[ 
https://issues.apache.org/jira/browse/GROOVY-12627?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18135688#comment-18135688
 ] 

ASF GitHub Bot commented on GROOVY-12627:
-----------------------------------------

codeconsole opened a new pull request, #3153:
URL: https://github.com/apache/groovy/pull/3153

   In an inner class, a property reached through the enclosing instance could 
be read from a closure, but writing it threw `MissingPropertyException` unless 
the receiver was a plain `this`:
   
   ```groovy
   class Base {
       private String z = 'outer'
       class Sub extends Base {
           def m() { { -> Sub.this.z = 'w' }(); Base.this.z }
       }
       def t() { new Sub().m() }
   }
   assert new Base().t() == 'w'   // MissingPropertyException: No such 
property: z for class: Base$Sub
   ```
   
   The same failure hit `s.z = 'w'` with `def s = this`, compound assignment 
and `++`, nested closures and lambdas, non-subclass inner classes 
(`Inner.this.z = 'w'`), and spread-dot reads and writes in a closure. Groovy 
4.0.33, 5.0.8 and 5.1.3 return `'w'`; 6.0.0 throws, since GROOVY-11823 replaced 
the generated inner-class `propertyMissing` with a metaclass fallback.
   
   A call site in a closure passes the closure class as its sender. `this.z` is 
not affected because the call site is flagged as a `this` call and the sender 
is moved out of the closure. A qualified `this` or a local variable is not 
flagged. `MetaClassImpl.getOuterProperty` and `MetaClassImpl.setOuterProperty`, 
the enclosing-instance fallback from GROOVY-11823, checked the nesting of the 
closure class and never applied. They now move the sender out to the class that 
declares the closure before checking.
   
   ## Tests
   
   `InnerClassTest.testNestedPropertyHandling8` and 
`InnerClassTest.testNestedPropertyHandling9` (dynamic and `@TypeChecked`) fail 
without the change. The `gls.*`, `bugs.*`, `groovy.lang.*`, 
`org.codehaus.groovy.runtime.*`, `org.codehaus.groovy.vmplugin.*`, 
`groovy.transform.stc.*` and `org.codehaus.groovy.classgen.asm.sc.*` tests 
pass, as do `groovy.ClosureTest` and `groovy.ThisAndSuperTest`. The Checkstyle 
and CodeNarc gates pass.
   




> Writing a property of the enclosing instance from a closure in an inner class 
> throws MissingPropertyException unless the receiver is a plain this
> -------------------------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: GROOVY-12627
>                 URL: https://issues.apache.org/jira/browse/GROOVY-12627
>             Project: Groovy
>          Issue Type: Bug
>    Affects Versions: 6.0.0
>            Reporter: Scott Murphy Heiberg
>            Priority: Major
>
> {code:groovy}
> class Base {
>     private String z = 'outer'
>     class Sub extends Base {
>         def m() { { -> Sub.this.z = 'w' }(); Base.this.z }
>     }
>     def t() { new Sub().m() }
> }
> assert new Base().t() == 'w'
> {code}
> Actual: groovy.lang.MissingPropertyException: No such property: z for class: 
> Base$Sub
> Expected: 'w'. The write goes to the enclosing instance, as this.z = 'w' in 
> the closure does, as Sub.this.z = 'w' outside the closure does, and as 
> reading Sub.this.z in the closure does. Groovy 4.0.33, 5.0.8 and 5.1.3 return 
> 'w'.
> The same happens for s.z = 'w' with def s = this (also under @TypeChecked), 
> for compound assignment and ++, for nested closures and lambdas, for a 
> non-subclass inner class (Inner.this.z = 'w'), and for a spread-dot read or 
> write in a closure.
> A call site in a closure passes the closure class as the sender. The 
> enclosing-instance fallback that GROOVY-11823 added to MetaClassImpl checks 
> the nesting of that closure class rather than of the class that declares the 
> closure, so it never applies there.
> Seen in 6.0.0 (regression); 4.0.33, 5.0.8 and 5.1.3 work.



--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

Reply via email to