macroexpand: stop pre-running the hygiene pass

[vtjnash]

This prepares us to delete the separate hygiene pass, and make it part of lowering resolve-scopes. This affects a few macros which expect to run macroexpand then mangle the result more, and the result of those macros is not already wrapped in esc(). The damage seems relatively minor however, and generally a good improvement.

[vtjnash]

@nanosoldier runtests()

[nanosoldier]

Your package evaluation job has completed - possible new issues were detected.
A full report can be found here.

[vtjnash]

Some findings so far (to be updated):

• Some package that appear to violate hygiene rules, now detectably broken by this:

  • Gen.jl
  • Enums.jl (https://s3.amazonaws.com/julialang-reports/nanosoldier/pkgeval/by_hash/5bdb3d1_vs_c6fc12c/HomotopyOpt.primary.log)

• It appears that many more places in Docs need to call the new unescape, which should resolve the breakage of many packages:

  • Graphics.jl
  • https://s3.amazonaws.com/julialang-reports/nanosoldier/pkgeval/by_hash/5bdb3d1_vs_c6fc12c/EnumX.primary.log
  • https://s3.amazonaws.com/julialang-reports/nanosoldier/pkgeval/by_hash/5bdb3d1_vs_c6fc12c/FileIO.primary.log
  • Optimization.jl
  • https://s3.amazonaws.com/julialang-reports/nanosoldier/pkgeval/by_hash/5bdb3d1_vs_c6fc12c/PreallocationTools.primary.log
  • SimpleTraits.jl

• Some places have tests for the results of macroexpand as a means of verifying their macro output. These tests should be somewhat simpler to write (as they will exactly test the macro output instead of the mangled output), but are non-trivial to fix. Inverting the output to return Moon's PR-style hygiene results would potentially help avoid breakage here (which is roughly what our hygiene application pass is supposed to do now), but does still result in an extra IR pass.

  • https://s3.amazonaws.com/julialang-reports/nanosoldier/pkgeval/by_hash/5bdb3d1_vs_c6fc12c/ArgCheck.primary.log
  • https://s3.amazonaws.com/julialang-reports/nanosoldier/pkgeval/by_hash/5bdb3d1_vs_c6fc12c/PyCall.primary.log