now have "lilypond" organization on GitHub

[Joseph Rushton Wakeling]

On 24/09/13 14:14, Graham Percival wrote:
> Umm, the whole point of the VM is to ensure that the contributor's
> setup is *right*.

As far as I can see, the whole point of the VM is to get round the fact that the 
range of environments you can use to hack on Lilypond is severely restricted.

I'm suggesting to you that this restriction is in itself a problem.

> That's way more fragile than having them use a VM.

I don't know if I understand the history here.  Were custom package repos tried 
as a first solution before the VM?  Was it a practical history of failure, or a 
pre-judgement?

I accept that faced with the situation of most devs being on Windows but Windows 
being impossible as a dev environment, a VM must be the only reliable solution.

> > (5) In the specific case of git-cl, my own experience was that it
> > was an absolute pain.  Multiple (custom) commands to be typed;
>
> err, the whole point of the VM is that we provided a GUI to take
> care of patch submission.
>
> > Compare that with GitHub: push patches to my own repo.  Click on
> > "Submit pull request."  Type a brief description and click a button
> > ... done.
>
> and then somebody else needs to check if it compiles or not.

Yes, and repositories can be set up to automatically run test suites on 
submitted pull requests.  This is how it should be done -- it makes no sense for 
a contributor to have to take any manual responsibility for submitting to the 
testing system.

Of course, it's a good habit as a contributor to run the test suite on one's own 
machine before submitting a pull request (I always did this with LP submissions, 
even though they were docs not code).  I imagine that one problem of using a VM 
is that it makes it much more difficult/slow to run such local tests?