Add :javap and :asmp to the REPL

[griggt]

I believe this draft to be fully functional, please give it a try if interested.

Unlike the Scala 2 implementation, this one has accomodations for working with JDK 16+. However, in light of the comments in scala/bug#12378, perhaps we should be looking to support an :asmp instead?

Closes lampepfl/dotty-feature-requests#80

[test_windows_full]
[test_java8]
[test_java11]
[test_java15]
[test_java17]
[test_java18]
[test_java19]

[som-snytt]

Can't have too many toys tools, but it's also true LOC are not free. OTOH existing tool works better than the not yet existing.

[griggt]

Now with experimental :asmp support.

[dwijnand]

@griggt is this still in draft or is it ready for review?

[SethTisue]

@griggt where does this stand? I'm interested in seeing this land and might be interested in helping. (or reviewing it, if you finish it yourself.)

[griggt]

It seems that our current version of scala-asm does not support Java 18 class files, so :asmp fails to disassemble them:

https://github.com/lampepfl/dotty/runs/5608623717?check_suite_focus=true#step:10:122

Error:  Test dotty.tools.repl.AsmpTests.java.lang.String signatures failed: java.lang.AssertionError: assertion failed: 
disassembly did not contain `public static varargs format(Ljava/lang/String;[Ljava/lang/Object;)Ljava/lang/String;`
Error:  Disassembly was:
Error:  Unsupported class file major version 62

By contrast the :javap implementation appears to be OK on Java 18.

--

Edit: we're on ASM 9.1, scala/scala is on 9.2
Upgrade PR: #14711

[som-snytt]

I've needed this so badly recently, I meant to come back and whatever the phrase is, put my shoulder to the whatever you put your shoulder to. I need to see what stuff compiles to.

Edit: I already forgot I made this comment two days ago. Actually, what I needed today is -Ygen-asmp, which I learned about recently on Scala 2.

[SethTisue]

We discussed this some at the Scala 3 issue spree today. One of the remaining issues is the interaction with @targetName, but I don't consider that a blocker (in part because it isn't an easy fix).

[griggt]

To expand on Seth's comment, I believe this PR correctly deals with classes with the @targetName annotation because I borrowed and adapted this binaryClassName code from ExtractDependencies. But I suspect there may be incremental compilation issues related to @targetName, and ultimately I think this binaryClassName code should be shared rather than duplicated.

I'm going to try to find some test cases and open a separate ticket for any incremental compilation issues.

[ckipp01]

It's been quite some time here, so I wanted to circle back around. Is there any plans to return to this or bring it over the finish line?

[SethTisue]

Is there a volunteer who would like to take this over and try to get it across the finish line?

[jxnu-liguobin]

I have been following it for a long time when I was still using Scala2, but I don't know why it didn't move forward.

[som-snytt]

PSA there is no finish line. It is an oval track of endless laps.

Edit: I think there is also a PR to move repl to a subproject.

The comments made me revisit this and the scala 2 PRs. I wonder why it invites so many LOC. The scala 2 effort got pushback from retronym on complexity, and he was partially appeased. He comments sagely that it would be nice to support multi-release jars, which would hide some multiplexing/multicomplexity at the expense of a (one-time) hit to build complexity.

Alternatively, support only javap tool interface (latest) and asm is whatever the asm jar version is.