The dependency issue could be solved separately in a much simpler change.
I agree improvements could be made for 2.8 testing but it is a big project
and I think it might be more worthwhile to have 3.x sooner.
Anyway we can discuss this further after we release the source.
On Fri, Feb 16, 2018
Several things I'd like to achieve:
* I'd like to support JUnit 4 in GWT2 (well, I'd actually like to support
JUnit 5 in both, but one step at a time) - JUnit 3 practically doesn't
exist outside of GWT2.
* There are probably better ways to do testing than the existing junit
shell wiring in
I haven't thought about separating it from J2CL, not sure if it is a good
idea.
Why do you need it for GWT2? For code sharing, we simply use Junit3 style
and we didn't feel much pain about it.
On Thu, Feb 15, 2018 at 12:15 PM Colin Alworth wrote:
> Anything we can do to
Anything we can do to facilitate this? Also, if it is released under j2cl,
is there an expectation that it could be factored out and made to work with
GWT2, and released generally instead of just under J2CL (in its binaries
and its deadlines)?
On Friday, January 26, 2018 at 9:17:39 PM UTC-6,
Yes, that's correct.
It is a simple system with two part generation.
First part in bazel that generates an empty java class with a class
annotation that points to JUnit suite to trigger APT.
The second part in APT that generates the wrapper around junit3/junit4
tests that are listed in the suite
Correct me if I'm wrong, this is relying on JUnit 4 suited to generate
appropriate goog.testing code, JUnit 3 test cases (GWTTestCase basically,
possibly simply TestCase), with a "new" emulation of those classes based on
JsInterop to goog.testing, right? (goog.testing behaving similar to JUnit
+dramaix +dankurka
It doesn't support runners nor rules; just the basic stuff.
@dramaix: Let's release the source of J2CL junit emulator under the J2CL
repo. We don't need to wait until it works for bazel.
On Fri, Jan 26, 2018 at 12:02 PM, Colin Alworth wrote:
> As we look
As we look at how to keep migrating GWT provided user modules to modern
best practices, the JUnit runner is a hairy one - it relies on a lot of the
practices we discourage, including a great many modules in GWT that we
might prefer it not use. It also assumes
As has been discussed on other