> I propose to fail the Fennec build if this middle set grows. +1 from me.
The current situation is unpleasant for consumers of GeckoView, because they’re pretty much taking a dependency on Fennec as a whole. We’ll need to cut our Gecko-side stuff down, too (a long project) but we should be aggressively working to minimize the Java-side dependencies of GeckoView users. I expect it’ll also expose some design warts in Fennec, which can only be a good thing. > * This will almost certainly make a small portion of Fennec > development more challenging, because you'll need to consider > whether the feature you're building is part of GeckoView and you'll > need to engineer certain features to be split between GeckoView and > Fennec. We should be doing that anyway. I’m all in favor of compiler errors when we’re being egregiously lazy. > For an even trickier example, GeckoProfile caches instances of > itself on GeckoApp instances. It's not immediately clear what > should own the GeckoProfile instance in a GeckoView App; no matter > what is chosen, there are all sorts of lifecycle issues to work > through in order to do something different than what is done now. Wes and I are tackling that (slowly) in Bug 1077590, Bug 1080038. > org.mozilla.gecko.Telemetry \ > org.mozilla.gecko.TelemetryContract \ Yeesh. _______________________________________________ mobile-firefox-dev mailing list [email protected] https://mail.mozilla.org/listinfo/mobile-firefox-dev
