cc @IgorMinar

The saucelabs failure is a flake and just needs to be restarted.

A refactor doesn't change behaviour or break compatibility. Behaviour is being changed here and there's nothing being fixed, so I guess this is a feat commit.

@alfaproject Yeah I don't know. It's literally the opposite of a feat though, since it's removing something from the publicApi. And yet, technically, it's removed from the publicApi when it's deprecated. I'll wait until @IgorMinar weighs in, but I have no issue changing the scope to whatever's most appropriate.

I expect that this will break a significant amount of code at google. I'm running a presubmit now to verify the impact.

If confirmed, we might need to create a ts-lint refactoring transform to make this change automatically and roll this out as part of ng update for v8.

@StephenFluin can we get a reliable estimate of this change on external users? I'm quite sure it's significant and therefore we could roll this out only with automated refactoring of apps and libraries (which in this case is safe and relatively straightforward).

internal presubmit results: http://test/OCL:229126105:BASE:229126324:1547447183003:cc4ee4a9

preliminary results: there is a handful of apps that still work after this change, but great majority of apps in google3 are broken.

this means that we'll need the ts-lint transform similar to several we built for rxjs in order to roll this change out. @CaerusKaru are you up for taking that on?

Now in order to make the transform compatible with ng update, we'll need a separate package that will contain the rule and the dependency on ts-lint (we can't add a dependency on ts-lint from @angular/platform-browser). So I'm proposing that we create @angular/platform-browser-compat (or possibly a more generic @angular/compat package), similar to rxjs-compat, that will contain the ts-lint transform and this package will declare the dependency on ts-lint.

We then write a ng update schematics for @angular/platform-browser that will install this compat package (and tslint), execute the refactoring, and remove the compat package once successful.

and yes, "refactor" is misleading. we might need a new category for this type of change, but in the meantime use "feat" as weird as it is because it's best category in terms of user visibility of the change in the changelog.

Around 80% of apps I just scanned use the DOCUMENT InjectionToken. I can't tell from compiled source where they import it from though, so I have a hard time guessing how many will be affected.

I wouldn't have guessed that it would be many people, as I would guess most developers access document directly and assume web-only use cases.

Blocked on #29237

Also blocked on angular/components#15470

bump @alexeagle @IgorMinar

This can be merged now that #29950 has been merged

This issue has been automatically locked due to inactivity.
Please file a new issue if you are encountering a similar or related problem.

Read more about our automatic conversation locking policy.

_{This action has been performed automatically by a bot.}