-
-
Notifications
You must be signed in to change notification settings - Fork 346
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Generalize handling of source folders in cross-platform/version scenarios #2531
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
file needs scalafmt
@@ -41,7 +41,13 @@ trait JavaModule | |||
override def zincWorker: ZincWorkerModule = outer.zincWorker | |||
override def skipIdea: Boolean = outer.skipIdea | |||
override def runUseArgsFile: Target[Boolean] = T { outer.runUseArgsFile() } | |||
override def sources = T.sources { | |||
for (src <- outer.sources()) yield { | |||
PathRef(this.millSourcePath / src.path.relativeTo(outer.millSourcePath)) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
What if the user added additional source folders, maybe pointing outside of outer.millSourcePath
?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Seems reasonable to mirror them inside test
I think? That's what I'm finding as I port builds over, e.g. maybe someone added src-2.13+/
or src-js-jvm/
, and it would make perfect sense to mirror them inside the test
submodule. Especially v.s. the current status quo, where the source folder layout of test folders is pretty manual and thus ad-hoc, "same as parent module" seems like a big simplification both in UX as well as internal machinery
We could filter these folders somehow if we really want, but IMO it's not necessary. In most common cases the only downside would be the test
module gets a few unused-but-reasonable source folders, and worst comes to worst if someone wants to change this mirroring they can always override it, which is what the CrossSbtModule#Tests
already do
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
One more consideration is that the current approach of shadowing the Tests
trait does not compose. While it works for one use case in CrossScalaModule
, you cannot stack a second use case in PlatformScalaModule
and have them work together to define the final sources list
We also already delegate a lot of targets from test module to host module, so delegating another one would fit right in
Originally bumped into these issues trying to upgrade fansi/os-lib/upickle to 0.11.0-M9
Generalize
CrossScalaModuleTests#sources
toJavaModuleTests#sources
.CrossScalaModule
inCrossScalaModuleTests
, we instead define that allTests
module within aJavaModule
of any kind most follow the parent module's source layout, relativized to the test module'smillSourcePath
.CrossScalaModule
,PlatformScalaModule
, etc. without needing it to be aware of every combination of such cross-modulesGeneralize
PlatformScalaModule#sources
to work for cross modules where the cross segment is the last one, i.e.foo.jvm[2.13.10]
vsfoo[2.13.10].jvm
Tested both changes with an additional example test included in the documentation. This demonstrates an alternative approach to that in
example/web/6-cross-version-platform-publishing/
, one commonly used in the com-lihaoyi libraries, so I think it's worth making a full example v.s. just testing this via unit tests