public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Jason Merrill <jason@redhat.com>
To: Patrick Palka <ppalka@redhat.com>, gcc-patches@gcc.gnu.org
Subject: Re: [PATCH] c++: ICE with failed __is_constructible constraint [PR100474]
Date: Tue, 29 Mar 2022 16:53:00 -0400	[thread overview]
Message-ID: <63ae6234-0e06-3b9b-f938-afb92cc5dbf6@redhat.com> (raw)
In-Reply-To: <20220329192230.255163-1-ppalka@redhat.com>

On 3/29/22 15:22, Patrick Palka wrote:
> Here we're crashing when diagnosing a failed __is_constructible constraint
> because diagnose_atomic_constraint don't know how to diagnose a trait
> that diagnose_trait_expr doesn't specifically handle.  This patch fixes
> this by falling through to the default case in this situation.
> 
> Bootstrapped and regtested on x86_64-pc-linux-gnu, does this look OK for
> trunk and perhaps 11?

Hmm, it seems reasonable, but I think it would be better to actually 
handle all the traits.  Removing the default, I get

> constraint.cc:3585:10: warning: enumeration value ‘CPTK_BASES’ not handled in switch [-Wswitch]
> constraint.cc:3585:10: warning: enumeration value ‘CPTK_DIRECT_BASES’ not handled in switch [-Wswitch]
> constraint.cc:3585:10: warning: enumeration value ‘CPTK_UNDERLYING_TYPE’ not handled in switch [-Wswitch]

These we don't need to handle.

> constraint.cc:3585:10: warning: enumeration value ‘CPTK_HAS_UNIQUE_OBJ_REPRESENTATIONS’ not handled in switch \
> constraint.cc:3585:10: warning: enumeration value ‘CPTK_IS_AGGREGATE’ not handled in switch [-Wswitch]
> constraint.cc:3585:10: warning: enumeration value ‘CPTK_IS_TRIVIALLY_ASSIGNABLE’ not handled in switch [-Wswit\
> constraint.cc:3585:10: warning: enumeration value ‘CPTK_IS_TRIVIALLY_CONSTRUCTIBLE’ not handled in switch [-Ws\
> constraint.cc:3585:10: warning: enumeration value ‘CPTK_IS_TRIVIALLY_COPYABLE’ not handled in switch [-Wswitch\
> constraint.cc:3585:10: warning: enumeration value ‘CPTK_IS_ASSIGNABLE’ not handled in switch [-Wswitch]
> constraint.cc:3585:10: warning: enumeration value ‘CPTK_IS_CONSTRUCTIBLE’ not handled in switch [-Wswitch]
> constraint.cc:3585:10: warning: enumeration value ‘CPTK_IS_NOTHROW_ASSIGNABLE’ not handled in switch [-Wswitch\
> constraint.cc:3585:10: warning: enumeration value ‘CPTK_IS_NOTHROW_CONSTRUCTIBLE’ not handled in switch [-Wswi\

These we should.

I think we should leave off the default so that when we add more traits 
we get a warning that we need to handle them here.

> 	PR c++/100474
> 
> gcc/cp/ChangeLog:
> 
> 	* constraint.cc (diagnose_trait_expr): Rename to ...
> 	(maybe_diagnose_trait_expr): ... this.  Return a boolean
> 	indicating whether we handled the trait.
> 	(diagnose_atomic_constraint) <case TRAIT_EXPR>: Fall through to
> 	the default case if maybe_diagnose_trait_expr returns false.
> 
> gcc/testsuite/ChangeLog:
> 
> 	* g++.dg/cpp2a/concepts-traits3.C: New test.
> ---
>   gcc/cp/constraint.cc                          | 16 +++++++++-------
>   gcc/testsuite/g++.dg/cpp2a/concepts-traits3.C | 10 ++++++++++
>   2 files changed, 19 insertions(+), 7 deletions(-)
>   create mode 100644 gcc/testsuite/g++.dg/cpp2a/concepts-traits3.C
> 
> diff --git a/gcc/cp/constraint.cc b/gcc/cp/constraint.cc
> index c5a991b9e71..27b1b9bb659 100644
> --- a/gcc/cp/constraint.cc
> +++ b/gcc/cp/constraint.cc
> @@ -3567,10 +3567,10 @@ get_constraint_error_location (tree t)
>     return input_location;
>   }
>   
> -/* Emit a diagnostic for a failed trait.  */
> +/* Maybe emit a friendlier diagnostic for the failed trait.  */
>   
> -static void
> -diagnose_trait_expr (tree expr, tree args)
> +static bool
> +maybe_diagnose_trait_expr (tree expr, tree args)
>   {
>     location_t loc = cp_expr_location (expr);
>   
> @@ -3655,8 +3655,9 @@ diagnose_trait_expr (tree expr, tree args)
>         inform (loc, "  %qT is not a union", t1);
>         break;
>       default:
> -      gcc_unreachable ();
> +      return false;
>       }
> +  return true;
>   }
>   
>   /* Diagnose a substitution failure in the atomic constraint T using ARGS.  */
> @@ -3685,9 +3686,6 @@ diagnose_atomic_constraint (tree t, tree args, tree result, sat_info info)
>     STRIP_ANY_LOCATION_WRAPPER (expr);
>     switch (TREE_CODE (expr))
>       {
> -    case TRAIT_EXPR:
> -      diagnose_trait_expr (expr, args);
> -      break;
>       case REQUIRES_EXPR:
>         gcc_checking_assert (info.diagnose_unsatisfaction_p ());
>         /* Clear in_decl before replaying the substitution to avoid emitting
> @@ -3696,6 +3694,10 @@ diagnose_atomic_constraint (tree t, tree args, tree result, sat_info info)
>         info.in_decl = NULL_TREE;
>         tsubst_requires_expr (expr, args, info);
>         break;
> +    case TRAIT_EXPR:
> +      if (maybe_diagnose_trait_expr (expr, args))
> +	break;
> +      /* Fall through.  */
>       default:
>         if (!same_type_p (TREE_TYPE (result), boolean_type_node))
>   	error_at (loc, "constraint %qE has type %qT, not %<bool%>",
> diff --git a/gcc/testsuite/g++.dg/cpp2a/concepts-traits3.C b/gcc/testsuite/g++.dg/cpp2a/concepts-traits3.C
> new file mode 100644
> index 00000000000..33152242988
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/cpp2a/concepts-traits3.C
> @@ -0,0 +1,10 @@
> +// PR c++/100474
> +// { dg-do compile { target c++20 } }
> +
> +template<class T, class... Args>
> +concept C = __is_constructible(T, Args...); // { dg-message "evaluated to 'false'" }
> +struct S {
> +  S() = delete;
> +};
> +
> +static_assert(C<S>); // { dg-error "assert" }


  reply	other threads:[~2022-03-29 20:53 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-03-29 19:22 Patrick Palka
2022-03-29 20:53 ` Jason Merrill [this message]
2022-03-30 12:56   ` Patrick Palka
2022-03-30 13:03     ` Jason Merrill

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=63ae6234-0e06-3b9b-f938-afb92cc5dbf6@redhat.com \
    --to=jason@redhat.com \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=ppalka@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).