GROOVY-10278, GROOVY-10281: refactor CompilerConfiguration somewhat

[eric-milles]

This is not complete, but does give some shape to recent discussions.

• ASM_API_VERSION goes to AsmClassGenerator
• DEFAULT_TARGET_BYTECODE and defaultTargetBytecode() were unnecessary additions
• setTargetBytecode(String) can use nearest value as described in GROOVY-10278
• isPostJDKn can be replaced with use of isAtLeast(String,String) -- the recent additions could be dropped, but I just moved them for now
• LoggableClassVisitor --> TraceClassVisitor (just need to work out the jarjar stuff for unit tests

Thoughts?

[sonatype-lift]

UnusedVariable: The parameter 'config' is never read. (details)
(at-me in a reply with help or ignore)

[daniellansun]

LoggableClassVisitor --> TraceClassVisitor

After reading tons of "minor edits" and this PR, I am still not sure why "inline" or "anonymous" code is preferred over code with meaningful names, e.g. method names, class names, etc.

-1 from me.

[eric-milles]

Such as what? If you gave some examples, I might be able to provide reasoning. The purpose of the draft was to discuss these items rather than just be dismissive.

What is the concern with leveraging TraceClassVisitor which is provided instead of maintaining LoggableClassVisitor which tries to do the same job and I've found many problems with the flushing and the multiple instance creation?

If you are referencing the replacement of isPostJDK5(), then my raised concerns earlier are that "is post" is not the correct semantic. It is really "is at least" and the use if isAtLeast(targetBytecode,JDK5) is conveys that semantic and does not require a new public API method for every Java release, which are coming every 6 months now, not every few years. Again, this is demonstrated for discussion purposes. It is not a must have.

I disagree with lifting one-time use values up to the top of a class file. I disagree with adding public API for something that might be useful but has no track record of need. These are specific examples that can foster further discussion.

[eric-milles]

Now if the new methods added were isAtLeastJDK9 and so on and isPostJDK5, etc. were removed or scheduled for removal, then my only concern would be the need to add a new method every six months. That kind of work should be automated IMO.

[paulk-asert]

According to a strict interpretation of ASF rules, a -1 on a code submission is a veto:
https://www.apache.org/foundation/voting.html#votes-on-code-modification
I'll close this issue and see if there are any non-controversial parts which can be cherry-picked.