Annotated Type bugs in OpenJDK 8
joel.franck at oracle.com
Mon Mar 9 20:07:41 UTC 2015
> On 28 feb 2015, at 03:09, Joshua Humphries <joshua at bluegosling.com> wrote:
> While working on tests for the changes I'm making, I've come across at least two other bugs.
> One is in core reflection: sun.reflect.annotation.TypeAnnotationParser#parseAnnotatedBounds is computing the wrong start index if the first type argument is a parameterized type but represents a class (vs. an interface). That is easy enough to include in the patch I am working on.
> Another seems to be in the compiler. I have a case where the wrong location info is being recorded (for a class type parameter bound), so when we rehydrate the annotations, they get attributed to the wrong part of the type parameter. (I haven't dug into this one yet at all, but I do at least have a repro case.)
> But this brings up a question: should I add additional bug reports for everything I find as I'm working on this patch? Since I am not a member, I cannot interact with the bug tracking system, other than to submit an anonymous bug through http://bugs.java.com/. (At least that I know of: if there are other ways for me to submit info for bugs, I'd be glad to hear it.)
If you post an example of the error here on compiler-dev I hope one of my colleagues will pick it up and file a bug.
> Also, are there any guidelines as to how/when to break changes into multiple patches? Would it be okay if I try to include all fixes in a single patch (which is what I'm currently doing)? Or do I need to make a separate patch per bug? Or does it depend on the size of the changes (e.g. handful of small changes may be fine to combine into single patch)?
I prefer smaller fixes and I would guess that potential reviewers for this will agree. Also since only parts of the work can be backported I think it makes sense to split it up.
More information about the compiler-dev