Fix #4066 ./bin/dotr -language:Scala2 fails

[newca12]

• allow to pass compilation option to ./bin/dotr
• restore standard notation for java options (using -J-Xmx512m instead of -JXmx512m introduce with Cannot pass jvm parameters to dotr #3509)
• allow to pass java options to ./bin/dotr -repl
• allow to pass java options to ./bin/dotc

I was a bit surprise to find dotr and dotc in such bad shape.
The code was and still is a bit fishy but as far as I can see now work in every situation.

[dottybot]

Hello, and thank you for opening this PR! 🎉

All contributors have signed the CLA, thank you! ❤️

Have an awesome day! ☀️

[newca12]

I did sign the CLA.

[Blaisorblade]

This comment matches the old behavior ("add them to JAVA_OPTS as well [as to the Scala options]"), could you please update accordingly? I double-checked the code, but I agree with your change.

[liufengyun]

This code dates back to scalac script, maybe we should keep it:

  case "$1" in
    -D*)
      # pass to scala as well: otherwise we lose it sometimes when we
      # need it, e.g. communicating with a server compiler.
      java_args+=("$1")
      scala_args+=("$1")
      # respect user-supplied -Dscala.usejavacp
      case "$1" in -Dscala.usejavacp*) OVERRIDE_USEJAVACP="";; esac
      shift
      ;;
    -J*)
      # as with -D, pass to scala even though it will almost
      # never be used.
      java_args+=("${1:2}")
      scala_args+=("$1")
      shift
      ;;

[Blaisorblade]

CLA signed.

[liufengyun]

Thanks a lot for the contribution @newca12 👍

Once it's ready, we can test it on windows in the following project https://github.com/lampepfl/packtest

[liufengyun]

Maybe rename not_only_options to num_of_options?

[liufengyun]

Maybe rename to something like addJvmOptionsRun/addJvmOptionsRepl or addJvmOptions/addDotcOptions, which is more clear?

[liufengyun]

This code dates back to scalac script, maybe we should keep it:

  case "$1" in
    -D*)
      # pass to scala as well: otherwise we lose it sometimes when we
      # need it, e.g. communicating with a server compiler.
      java_args+=("$1")
      scala_args+=("$1")
      # respect user-supplied -Dscala.usejavacp
      case "$1" in -Dscala.usejavacp*) OVERRIDE_USEJAVACP="";; esac
      shift
      ;;
    -J*)
      # as with -D, pass to scala even though it will almost
      # never be used.
      java_args+=("${1:2}")
      scala_args+=("$1")
      shift
      ;;

[newca12]

@Blaisorblade, the real signification of "as well" come from a 8 years old paulp commit scala/scala@80b0d24#diff-eb4553171a4a79a0be4ea69f883e75b2R72
@liufengyun I cannot find any evidence that scala_args is used in Scala. Sound like dead code for me.
This PR does not fix problems with Windows.

[Blaisorblade]

@liufengyun I cannot find any evidence that scala_args is used in Scala. Sound like dead code for me.

Right, that comment hints at fsc, which we don't have: the options for the fsc client had to be parsed and passed by hand to the existing fsc server.

[liufengyun]

@liufengyun I cannot find any evidence that scala_args is used in Scala. Sound like dead code for me.

Right, that comment hints at fsc, which we don't have: the options for the fsc client had to be parsed and passed by hand to the existing fsc server.

OK, I agree with you, thanks for the clarification.

[liufengyun]

Test succeeded on Linux/Windows/OSX lampepfl/packtest#14

[Blaisorblade]

Assigning to @liufengyun for merge if he has no further comments.

[liufengyun]

Thanks a lot for the contribution @newca12 🎉