search

whitequark / parser

A Ruby parser.
Other
1.57k stars 197 forks source link

+ ruby33.y: extract p_assoc and p_in rules #976

Closed iliabylich closed 8 months ago

iliabylich commented 8 months ago

This commit tracks upstream commit ruby/ruby@3b05238.

Closes https://github.com/whitequark/parser/issues/951.

Racc doesn't have access to a result of a p_assoc rule in the last mid-rule of the diff (probably because of a lookahead?) and so it can't read $ctxt.in_kwarg. If that's true for Lrama then it's a very nasty bug that I don't see exactly how to reproduce :(. It could be something like this:

def foo a: case 1; in 1; then 1; end, b:
  p 42
end

that gets rejected (this particular snippet is accepted by MRI which is correct). It could even be a UB if generated code reads uninitialised memory and transmutes it to ctxt (then it becomes a heisenbug) but I hope Ruby CI runs ASAN for all of its tests.

iliabylich commented 8 months ago

cc @nobu, I hope it's a false alarm and Lrama does everything right but could you PTAL? Do all mid-rules have access to results of previous mid-rules in the scope of the same derivation?