Forward port sbt/sbt#3701, JDK9/Scala 2.{10,11} overcompilation #450
Conversation
| } binaryDependency(classFile, binaryClassName) | ||
| jarFile <- Option(zip.file) | ||
| if !jarFile.isDirectory // workaround for JDK9 and Scala 2.10/2.11, see https://github.com/sbt/sbt/pull/3701 | ||
| } binaryDependency(jarFile, binaryClassName) |
There was a problem hiding this comment.
This is the only change that fixes an observable bug. I didn't figure out why the bug in SBT 0.13 fixed by -Dscala.ext.dirs was not necessary here.
There was a problem hiding this comment.
Can you elaborate why Scalac reports the underlying source to be a directory under JDK9? I'm curious.
There was a problem hiding this comment.
To integrate the JEP-220 support into Scala 2.10/2.11, I made the jrt:// virtual file system appear as a ZipFile. This was done by subclassing, but the superclass had inadvertently overriden override val underlyingSource = Some(jarFile). I could not override this to return None without making a binary compatible change to the method I was overriding (widening to Option).
So I return Some($JAVA_HOME) instead, but that messed up Zinc.
Rocks and hard places, all around. Filtering here in Zinc seemed the least bad option for now, but we might need to revisit it to make Zinc properly invalidate after the JDK is upgraded.
There was a problem hiding this comment.
Great to know this, thanks for the explanation.
retronym
force-pushed
the
ticket/dirDependency
branch
3 times, most recently
from
152a6c3 to
5812284
Compare
| @@ -13,6 +13,7 @@ import xsbti.ArtifactInfo | |||
| import scala.util | |||
| import java.io.File | |||
| import CompilerArguments.{ absString, BootClasspathOption } | |||
| import sbt.IO | |||
There was a problem hiding this comment.
This is causing a compilation error and can be removed.
| def findBoot: String = { | ||
| import scala.collection.JavaConverters._ | ||
| System.getProperties.asScala.iterator.collectFirst { | ||
| case (k, v) if k.endsWith(".boot.class.path") => v |
There was a problem hiding this comment.
Have you done this to avoid making it specific to scala.boot.class.path? Is there any other valuable system variable that users .boot.class.path as the suffix?
There was a problem hiding this comment.
The code is from scalac. I believe the intent is to support non OpenJDK based JVMs that don't use sun.boot.classpath.
https://bugs.eclipse.org/bugs/show_bug.cgi?id=188648#c6
http://mail.openjdk.java.net/pipermail/build-dev/2013-November/011103.html
| } binaryDependency(classFile, binaryClassName) | ||
| jarFile <- Option(zip.file) | ||
| if !jarFile.isDirectory // workaround for JDK9 and Scala 2.10/2.11, see https://github.com/sbt/sbt/pull/3701 | ||
| } binaryDependency(jarFile, binaryClassName) |
There was a problem hiding this comment.
Can you elaborate why Scalac reports the underlying source to be a directory under JDK9? I'm curious.
retronym
force-pushed
the
ticket/dirDependency
branch
from
5812284 to
1c8eefb
Compare
retronym
force-pushed
the
ticket/dirDependency
branch
from
1c8eefb to
641a491
Compare
|
This is good to go on my side. |
There was a problem hiding this comment.
I'm happy with this. And given @eed3si9n merged the 0.13 PR sbt/sbt#3701 my guess is he's cool with this too.
No description provided.