RFR: JDK-8243047: javac may crash when processing exists in class initializers

[Vicente Romero]

Hi Jan,

The fix looks good, there seems to be an issue in the combo test though 
at enum: Block, there is no difference in the arguments of STATIC and 
INSTANCE and in the arguments of STATIC_INITIALIZER and INITIALIZER,

Thanks,
Vicente

On 4/17/20 5:25 AM, Jan Lahoda wrote:
> Hi,
>
> For code like:
> ---
> public class ExitInInitializer {
>     {return;}
> }
> ---
>
> If javac is invoked like:
> $ javac -XDshould-stop.at=FLOW -XDdev ExitInInitializer.java
>
> javac eventually crashes with:
> java.lang.AssertionError
>         at 
> jdk.compiler/com.sun.tools.javac.util.Assert.error(Assert.java:155)
>         at 
> jdk.compiler/com.sun.tools.javac.util.Assert.check(Assert.java:46)
>         at 
> jdk.compiler/com.sun.tools.javac.comp.Flow$AliveAnalyzer.visitMethodDef(Flow.java:558)
>
> The reason is that the the (invalid) "return" in the intializer is 
> added to pendingExits, but not removed as the end of the initializer, 
> and when analyzing method, javac checks pendingExits are empty.
>
> The proposed solution is to clear pendingExits at the end of 
> initializers. The asserts that an error had to be reported for such 
> exits is preserved, so there shouldn't be a danger of accidentally 
> letting incorrect compilation pass.
>
> Webrev:
> http://cr.openjdk.java.net/~jlahoda/8243047/webrev.00/
>
> JBS: https://bugs.openjdk.java.net/browse/JDK-8243047
>
> How does this look?
>
> Thanks,
>     Jan