| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
return values, converted to JSON
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Remove method defns for zipMap/zip that will be generated.
The trait Applyer will now extend from ApplyerGenerated which will contain the code-generated definitions.
* Script to generate code for Applicative
* Generate the zip methods in Target
* Generate zip methods in ApplicativeTests
* Make sure the full 22-arities are generated
* Move code generator into build.sc.
Remove generate.sc from directories where code will be generated.
* Generate code as part of the SBT also
* Properly wire up the test sources
* Generate all the code in one place
|
|
|
|
| |
`HelloWorldTests`, to try and speed up `test.sh` runs
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
dependencies between `Ctx` -> `Discovered` -> `Task`
- `mill idea` works now using `GenIdea` as a standalone `T.command` making use of the new contextually-available `Mapping`
- Limit implicit `ReplApplyHandler` to `--repl` only, to avoid it kicking in if `build.sc` scripts are screwed up and adding further confusion
|
|
|
|
| |
Tasks, for usage within `GenIdea` and similar
|
|
|
|
| |
system
|
|
|
|
|
|
| |
master branch
Remove the `out/run.sc` entrypoint script, using Ammonite's new `codeWrapper` API to synthesize the necessary wrapper/forwarder objects to substitute it
|
| |
|
|\ |
|
| |
| |
| |
| | |
now you can publish your module with `mill run MyModule.publish --credentials $SONATYPE_CREDENTIALS --gpgPassphrase $GPG_PASSPHRASE`
|
| | |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
* Report errors when dependency resolution fails
Function `Lib.reolveDependencies` now checks the result of the coursier dependency
resolution for errors. The return value of the fuction is now a
`Result[Seq[PathRef]]` and errors are signaled by returning an instance
of Failure when errors occur while resolving dependencies.
`ScalaModule` now has a new task called `resolve` which allows for
triggering dependency resolution directly from the cmd line like:
`mill run resolve`.
`ResolveDepsTests` was added to verify failed resolution is properly
detected and reported.
* Remove type signatures in ScalaModule.scala
* Updates from PR comments
|
|/
|
|
| |
around their generated `Discovered.make` structures by hand, making it harder to accidentally generate the same `Discovered` twice and slow down runtime/compilation
|
|
|
|
|
|
|
|
| |
implementations, added a `Task.sequence` that does what `traverse` used to do
- Added a `test.sh` script to easily kick off self-hosted unit test runs
- Tweak `ScalaModule` to fall back to the old behavior of including the transitive classpath during compilation
|
|
|
|
| |
`codeWrapper` to inherit from `mill.Module`
|
| |
|
|
|
|
|
|
|
|
| |
keyword when overriding a field within a `mill.Module`
This only applies to `mill.Module`s, not overrides elsewhere which still require the keyword. `mill.Module`s tend to have lots and lots of overriding, so the keyword is basically noise.
Also includes the necessary build changes to enable the locally-built Scalac plugin when compiling the test suite. Note that no changes are necessary for the executable assembly, because the `scalac-plugin.xml` will be included in the assembly and get picked up by the Ammonite scalac plugin classloader automatically
|
| |
|
|
|
|
|
|
| |
managed by the `Evaluator`, which is now a stateful object that shouldn't be thrown away every time.
We still need to update the code/test-suite to make the `Evaluator` hang around in between `evaluate` calls
|
| |
|
|
|
|
|
|
|
|
| |
`ReplApplyHandler` keep using the same one every time to avoid busting caches due to REPL commands being added to the classpath
Reverts https://github.com/lihaoyi/mill/pull/60, which seems to break `mill.scalaplugin.AcyclicTests.scala2123` (reproducible in master)
Tweak `build.sbt` to properly set the forked test working directory in `test-only` as well as `test`
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
| |
query for terms like `Core.compile`, but still can't run `Core.compile()`
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
into Mill as JVM properties.
This makes the dependency between the compiler-bridge jar and the Mill executable explicit, allowing us to swap out the locations compiler-bridge jars (which end up in different places, depending on whether you're building with SBT or Mill) or eventually making them load from Maven Central in a "release" Mill executable
Since Mill (and uTest) both do not support SBT-style test arguments, we need to use `forkTest` instead of `test` to run the Mill tests passing the compiler-jar locations as JVM props. This necessitated some fixes to make `forkTest` behave properly
|
|
|
|
| |
`log.outputStream`
|
| |
|
| |
|
|
|
|
|
| |
- Convert `ScalaModule.{compile,assembly}` to use the new implicit `T.ctx()`
- Add a `log: PrintStream` to the `T.ctx()`, in preparation for per-task logging
|
|
|
|
|
|
|
|
|
|
| |
cross-minor-version `AcyclicTests` running using the `mill` test runner
Fixed `Discovered` to only pick up public `Target`s, which makes it skip things like the `super$compile` of an overriden `compile` target.
Split up `sources` and `allSources`
TBD how to properly switch the `compilerBridgeJar` between the different output paths of SBT's compile-dir and Mill's compile-dir, and eventually to a maven-central artifact too
|
|
|
|
| |
neg tests
|
|\ |
|
| |
| |
| | |
Inherited methods from Object can be checked for owner instead of looking at the name of the method. This should be cleaner and faster, though probably below the margin of error at this point.
|
| |
| |
| |
| | |
metadata.mill.json invalidation bug
|
|/
|
|
| |
directory, but only when run alone (???)
|
|
|
|
|
| |
Use the Ammonite classloader to check if the classpath has changed
Fix #7
|
| |
|
|
|
|
| |
Also workaround a few scala-reflect bugs in `Discovered` and `Router` that were making this not-work
|
|
|
|
| |
defaultCommandName member
|
| |
|