7122142 : (ann) Race condition between isAnnotationPresent and getAnnotations

[Joel Borggrén-Franck]

Hi Peter,

As Alan said, a big thanks for looking into this.

I have been toying with a slightly different approach to breaking the infinite recursion by pre-construct AnnotationType instances for Retention and Inherited. While cleaner in some places others became uglier. I'm fine with this approach.

As Alan said, you need to update the patch with a modification in TypeAnnotationParser. After doing that all java/lang tests in the jdk/test directory passes.

Also, can you please explain to me why the update race is safe. I have done the/some research myself but I would like to know which angles you have covered.

Given that and that you fix Alan's comments I think this is good to go.

Thanks again!

cheers
/Joel

On Jun 24, 2013, at 8:23 PM, Peter Levart <peter.levart at gmail.com> wrote:

> On 06/19/2013 08:54 PM, Alan Bateman wrote:
>> Thank you for coming back to this.
>> 
>> I've looked over the webrev and the approach looks good to me. Joel might want to look at this too. Do you think you could include a test (as we try to include a test with all fixes if we can)?
>> 
>> It would be good to remove the synchronizaiton on initAnnotationsIfNecessary too, but one step as time (and smaller changes are always easier to discuss).
>> 
>> -Alan
> 
> Hi Alan,
> 
> I have prepared the 2nd revision of the patch:
> 
> http://cr.openjdk.java.net/~plevart/jdk8-tl/AnnotationType/webrev.02/
> 
> There is a little change in AnnotationParser's alternative parsing method. This time the parser does not assume anything about annotations being parsed (previously it assumed that they had RUNTIME retention). The only difference from standard parsing is that only the select annotation types are parsed and the rest are quickly skipped. Infinite recursion is broken by the special cased evaluation in AnnotationType constructor:
> 
> 129         // Initialize retention, & inherited fields.  Special treatment
> 130         // of the corresponding annotation types breaks infinite recursion.
> 131         if (annotationClass != Retention.class &&
> 132             annotationClass != Inherited.class) {
> 133             JavaLangAccess jla = sun.misc.SharedSecrets.getJavaLangAccess();
> 134             Map<Class<? extends Annotation>, Annotation> metaAnnotations =
> 135                 AnnotationParser.parseAnnotations(
> 136                     jla.getRawClassAnnotations(annotationClass),
> 137                     jla.getConstantPool(annotationClass),
> 138                     annotationClass,
> 139                     Retention.class, Inherited.class
> 140                 );
> 141             Retention ret = (Retention) metaAnnotations.get(Retention.class);
> 142             retention = (ret == null ? RetentionPolicy.CLASS : ret.value());
> 143             inherited = metaAnnotations.containsKey(Inherited.class);
> 144         }
> 145         else {
> 146             retention = RetentionPolicy.RUNTIME;
> 147             inherited = false;
> 148         }
> 
> 
> This is enough to break recursion. The RUNTIME and !inherited assumptions for @Retention and @Inherited meta-annotations are therefore localized in this code only.
> 
> I have also added two tests. The one for detecting deadlock situation and the other for consistent parsing of mutually recursive annotations in presence of separate compilation.
> 
> 
> 
> Regards, Peter
>