Code-size difference on a large app is very small (0.007% worse), but
this cuts the time spent in type-propagation by more than half
(23s -> 10s on my machine).
I saw a 0.4% code-size regression in pop-pop-win, but that is
recovered once we are smarter about generating simple is-checks
when the "raw" type can be used because the type arguments
match the bounds (see separate CL).
Change-Id: I45c0fb1afabdca3d045aec576c131ed764dc31ce
Reviewed-on: https://dart-review.googlesource.com/c/94120
Commit-Queue: Sigmund Cherem <sigmund@google.com>
Reviewed-by: Stephen Adams <sra@google.com>
* All helpers in /helpers/
* Tests of user-facing behavior in /end_to_end/
* Tests of internal models and unittests of internal classes in /model/
Change-Id: Ic7a7e58aa01f92f8572514ae81230804f36af67e
Reviewed-on: https://dart-review.googlesource.com/72644
Commit-Queue: Johnni Winther <johnniwinther@google.com>
Reviewed-by: Stephen Adams <sra@google.com>
Sigmund Cherem