diff mbox series

[v2] c++: wrong error with constexpr COMPOUND_EXPR [PR105321]

Message ID YmFaMPqs0GrM51LF@redhat.com
State New
Headers show
Series [v2] c++: wrong error with constexpr COMPOUND_EXPR [PR105321] | expand

Commit Message

Marek Polacek April 21, 2022, 1:20 p.m. UTC 
On Thu, Apr 21, 2022 at 08:56:23AM -0400, Jason Merrill wrote:
> On 4/20/22 18:40, Marek Polacek wrote:
> > Here we issue a bogus error for the first assert in the test.  Therein
> > we have
> > 
> > <retval> = (void) (VIEW_CONVERT_EXPR<bool>(yes) || handle_error ());, VIEW_CONVERT_EXPR<int>(value);
> > 
> > which has a COMPOUND_EXPR, so we get to cxx_eval_constant_expression
> > <case COMPOUND_EXPR>.  The problem here is that we call
> > 
> > 7044             /* Check that the LHS is constant and then discard it.  */
> > 7045             cxx_eval_constant_expression (ctx, op0,
> > 7046                                           true, non_constant_p, overflow_p,
> > 7047                                           jump_target);
> > 
> > where lval is always true, so the PARM_DECL 'yes' is not evaluated into
> > its value.  r218832 changed the argument for 'lval' from false to true:
> > 
> > 	(cxx_eval_constant_expression) [COMPOUND_EXPR]: Pass true for lval.
> > 
> > but I think we want to pass 'lval' instead.  Jakub tells me that's what
> > we do for "(void) expr" as well.  [expr.comma] says that the left expression
> > is a discarded-value expression, but [expr.context] doesn't suggest that
> > we should always be passing false for lval as pre-r218832.
> 
> In a discarded-value expression, we don't do the lvalue-rvalue conversion;
> whether we want an lvalue for the RHS of the comma is irrelevant.

Ah, that's what I misread -- [expr.context]/2.8 cares only about the right operand :(.
 
> The bug here seems to be that we aren't doing the l->r conversion for the
> LHS of the TRUTH_OR_EXPR; I'd think that cxx_eval_logical_expression should
> pass false for lval to both recursive calls, there's no case where we
> actually expect an lvalue from a TRUTH_*.

Yeah, that makes sense.

Bootstrap/regtest running on x86_64-pc-linux-gnu, ok for trunk/11.3 if it
passes?

-- >8 --
Here we issue a bogus error for the first assert in the test.  Therein
we have

<retval> = (void) (VIEW_CONVERT_EXPR<bool>(yes) || handle_error ());, VIEW_CONVERT_EXPR<int>(value);

which has a COMPOUND_EXPR, so we get to cxx_eval_constant_expression
<case COMPOUND_EXPR>.  The problem here is that we call

7044             /* Check that the LHS is constant and then discard it.  */
7045             cxx_eval_constant_expression (ctx, op0,
7046                                           true, non_constant_p, overflow_p,
7047                                           jump_target);

where lval is always true, so the PARM_DECL 'yes' is not evaluated into
its value.

Fixed by always passing false for 'lval' in cxx_eval_logical_expression;
there's no case where we actually expect an lvalue from a TRUTH_*.

	PR c++/105321

gcc/cp/ChangeLog:

	* constexpr.cc (cxx_eval_logical_expression): Always pass false for lval
	to cxx_eval_constant_expression.

gcc/testsuite/ChangeLog:

	* g++.dg/cpp0x/constexpr-105321.C: New test.
---
 gcc/cp/constexpr.cc                           |  9 ++++-----
 gcc/testsuite/g++.dg/cpp0x/constexpr-105321.C | 18 ++++++++++++++++++
 2 files changed, 22 insertions(+), 5 deletions(-)
 create mode 100644 gcc/testsuite/g++.dg/cpp0x/constexpr-105321.C


base-commit: 1e6c0e69af8da436e1d1d2d23d8c38410d78ecf2

Comments

Jason Merrill April 21, 2022, 2:15 p.m. UTC | #1 
Ok.

