Remove extra ElementTypes in all synthetic annotations

[theron-wang]

This change allows na-705 to be reproducible. I discovered this one change was able to reproduce the error when running Specimin with --jarPath path/to/checker-qual.jar. Since @Retention(RetentionPolicy.RUNTIME) is the most generous, it should work for all cases (i.e. CLASS and SOURCE). (see comment below)

[kelloggm]

Nice find!

[theron-wang]

Never mind, it looks like @Retention(RetentionPolicy.RUNTIME) was a red herring (the checker jar also only used @Target(TYPE, TYPE_USE). The issue for this was an over-correction in #327, where I included all ElementTypes in @Target. In fact, the only change I needed to make was to add ElementType.Type. I believe that the extra definition of ElementType.Field, in this case, caused the annotation to be incorrectly processed by NullAway.

[kelloggm]

@theron-wang I think this is still off - now not only is na-705 failing, but na-97 is, too.

[theron-wang]

@kelloggm Looks like ElementType.FIELD is needed to reproduce na-97, but its presence in na-705 prevents the NullAway from reproducing the bug. I tried removing @Target altogether for both examples, but na-705 still is not being reproduced. I believe that the issue is that na-705 uses CF annotations, which are type use, versus the Javax annotations in na-97, which do not have a target by default. I don't think there is way to differentiate between the two cases (we can't check by context, since both na-97 and na-705 @Nullable annotations apply to fields), so the best way to still reproduce na-705 is to run it with the checker-qual.jar. What are your thoughts? We could just close this PR and add a checker-qual dependency in specimin-evaluation for na-705, but that is an unsatisfying solution.

[theron-wang]

Closed as per njit-jerse/specimin-evaluation#12.