Upgrade ASM to 6.2

[Godin]

No description provided.

[Godin]

@marchof could you please have a look? I didn't update changelog to first discuss below points.

---

All changes in our tests, except one about ClassWriter.COMPUTE_FRAMES, are because starting from version 6.1 ASM is more strict about allowed arguments.

---

Should be noted that some regressions were already found in ASM 6.2, however to me seems that JaCoCo can't be affected by them, but might be worth double-checking - AFAIK as of 2 July 2018 from the list of issues (https://gitlab.ow2.org/asm/asm/issues) and changes after ASM 6.2 (https://gitlab.ow2.org/asm/asm/compare/ASM_6_2...master):

• https://gitlab.ow2.org/asm/asm/issues/317835 - we don't use ClassWriter.COMPUTE_MAXS and therefore MethodWriter.COMPUTE_MAX_STACK_AND_LOCAL_FROM_FRAMES
• https://gitlab.ow2.org/asm/asm/issues/317837 - we use GeneratorAdapter only in tests

And seems that while https://gitlab.ow2.org/asm/asm/issues/317833 was fixed after 6.2, problem was already there with ASM 6.0, also we don't use ClassWriter.COMPUTE_MAXS and don't change number of arguments of methods.

---

Also should be noted that ASM 6.2 contains experimental support for the upcoming JDK 11 features (nest mates JEP-181 and ConstantDynamic JEP-309) - see https://mail.ow2.org/wws/arc/asm/2018-05/msg00004.html Furthermore ClassReader doesn't reject JDK 11 classes even if flag Opcodes.ASM7_EXPERIMENTAL is not used - see https://mail.ow2.org/wws/arc/asm/2018-05/msg00000.html

While seems that this allows to use our ModifiedSystemClassRuntime under JDK 11 and perform on-the-fly instrumentation of classes compiled into bytecode version 55 without other changes (ClassReader reads new ConstantDynamic constant pool entries and new nest mates attributes, ClassWriter copies them), just an update of ASM alone is clearly not enough for proper full support of JDK 11:

• ContentTypeDetector should be updated for generation of report and offline instrumentation for classes compiled into bytecode version 55
• TryWithResourcesJavacFilter also requires update - need to finish work started in Update TryWithResourcesJavacFilter for javac 11 #669
• need to have a look on JEP 12 ( http://openjdk.java.net/jeps/12 ) as it changes meaning of a minor segment of bytecode version

Also even if JDK 11 is now in Rampdown Phase One, AFAIK there is no strong guarantees that bytecode won't change, and this is also why support for JDK 11 features is experimental in ASM 6.2. IMO the best what we can do - is also "experimental support", i.e. without any guarantees. WDYT? In any case, I would prefer to separate ASM update from JDK 11 support.

[marchof]

@Godin +1 for adding experimental JDK 11 support in a separate PR

[marchof]

Thanks!

[Godin]

@marchof and just to be sure - what about known regressions in ASM 6.2? do you agree with my analysis that they don't affect us? any concerns about possible others, given that 6.1 was a massive refactoring?

[marchof]

@Godin I agree with you that the regressions do not affect usage in JaCoCo. And hey, our tests are green ;)

[Godin]

@marchof well, as history shows - our tests are not totally ehaustive and bulletproof 😉

I updated changelog and will do the merge. Thank you for review!

[Godin]

@marchof in fact while

ClassReader reads new ConstantDynamic constant pool entries and new nest mates attributes, ClassWriter copies them

is correct, ClassVisitor without Opcodes.ASM7_EXPERIMENTAL throws UnsupportedOperationException when ClassReader encounters/visits nest mates 😞

[Godin]

For the record: ticket to upgrade ASM in Eclipse Orbit - https://bugs.eclipse.org/bugs/show_bug.cgi?id=536922