On Thu, Apr 21, 2022, 9:20 AM Marek Polacek <polacek@redhat.com> wrote:

> On Thu, Apr 21, 2022 at 08:56:23AM -0400, Jason Merrill wrote:
> > On 4/20/22 18:40, Marek Polacek wrote:
> > > Here we issue a bogus error for the first assert in the test.  Therein
> > > we have
> > >
> > > <retval> = (void) (VIEW_CONVERT_EXPR<bool>(yes) || handle_error ());,
> VIEW_CONVERT_EXPR<int>(value);
> > >
> > > which has a COMPOUND_EXPR, so we get to cxx_eval_constant_expression
> > > <case COMPOUND_EXPR>.  The problem here is that we call
> > >
> > > 7044             /* Check that the LHS is constant and then discard
> it.  */
> > > 7045             cxx_eval_constant_expression (ctx, op0,
> > > 7046                                           true, non_constant_p,
> overflow_p,
> > > 7047                                           jump_target);
> > >
> > > where lval is always true, so the PARM_DECL 'yes' is not evaluated into
> > > its value.  r218832 changed the argument for 'lval' from false to true:
> > >
> > >     (cxx_eval_constant_expression) [COMPOUND_EXPR]: Pass true for lval.
> > >
> > > but I think we want to pass 'lval' instead.  Jakub tells me that's what
> > > we do for "(void) expr" as well.  [expr.comma] says that the left
> expression
> > > is a discarded-value expression, but [expr.context] doesn't suggest
> that
> > > we should always be passing false for lval as pre-r218832.
> >
> > In a discarded-value expression, we don't do the lvalue-rvalue
> conversion;
> > whether we want an lvalue for the RHS of the comma is irrelevant.
>
> Ah, that's what I misread -- [expr.context]/2.8 cares only about the right
> operand :(.
>
> > The bug here seems to be that we aren't doing the l->r conversion for the
> > LHS of the TRUTH_OR_EXPR; I'd think that cxx_eval_logical_expression
> should
> > pass false for lval to both recursive calls, there's no case where we
> > actually expect an lvalue from a TRUTH_*.
>
> Yeah, that makes sense.
>
> Bootstrap/regtest running on x86_64-pc-linux-gnu, ok for trunk/11.3 if it
> passes?
>
> -- >8 --
> Here we issue a bogus error for the first assert in the test.  Therein
> we have
>
> <retval> = (void) (VIEW_CONVERT_EXPR<bool>(yes) || handle_error ());,
> VIEW_CONVERT_EXPR<int>(value);
>
> which has a COMPOUND_EXPR, so we get to cxx_eval_constant_expression
> <case COMPOUND_EXPR>.  The problem here is that we call
>
> 7044             /* Check that the LHS is constant and then discard it.  */
> 7045             cxx_eval_constant_expression (ctx, op0,
> 7046                                           true, non_constant_p,
> overflow_p,
> 7047                                           jump_target);
>
> where lval is always true, so the PARM_DECL 'yes' is not evaluated into
> its value.
>
> Fixed by always passing false for 'lval' in cxx_eval_logical_expression;
> there's no case where we actually expect an lvalue from a TRUTH_*.
>
>         PR c++/105321
>
> gcc/cp/ChangeLog:
>
>         * constexpr.cc (cxx_eval_logical_expression): Always pass false
> for lval
>         to cxx_eval_constant_expression.
>
> gcc/testsuite/ChangeLog:
>
>         * g++.dg/cpp0x/constexpr-105321.C: New test.
> ---
>  gcc/cp/constexpr.cc                           |  9 ++++-----
>  gcc/testsuite/g++.dg/cpp0x/constexpr-105321.C | 18 ++++++++++++++++++
>  2 files changed, 22 insertions(+), 5 deletions(-)
>  create mode 100644 gcc/testsuite/g++.dg/cpp0x/constexpr-105321.C
>
> diff --git a/gcc/cp/constexpr.cc b/gcc/cp/constexpr.cc
> index e89440e770f..fa65290e938 100644
> --- a/gcc/cp/constexpr.cc
> +++ b/gcc/cp/constexpr.cc
> @@ -4566,19 +4566,18 @@ cxx_eval_bit_cast (const constexpr_ctx *ctx, tree
> t, bool *non_constant_p,
>  static tree
>  cxx_eval_logical_expression (const constexpr_ctx *ctx, tree t,
>                               tree bailout_value, tree continue_value,
> -                            bool lval,
> -                            bool *non_constant_p, bool *overflow_p)
> +                            bool, bool *non_constant_p, bool *overflow_p)
>  {
>    tree r;
>    tree lhs = cxx_eval_constant_expression (ctx, TREE_OPERAND (t, 0),
> -                                          lval,
> -                                          non_constant_p, overflow_p);
> +                                          /*lval*/false, non_constant_p,
> +                                          overflow_p);
>    VERIFY_CONSTANT (lhs);
>    if (tree_int_cst_equal (lhs, bailout_value))
>      return lhs;
>    gcc_assert (tree_int_cst_equal (lhs, continue_value));
>    r = cxx_eval_constant_expression (ctx, TREE_OPERAND (t, 1),
> -                                   lval, non_constant_p,
> +                                   /*lval*/false, non_constant_p,
>                                     overflow_p);
>    VERIFY_CONSTANT (r);
>    return r;
> diff --git a/gcc/testsuite/g++.dg/cpp0x/constexpr-105321.C
> b/gcc/testsuite/g++.dg/cpp0x/constexpr-105321.C
> new file mode 100644
> index 00000000000..adb6830ff22
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/cpp0x/constexpr-105321.C
> @@ -0,0 +1,18 @@
> +// PR c++/105321
> +// { dg-do compile { target c++11 } }
> +
> +bool handle_error();
> +
> +constexpr int echo(int value, bool yes = true) noexcept
> +{
> +    return (yes || handle_error()), value;
> +}
> +
> +static_assert(echo(10) == 10, "");
> +
> +constexpr int echo2(int value, bool no = false) noexcept
> +{
> +    return (!no || handle_error()), value;
> +}
> +
> +static_assert(echo2(10) == 10, "");
>
> base-commit: 1e6c0e69af8da436e1d1d2d23d8c38410d78ecf2
> --
> 2.35.1
>
>
Jakub Jelinek April 21, 2022, 2:22 p.m. UTC | #2 
On Thu, Apr 21, 2022 at 09:20:48AM -0400, Marek Polacek via Gcc-patches wrote:
> --- a/gcc/cp/constexpr.cc
> +++ b/gcc/cp/constexpr.cc
> @@ -4566,19 +4566,18 @@ cxx_eval_bit_cast (const constexpr_ctx *ctx, tree t, bool *non_constant_p,
>  static tree
>  cxx_eval_logical_expression (const constexpr_ctx *ctx, tree t,
>                               tree bailout_value, tree continue_value,
> -			     bool lval,
> -			     bool *non_constant_p, bool *overflow_p)
> +			     bool, bool *non_constant_p, bool *overflow_p)

