Comment author: craff

I wrote major, because this can break bootstraping of -pp or -ppx extensions
that use ascii pretty printing as a simple way to boostrap (by producing
.ml/.mli file that do not use the extension)

Actually this broke decap bootstrap at some point and forced me not to use the "let open" syntax inside conditions

Moreover, this should be an easy fix ...

Comment author: @gasche

Indeed, this was very simple to fix:
c676334
Thanks for the report!

Despite a few rewrites, the pretty-printing machinery is still baroque and I wouldn't trust it. It would be nice to overhaul it to exactly match the precedence levels and nonterminals used in the parser, to get verifiably correct minimal parentheses insertion. I'd rather get the parser converted to Menhir first to work from a nicer grammar.

Comment author: craff

I use extensively the pretty printing to bootstrap decap and this is the first time I have a problem ...

What would be a real gain is to change the grammar itself, the priority of
if then else should not change by adding a let / fun / etc ...

This is a nightmare if you do not use the yacc way of giving precedence,
which is the case when pretty printing.

The nice things, is that changing the grammar at this place actually breaks little code (I tested), and can be easily detected by parsing both with the old and the new parser.

I would suggest a few deep revision of the syntax (in places that hurt students):

• if then else priority
• optional endif, endfun, endmatch endfunction
• "in" replaced by ";" (YES)
• let replaced by val in structure (syntax error sooner in the code)

If this change are not mandatory, a smooth transition is possible

Comment author: craff

And thanks for the fix !!!