[BUG] Compound assignment chosen over postfix operator

[JohelEGP]

Title: Compound assignment chosen over postfix operator

This seems at odds with
https://github.com/hsutter/cppfront/wiki/Design-note:-Unambiguous-parsing,
https://github.com/hsutter/cppfront/wiki/Design-note%3A-Postfix-operators, and
https://github.com/hsutter/cppfront/wiki/Design-note%3A-Postfix-unary-operators-vs-binary-operators.

Minimal reproducer (https://godbolt.org/z/PYEnvh6Px):

f4: (inout v0: i8, copy v1: *i8) = {
  v0&*= 1;
}
main: () = {
  v0: i8 = 1;
  f4(v0, v0&);
}

Commands:

cppfront x.cpp2
clang++17 -std=c++2b -stdlib=libc++ -I $CPPFRONT_INCLUDE_DIR x.cpp

Expected result:
*&v0 = 1 to be generated instead of
&v0 *= 1,
because v0&* can be parsed as a postfix-expression
because v0& can be parsed as a postfix-expression
because v0 can be parsed as a postfix-expression
and those match first,
before assignment-expression assignment-operator logical-or-expression
has a chance to use *= as an assignment-operator,
and because the leftover = 1 can match that assignment-operator logical-or-expression.

Actual result and error:

Generated Cpp1.

//=== Cpp2 type declarations ====================================================


#include "cpp2util.h"



//=== Cpp2 type definitions and function declarations ===========================

#line 1 "/app/main.cpp2"
auto f4(cpp2::i8& v0, cpp2::i8* v1) -> void;
  

auto main() -> int;
  

//=== Cpp2 function definitions =================================================

#line 1 "/app/main.cpp2"
auto f4(cpp2::i8& v0, cpp2::i8* v1) -> void{
  &v0 *= 1;
}
auto main() -> int{
  cpp2::i8 v0 {1}; 
  f4(v0, &v0);
}

main.cpp2:2:7: error: invalid operands to binary expression ('cpp2::i8 *' (aka 'signed char *') and 'int')
  &v0 *= 1;
  ~~~ ^  ~
1 error generated.

[hsutter]

Thanks. It's max munch... *= is one token, so the above parses can't happen... the only * token in this code is on the parameter.

Distilling even slightly further:

p: *i8 = f();
p*= 1;

That should clearly be compound assignment [edit: and semantically illegal since we can't multiply-assign pointers] even without the whitespace, I think... what do you think?

[JohelEGP]

It's max munch...

Thank you. I'll have to study this.

[hsutter]

I should distill even further:

i := 42;
i*= 1;

[JohelEGP]

It's max munch...

Thank you. I'll have to study this.

Maybe later. I remembered what should be the source of this parsing terminology. Translation phase 3 tokenizes before phase 7 reads the sequence according to the grammar.

[JohelEGP]

The max much rule must come from

(3.3)
Otherwise, the next preprocessing token is the longest sequence of characters that could constitute a preprocessing token, even if that would cause further lexical analysis to fail, except that a header-name ([lex.header]) is only formed
(3.3.1)
after the include or import preprocessing token in an #include ([cpp.include]) or import ([cpp.import]) directive, or
(3.3.2)
within a has-include-expression.

[Example 1: #define R "x"
const char* s = R"y"; // ill-formed raw string, not "x" "y"
— end example]

[JohelEGP]

Is it safe to say that https://eel.is/c++draft/lex#digraph doesn't apply to Cpp2, and v0<:int = 0> should lower to v0<int{0}> and not interpret the <: as a [ (https://godbolt.org/z/oeKs3xad9)? If so, I think we can close this.

[hsutter]

Right, Cpp2 eliminates all the things on this cppreference page as special features... no digraphs or trigraphs as they aren't among the specified tokens, and no alternative operator macro names which were reclaimed for normal use in #328.