Update build matrix for Java 14 #1571
Conversation
This allows us to check that ErrorProne works on Java 12. It is part of issue google#1106 to make ErrorProne work on newer versions of Java. It will allow us to verify that google#1110 indeed fixes the problem with Java 12.
|
12 is EOL since last September, 13 for nearly one month already. Shouldn't this use JDK 14 instead? FWIW google-java-format doesn't even try to be compatible with 13: google/google-java-format#436 (comment) |
|
It should always use the latest version of Java. That means at minimum an update every 6 months. |
|
The compilation error I am currently hitting is a method signature changed in JDK 12 but remains in JDK 14. So we will need to fix that also for JDK 14. We are discussing mitigations in an internal thread. |
TimvdLippe
changed the title
|
You may also want to migrate to Xenial or even Bionic rather than Trusty, and openjdk8 rather than oraclejdk8 |
Note: it is expected that this will fail on JDK 8 and 11.
This is not strictly necessary for JDK 14 and we want a minimal diff at this point in time.
| } | ||
|
|
||
| public abstract JCPattern getPattern(); |
There was a problem hiding this comment.
This method is problematic, as JCPattern has only been introduced in JDK 14. It implements the interface com.sun.source.tree.PatternTree, which seems like a marker interface: http://hg.openjdk.java.net/jdk/jdk14/file/6c954123ee8d/src/jdk.compiler/share/classes/com/sun/source/tree/PatternTree.java
I am not sure how we can implement this interface on JDK 14 without breaking JDK 8/11.
| sym.type = typeVar; | ||
| typeVarCache.put(var.getName(), typeVar); | ||
| // Any recursive uses of var will point to the same TypeVar object generated above. | ||
| typeVar.bound = var.getUpperBound().inline(this); |
There was a problem hiding this comment.
These fields are removed and are instead put in the constructor. Looking at the code, it seems to make sense to put them in the constructor in the first place, since we would still override them. This should be 8/11-compatible.
| @@ -282,7 +282,7 @@ private static int getPrecedence(JCTree leaf, Context context) { | |||
| return (typeCast.expr == leaf) ? TreeInfo.prefixPrec : TreeInfo.noPrec; | |||
| } else if (parent instanceof JCInstanceOf) { | |||
| JCInstanceOf instanceOf = (JCInstanceOf) parent; | |||
| return TreeInfo.ordPrec + ((instanceOf.clazz == leaf) ? 1 : 0); | |||
| return TreeInfo.ordPrec + ((instanceOf.getType() == leaf) ? 1 : 0); | |||
There was a problem hiding this comment.
Similar as below, the field has been removed and we should use getType() (which is a getter for the field).
| return result.exists() ? result : null; | ||
| } catch (ReflectiveOperationException e) { | ||
| throw new LinkageError(e.getMessage(), e); | ||
| } | ||
| } | ||
|
|
||
| private static final class FindIdent { |
There was a problem hiding this comment.
Pulled all of this in from #1439 to fix a lot of test failures.
| @@ -665,7 +665,8 @@ public Boolean visitIdentifier(IdentifierTree node, Unifier unifier) { | |||
| return chooseSubtrees( | |||
| state, | |||
| s -> unifyStatements(node.getStatements(), s), | |||
| stmts -> maker().Case((JCExpression) node.getExpression(), stmts)); | |||
| s -> unify(node.getBody(), s), | |||
| (stmts, body) -> maker().Case(CaseTree.CaseKind.STATEMENT, List.<JCExpression>of((JCExpression) node.getExpression()), stmts, body)); | |||
There was a problem hiding this comment.
We have to provide 2 additional arguments on JDK 14. It seems like this should be fixed similarly to the FindIdentifiers solution (e.g. reflection).
Wasn't actually necessary for Java 14
|
CI is now passing for 12 and 13: 937b3d2. (14+ is still WIP: #1106 (comment)) |
This allows us to check that ErrorProne works on Java 12. It is part of
issue #1106 to make ErrorProne work on newer versions of Java.
It will allow us to verify that #1110 indeed fixes the problem with Java
12.