Numerous updates to conversion service

[philwebb]

Changes are primarily to support conversion of enums with interfaces and to allow performance optimization when converting byte arrays.

See individual commits for details

Issues: SPR-9566, SPR-9692, SPR-9927, SPR-9928

[philwebb]

@cbeams This commit has now been reworked following the review comments in:
philwebb#1

Specifically:

• The use of ExpectedException causes NoSuchMethodErrors for me under both Eclipse and Gradle. Presumably you're getting different results? Please run tests like TypeDescriptorTests to confirm, and let's figure out how to make everything work.

I have removed ExpectedException for now, we can return to this later

• TypeDescriptor has been made final - why? What do you think the likelihood is of this being a breaking change for people, i.e. how likely is it that folks might have extended this type?

TypeDescriptor is no longer final

• Numerous updates to conversion service #132 contains commits for both SPR-9566 and SPR-9692 - why, i.e. what is the relationship between these two, i.e. why combine both in a single PR?

This is mainly due to the fact that both issues touch the same areas of code so create atomic pull requests is hard. This pull request still deals with a number of issues but now the commits are much more logically organized. I have tried to order them in terms of how controversial they are so that you can work from the bottom up

• both SPR-9566 and SPR-9692 have their fixVersion set for both 3.1.3 and 3.2 RC1. Given the scope of these changes, I am hesitant to backport them to 3.1.3 (still need to understand the complete impact even to 3.2). Do you have a strong case for backporting one or both?

Neither should be backported IMHO. I have removed 3.1.3 from the fixVersion field

• It appears that these changes modify the conversion algorithm to match superclasses with a higher precedence than interfaces. As per your comments in SPR-9692, this would seem to introduce a backward-compatibility problem. Do you have a rationale as to why this change is merited, i.e. that the previous behavior is simply wrong, that it is very unlikely to actually affect anyone, etc.?

I spoke to @jhoeller about why the search changed from 3.0 to 3.1, he recalled that there was a good reason but not exactly what it was. The new algorithm does not fundamentally change the way that the type hierarchy is searched, it just considers java. and javax. packages after other packages (see the updated commit message).

Thanks for the review, it was very helpful!

[philwebb]

@cbeams I have updated the request again becuase I thought of a better way to fix the enum issue that does not involve changing the class search order. See commit 9272f48. The new implementation simply makes EnumToStringConverter conditional on the fact that no interfaces can be converted.

[cbeams]

Thanks, Phil, this is now merged into SpringSource/master.