Re: [PATCH] c++: CWG 2359, wrong copy-init with designated init [PR91319]

[Marek Polacek <polacek@redhat.com>]

On Mon, Aug 28, 2023 at 06:27:26PM -0400, Jason Merrill wrote:
> On 8/25/23 12:44, Marek Polacek wrote:
> > Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk?
> > 
> > -- >8 --
> > 
> > This CWG clarifies that designated initializer support direct-initialization.
> > Just be careful what Note 2 in [dcl.init.aggr]/4.2 says: "If the
> > initialization is by designated-initializer-clause, its form determines
> > whether copy-initialization or direct-initialization is performed."  Hence
> > this patch sets CONSTRUCTOR_IS_DIRECT_INIT only when we are dealing with
> > ".x{}", but not ".x = {}".
> > 
> > 	PR c++/91319
> > 
> > gcc/cp/ChangeLog:
> > 
> > 	* parser.cc (cp_parser_initializer_list): Set CONSTRUCTOR_IS_DIRECT_INIT
> > 	when the designated initializer is of the .x{} form.
> > 
> > gcc/testsuite/ChangeLog:
> > 
> > 	* g++.dg/cpp2a/desig30.C: New test.
> > ---
> >   gcc/cp/parser.cc                     |  6 ++++++
> >   gcc/testsuite/g++.dg/cpp2a/desig30.C | 22 ++++++++++++++++++++++
> >   2 files changed, 28 insertions(+)
> >   create mode 100644 gcc/testsuite/g++.dg/cpp2a/desig30.C
> > 
> > diff --git a/gcc/cp/parser.cc b/gcc/cp/parser.cc
> > index eeb22e44fb4..b3d5c65b469 100644
> > --- a/gcc/cp/parser.cc
> > +++ b/gcc/cp/parser.cc
> > @@ -25718,6 +25718,7 @@ cp_parser_initializer_list (cp_parser* parser, bool* non_constant_p,
> >         tree designator;
> >         tree initializer;
> >         bool clause_non_constant_p;
> > +      bool direct_p = false;
> >         location_t loc = cp_lexer_peek_token (parser->lexer)->location;
> >         /* Handle the C++20 syntax, '. id ='.  */
> > @@ -25740,6 +25741,8 @@ cp_parser_initializer_list (cp_parser* parser, bool* non_constant_p,
> >   	  if (cp_lexer_next_token_is (parser->lexer, CPP_EQ))
> >   	    /* Consume the `='.  */
> >   	    cp_lexer_consume_token (parser->lexer);
> > +	  else
> > +	    direct_p = true;
> >   	}
> >         /* Also, if the next token is an identifier and the following one is a
> >   	 colon, we are looking at the GNU designated-initializer
> > @@ -25817,6 +25820,9 @@ cp_parser_initializer_list (cp_parser* parser, bool* non_constant_p,
> >         if (clause_non_constant_p && non_constant_p)
> >   	*non_constant_p = true;
> > +      if (TREE_CODE (initializer) == CONSTRUCTOR)
> > +	CONSTRUCTOR_IS_DIRECT_INIT (initializer) |= direct_p;
> 
> Why |= rather than = ?

CONSTRUCTOR_IS_DIRECT_INIT could already have been set earlier so using
= might wrongly clear it.  I saw this in direct-enum-init1.C.

Marek