Wouldn't it be better to remove the unused lval argument from
cxx_eval_logical_expression and adjust the 2 callers?

	Jakub
Jason Merrill April 21, 2022, 2:25 p.m. UTC | #3 
Yes, also ok with that change.

On Thu, Apr 21, 2022, 10:22 AM Jakub Jelinek <jakub@redhat.com> wrote:

> On Thu, Apr 21, 2022 at 09:20:48AM -0400, Marek Polacek via Gcc-patches
> wrote:
> > --- a/gcc/cp/constexpr.cc
> > +++ b/gcc/cp/constexpr.cc
> > @@ -4566,19 +4566,18 @@ cxx_eval_bit_cast (const constexpr_ctx *ctx,
> tree t, bool *non_constant_p,
> >  static tree
> >  cxx_eval_logical_expression (const constexpr_ctx *ctx, tree t,
> >                               tree bailout_value, tree continue_value,
> > -                          bool lval,
> > -                          bool *non_constant_p, bool *overflow_p)
> > +                          bool, bool *non_constant_p, bool *overflow_p)
>
> Wouldn't it be better to remove the unused lval argument from
> cxx_eval_logical_expression and adjust the 2 callers?
>
>         Jakub
>
>
Marek Polacek April 21, 2022, 2:34 p.m. UTC | #4 
On Thu, Apr 21, 2022 at 04:22:03PM +0200, Jakub Jelinek wrote:
> On Thu, Apr 21, 2022 at 09:20:48AM -0400, Marek Polacek via Gcc-patches wrote:
> > --- a/gcc/cp/constexpr.cc
> > +++ b/gcc/cp/constexpr.cc
> > @@ -4566,19 +4566,18 @@ cxx_eval_bit_cast (const constexpr_ctx *ctx, tree t, bool *non_constant_p,
> >  static tree
> >  cxx_eval_logical_expression (const constexpr_ctx *ctx, tree t,
> >                               tree bailout_value, tree continue_value,
> > -			     bool lval,
> > -			     bool *non_constant_p, bool *overflow_p)
> > +			     bool, bool *non_constant_p, bool *overflow_p)
> 
> Wouldn't it be better to remove the unused lval argument from
> cxx_eval_logical_expression and adjust the 2 callers?

