On 10/1/18 10:46 AM, Martin Buchholz wrote:
This is one example of a change that should leave zero impact on the generated bytecode, and that should be testable, if only by comparing the .class files for identical size.
This is a good point. I did something similar when I did the "diamond" conversion way back in JDK 7. That conversion didn't change any bytecodes, I compared class files to ensure the results were identical after the change. That provided greater assurance of correctness for large changesets.
s'marks
