← Back to team overview

kicad-developers team mailing list archive

Re: suggestion: git branching model


On Wed, Dec 11, 2019 at 5:35 PM Seth Hillbrand <seth@xxxxxxxxxxxxx> wrote:
> On 2019-12-11 08:17, Jonatan Liljedahl wrote:
> > Ok, good to hear!
> >
> > How about moving the new symbol inheritance stuff into such a feature
> > branch until it's working? (See
> > https://gitlab.com/kicad/code/kicad/issues/3658) I understand it would
> > mean less testing, but in its current state it's impossible to test or
> > work on anything else since kicad keeps crashing.
> >
> This will happen sometimes.  We try to avoid it but we can't always
> foresee the crashes, especially with large features.  Frequently the
> crashes are actually due to the difficulty of rebasing and merging the
> branch.
> Asking the development team to keep the development branch bug-free is
> not realistic.  This issue will get resolved.  Until then, please use
> the older revision.

Of course, I'm not expecting the dev branch to be bug-free. I was just
thinking about git branching models and wondering if things could be
improved, specifically about larger feature branches. I didn't know
that the symbol inheritance branch was available for testing from
Waynes personal repo before merge, possibly because I'm new here and
just recently joined the dev-list. I'm sorry if I stepped on any toes!

Hopefully all this will be much easier now with GitLab, where everyone
can see and access available merge requests in a simple way.

I think the trick with feature branches is to rebase against master
very frequently, to avoid a pile of merge conflicts.

> If you are interested in helping test feature branches to minimize the
> crashes in the future, you might test and review the feature branch at
> https://gitlab.com/kicad/code/kicad/merge_requests/15 .  It will be
> merged soon but has not had any testing outside of my own at the moment.
>   This is the downside of feature branches in a volunteer-based
> development model.

I will take a look! However, since it's rebased on current master,
which is crashing a lot, it's hard to test and know what broke because
of which change.
When the current issues on master are sorted out and !15 rebased
against it, it will be easier.


Follow ups