I'm going to fix it with this patch, sorry for not doing it in the
original patch.

gcc/cp/ChangeLog:

	* constexpr.cc (cxx_eval_logical_expression): Remove unused
	parameter.
	(cxx_eval_constant_expression) <case TRUTH_ANDIF_EXPR>,
	<case TRUTH_OR_EXPR>: Adjust calls to cxx_eval_logical_expression.
---
 gcc/cp/constexpr.cc | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/gcc/cp/constexpr.cc b/gcc/cp/constexpr.cc
index fa65290e938..47d5113ace2 100644
--- a/gcc/cp/constexpr.cc
+++ b/gcc/cp/constexpr.cc
@@ -4566,7 +4566,7 @@ cxx_eval_bit_cast (const constexpr_ctx *ctx, tree t, bool *non_constant_p,
 static tree
 cxx_eval_logical_expression (const constexpr_ctx *ctx, tree t,
                              tree bailout_value, tree continue_value,
-			     bool, bool *non_constant_p, bool *overflow_p)
+			     bool *non_constant_p, bool *overflow_p)
 {
   tree r;
   tree lhs = cxx_eval_constant_expression (ctx, TREE_OPERAND (t, 0),
@@ -7105,7 +7105,6 @@ cxx_eval_constant_expression (const constexpr_ctx *ctx, tree t,
     case TRUTH_ANDIF_EXPR:
       r = cxx_eval_logical_expression (ctx, t, boolean_false_node,
 				       boolean_true_node,
-				       lval,
 				       non_constant_p, overflow_p);
       break;
 
@@ -7113,7 +7112,6 @@ cxx_eval_constant_expression (const constexpr_ctx *ctx, tree t,
     case TRUTH_ORIF_EXPR:
       r = cxx_eval_logical_expression (ctx, t, boolean_true_node,
 				       boolean_false_node,
-				       lval,
 				       non_constant_p, overflow_p);
       break;
 

base-commit: 93b65ed9706e2ceb4be7b28c9ff9be759e68a614
diff mbox series

Patch

diff --git a/gcc/cp/constexpr.cc b/gcc/cp/constexpr.cc
index e89440e770f..fa65290e938 100644
--- a/gcc/cp/constexpr.cc
+++ b/gcc/cp/constexpr.cc
@@ -4566,19 +4566,18 @@  cxx_eval_bit_cast (const constexpr_ctx *ctx, tree t, bool *non_constant_p,
 static tree
 cxx_eval_logical_expression (const constexpr_ctx *ctx, tree t,
                              tree bailout_value, tree continue_value,
-			     bool lval,
-			     bool *non_constant_p, bool *overflow_p)
+			     bool, bool *non_constant_p, bool *overflow_p)
 {
   tree r;
   tree lhs = cxx_eval_constant_expression (ctx, TREE_OPERAND (t, 0),
-					   lval,
-					   non_constant_p, overflow_p);
+					   /*lval*/false, non_constant_p,
+					   overflow_p);
   VERIFY_CONSTANT (lhs);
   if (tree_int_cst_equal (lhs, bailout_value))
     return lhs;
   gcc_assert (tree_int_cst_equal (lhs, continue_value));
   r = cxx_eval_constant_expression (ctx, TREE_OPERAND (t, 1),
-				    lval, non_constant_p,
+				    /*lval*/false, non_constant_p,
 				    overflow_p);
   VERIFY_CONSTANT (r);
   return r;
diff --git a/gcc/testsuite/g++.dg/cpp0x/constexpr-105321.C b/gcc/testsuite/g++.dg/cpp0x/constexpr-105321.C
new file mode 100644
index 00000000000..adb6830ff22
--- /dev/null
+++ b/gcc/testsuite/g++.dg/cpp0x/constexpr-105321.C
@@ -0,0 +1,18 @@ 
+// PR c++/105321
+// { dg-do compile { target c++11 } }
+
+bool handle_error();
+
+constexpr int echo(int value, bool yes = true) noexcept
+{
+    return (yes || handle_error()), value;
+}
+
+static_assert(echo(10) == 10, "");
+
+constexpr int echo2(int value, bool no = false) noexcept
+{
+    return (!no || handle_error()), value;
+}
+
+static_assert(echo2(10) == 10, "");