public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Jason Merrill <jason@redhat.com>
To: Marek Polacek <polacek@redhat.com>,
	GCC Patches <gcc-patches@gcc.gnu.org>
Subject: Re: [PATCH] c++: ICE with constexpr dtor on ARM [PR105529]
Date: Tue, 10 May 2022 00:13:50 -0400	[thread overview]
Message-ID: <51ee566a-6b41-09fe-cfef-e9cb44e53935@redhat.com> (raw)
In-Reply-To: <20220509224107.483598-1-polacek@redhat.com>

On 5/9/22 18:41, Marek Polacek wrote:
> When compiling this test on ARM with -O, we ICE in cxx_eval_store_expression
> while evaluating a CALL_EXPR allocator::~allocator (&D.4529).  Its body
> has this store:
> 
>    <retval> = this
> 
> The RHS is evaluated into &D.4529 of type allocator *.  The object,
> <retval>, is of type void *.  Their types don't match

That seems like the bug, which I have a patch for.

> so we go into the new
> 
>    if (empty_base || !(same_type_ignoring_top_level_qualifiers_p
>                        (TREE_TYPE (init), type)))
>      {
>        empty_base = true;
>        gcc_assert (is_empty_class (TREE_TYPE (init)));
> 
> branch where we crash because is_empty_class will always say false for
> a pointer type.  It seems like the new branch expects code like
> 
>    D.2181 = {}
> 
> where the type of the LHS is the derived type and the type of the RHS is
> the base type.  But it only expects to see actual class types, not
> pointers to them.  With this patch we will do what we did before the
> empty bases cleanup patch: set *valp and return the initializer.

It seems reasonable to expect that the initializer has the same type as 
the target; otherwise other things are likely to get confused.

I'll incorporate your testcases in my patch, thanks.

> Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk?
> 
> 	PR c++/105529
> 
> gcc/cp/ChangeLog:
> 
> 	* constexpr.cc (cxx_eval_store_expression): Check CLASS_TYPE_P before
> 	checking initialization of an empty base.
> 
> gcc/testsuite/ChangeLog:
> 
> 	* g++.dg/cpp2a/constexpr-dtor13.C: New test.
> 	* g++.dg/cpp2a/constexpr-dtor14.C: New test.
> ---
>   gcc/cp/constexpr.cc                           |  5 +++--
>   gcc/testsuite/g++.dg/cpp2a/constexpr-dtor13.C | 11 +++++++++++
>   gcc/testsuite/g++.dg/cpp2a/constexpr-dtor14.C | 13 +++++++++++++
>   3 files changed, 27 insertions(+), 2 deletions(-)
>   create mode 100644 gcc/testsuite/g++.dg/cpp2a/constexpr-dtor13.C
>   create mode 100644 gcc/testsuite/g++.dg/cpp2a/constexpr-dtor14.C
> 
> diff --git a/gcc/cp/constexpr.cc b/gcc/cp/constexpr.cc
> index e560d842e8c..e4498c15f1e 100644
> --- a/gcc/cp/constexpr.cc
> +++ b/gcc/cp/constexpr.cc
> @@ -5916,8 +5916,9 @@ cxx_eval_store_expression (const constexpr_ctx *ctx, tree t,
>   
>     gcc_checking_assert (!*valp || (same_type_ignoring_top_level_qualifiers_p
>   				  (TREE_TYPE (*valp), type)));
> -  if (empty_base || !(same_type_ignoring_top_level_qualifiers_p
> -		      (TREE_TYPE (init), type)))
> +  if (empty_base || (CLASS_TYPE_P (TREE_TYPE (init))
> +		     && !(same_type_ignoring_top_level_qualifiers_p
> +			  (TREE_TYPE (init), type))))
>       {
>         /* For initialization of an empty base, the original target will be
>          *(base*)this, evaluation of which resolves to the object
> diff --git a/gcc/testsuite/g++.dg/cpp2a/constexpr-dtor13.C b/gcc/testsuite/g++.dg/cpp2a/constexpr-dtor13.C
> new file mode 100644
> index 00000000000..7b289614fc1
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/cpp2a/constexpr-dtor13.C
> @@ -0,0 +1,11 @@
> +// PR c++/105529
> +// { dg-do compile { target c++20 } }
> +// { dg-options "-O" }
> +
> +struct allocator {
> +  constexpr ~allocator() {}
> +};
> +struct S {
> +  S(int, int, allocator = allocator());
> +};
> +void to_string() { S(0, '\0'); }
> diff --git a/gcc/testsuite/g++.dg/cpp2a/constexpr-dtor14.C b/gcc/testsuite/g++.dg/cpp2a/constexpr-dtor14.C
> new file mode 100644
> index 00000000000..9c55121eb8a
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/cpp2a/constexpr-dtor14.C
> @@ -0,0 +1,13 @@
> +// PR c++/105529
> +// { dg-do compile { target c++20 } }
> +// { dg-options "-O" }
> +// Like constexpr-dtor13.C, except that allocator is not an empty class.
> +
> +struct allocator {
> +  constexpr ~allocator() {}
> +  int a;
> +};
> +struct S {
> +  S(int, int, allocator = allocator());
> +};
> +void to_string() { S(0, '\0'); }
> 
> base-commit: 8a39c7c128dbf4405821778724c5a1017ab12e06


      reply	other threads:[~2022-05-10  4:13 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-05-09 22:41 Marek Polacek
2022-05-10  4:13 ` Jason Merrill [this message]

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=51ee566a-6b41-09fe-cfef-e9cb44e53935@redhat.com \
    --to=jason@redhat.com \
    --cc=gcc-patches@gcc.gnu.org \
    --cc=polacek@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).