More boolean simplifications in generated code.

[cristianoc]

This handles all the boolean simplifications in #7142.
What's left from that issue is conditionals b ? x : y, that can be done separately.

[zth]

Really nice! 👏

[yallop]

Does this need a check that e1 doesn't have side effects?

[cristianoc]

Thanks! It does.

[cristianoc]

0d49895

[zth]

@yallop nice find! See any more issues of that kind? 👀

[yallop]

Yes, there seem to be a few. To pick some at random:

• in eq_null_undefined_boolean

  match (a.expression_desc, b.expression_desc) with
  | ( (Null | Undefined _),
     (Bool _ | Number _ | Typeof _ | Fun _ | Array _ | Caml_block _) ) ->

  the Typeof or Array or Caml_block could have side effects, so can't always be safely dropped
• in int_comp

    | Ceq, Optional_block _, Undefined _ | Ceq, Undefined _, Optional_block _ ->
   false_

  if Optional_block can have side effects then it shouldn't be dropped
• in simplify_or

      | _, Bool true -> Some true_

  the left operand shouldn't be dropped if it has side effects.

This is just a selection; there are some similar cases in those functions and others. I don't have all the development tools set up, so can't easily send a PR, unfortunately.

[cristianoc]

Yes, there seem to be a few. To pick some at random:

• in eq_null_undefined_boolean

  match (a.expression_desc, b.expression_desc) with
  | ( (Null | Undefined _),
     (Bool _ | Number _ | Typeof _ | Fun _ | Array _ | Caml_block _) ) ->

  the Typeof or Array or Caml_block could have side effects, so can't always be safely dropped

• in int_comp

    | Ceq, Optional_block _, Undefined _ | Ceq, Undefined _, Optional_block _ ->
   false_

  if Optional_block can have side effects then it shouldn't be dropped

• in simplify_or

      | _, Bool true -> Some true_

  the left operand shouldn't be dropped if it has side effects.

This is just a selection; there are some similar cases in those functions and others. I don't have all the development tools set up, so can't easily send a PR, unfortunately.

Thanks a lot @yallop !
I've fixed a bunch more in #7151