Am Mi., 12. Sep. 2018 um 22:01 Uhr schrieb Chris Brody < [email protected]>:
> On Wed, Sep 12, 2018 at 3:48 PM <[email protected]> wrote: > > > > Am Mi., 12. Sep. 2018 um 21:36 Uhr schrieb Jan Piotrowski < > > [email protected]>: > > > > > This uses specific nightlies the developer has to select manually, > right? > > > > > I would say it would have to, since nightly alone would be an unstable > > moving target that would regularly break our CI tests. > > I would disagree. I thought that work in master branch *should* > already be considered to be a non-stable "dev" version. If a change in > a lower-level dependency would causes another Cordova package to stop > working, I would rather we discover and resolve this asap. > Well then we agree to disagree ;) IMHO, CI tests should run against a known-to-work state, so we can reliably see from the CI tests if a __PR__ breaks the tests or not. If we want to test integration between Cordova components, we should run tests with all the master branches linked together during our nightly builds. Plus, with package-lock.json committed no version update would be picked up automatically. That's the idea behind committing it in the first place. > > > How do other libraries handle this inter-connectedness between > libraries? > > > > You could have a look at Babel. They have a lot of packages (managed by > > lerna AFAIK) > > Does that mean we should consider migrating to some kind of a monorepo? > No, I was not suggesting anything. Just mentioning that they used it.
