Try to get mb-math working with TruffleRuby

[mike-bourgeous]

Based on a conversation elsewhere, it would be nice to see if TruffleRuby can run my math and sound code faster than CRuby. This would help with the synthesizer demo from mb-surround and my future videos.

• [ ] https://github.com/oracle/truffleruby/issues/1589 for the GNUplot wrapper
• [ ] Make sure RSpec tests pass in TruffleRuby
• [x] Add TruffleRuby to CI matrix

[mike-bourgeous]

I've worked around the absence of PTY by rescuing LoadError in plot.rb. See commit e08d9573b09fd96863055f01db70394c83eedc56.

[mike-bourgeous]

So far I've run into two TruffleRuby crashes/uncaught exceptions:

• One while running rspec
  • Infinite or NaN (java.lang.NumberFormatException) within my PrecisionMethods#ceil
• One while running pry (on mb-util)
  • Index: 46548, Size: 46548 (java.lang.IndexOutOfBoundsException) within TruffleRuby's readline implementation

[mike-bourgeous]

[mike-bourgeous]

When rspec does run, in addition to the expected plotting failures, there are parsing failures for complex numbers, and differences in default rounding mode.

[mike-bourgeous]

Saw a repeatable segfault in rspec with seed 17464:

truffleruby-mb-math-segfault.txt

[mike-bourgeous]

Failures from https://github.com/mike-bourgeous/mb-math/runs/4185434515?check_suite_focus=true --

• 1) MB::M.parse_complex raises an error for "invalid" Failure/Error: expect { MB::M.parse_complex(k) }.to raise_error(v) expected ArgumentError but nothing was raised

• 2) MB::M.parse_complex parses "1 + 1i" correctly Failure/Error: expect(MB::M.round(MB::M.parse_complex(k), 10)).to eq(MB::M.round(v, 10))

  expected: (1+1i) got: (1+0i)

• 10) MB::M::PrecisionMethods#round with positive figs rounds NArrays Failure/Error: expect(MB::M.round(Numo::SFloat[-0.00125, 1.23456], 4)).to eq([-0.0013, 1.2346])

  expected: [-0.0013, 1.2346]
       got: Numo::SFloat#shape=[2]
  [-0.0012, 1.2346]

• truffleruby: an internal exception escaped out of the interpreter, please report it to https://github.com/oracle/truffleruby/issues.

  Infinite or NaN (java.lang.NumberFormatException)
  from java.math.BigDecimal.<init>(BigDecimal.java:923)
  from java.math.BigDecimal.<init>(BigDecimal.java:900)
  from org.truffleruby.core.numeric.BigIntegerOps.fromDouble(BigIntegerOps.java:263)
  from org.truffleruby.core.numeric.FixnumOrBignumNode.fixnumOrBignum(FixnumOrBignumNode.java:65)
  from org.truffleruby.core.numeric.FloatNodes$FloatCeilPrimitiveNode.ceil(FloatNodes.java:554)
  from org.truffleruby.core.numeric.FloatNodesFactory$FloatCeilPrimitiveNodeFactory$FloatCeilPrimitiveNodeGen.execute(FloatNodesFactory.java:2946)
  from org.truffleruby.language.control.IfElseNode.execute(IfElseNode.java:43)
  from org.truffleruby.language.control.SequenceNode.execute(SequenceNode.java:36)
  from org.truffleruby.language.RubyMethodRootNode.execute(RubyMethodRootNode.java:58)
  from org.graalvm.compiler.truffle.runtime.OptimizedCallTarget.executeRootNode(OptimizedCallTarget.java:650)
    /home/runner/work/mb-math/mb-math/lib/mb/m/precision_methods.rb:42:in `ceil'
  from /home/runner/work/mb-math/mb-math/lib/mb/m/precision_methods.rb:42:in `sigfigs'
  from /home/runner/work/mb-math/mb-math/spec/lib/mb/m/special_functions_spec.rb:86:in `block (4 levels) in <top (required)>'

[gogainda]

@mike-bourgeous I would recommend to open up corresponding issues in truffleruby repository

[mike-bourgeous]

Probably not going to work with TruffleRuby for now. May revisit in the future.