[PATCH] Add LVAL argument to c_fully_fold* and propagate it through (PR c/66618, PR c/69960)
Hi!
Here is an attempt to fix these two PRs. The C++ FE already has an LVAL
bool that it propagates through constexpr.c functions, or in
cp-gimplify.c through calling cp_fold_{maybe_,}rvalue where appropriate.
The C c_fully_fold was instead just calling decl_constant_value_for_optimization
in some spots where we've previously called c_fully_fold{_internal,} on
some rvalue.
decl_constant_value_for_optimization wasn't doing anything when -O0,
so we got different behavior between -O0 and -O1+ on what is accepted
in static initializers. Furthermore, we needed the hack to ignore
vars with ARRAY_TYPE or BLKmode, so that we actually wouldn't fold lvalues
to rvalues.
This patch adds LVAL to c_fully_fold{_internal,} and does what
decl_constant_value_for_optimization did inside of it (without the hacks,
furthermore, and additionally for -O0 when in initializers for consistency),
of course only if LVAL is false. Additionally, I've added folding of
"foo"[2] into 'o'. We have it in gimple-fold.c or so, so that one
isn't performed when not in_init.
Not sure about the COND_EXPR/VEC_COND_EXPR cases, right now I'm passing
false as LVAL for the first operand (condition) and lval as LVAL for the
other two (i.e. if called with lval == true on the whole *_COND_EXPR
decl_constant_value_for_optimization etc. isn't performed on op1/op2, while
without it it is). Can one take address of the whole COND_EXPR, or
have it on LHS of anything in C? If not, perhaps I could just pass false
in all the 3 calls. Though, then likely it would be called with lval == false
anyway.
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
2017-11-08 Jakub Jelinek
PR c/66618
PR c/69960
c-family/
* c-common.h (c_fully_fold): Add LVAL argument defaulted to false.
c/
* c-parser.c (c_parser_omp_atomic): Pass true as LVAL to c_fully_fold
where needed.
* c-typeck.c (build_unary_op, build_modify_expr, build_asm_expr,
handle_omp_array_sections): Likewise.
(digest_init): Don't call decl_constant_value_for_optimization.
* c-tree.h (decl_constant_value_for_optimization): Removed.
* c-fold.c (c_fully_fold_internal): Add LVAL argument, propagate
it through recursive calls. For VAR_P call decl_constant_value and
unshare if not LVAL and either optimizing or IN_INIT. Remove
decl_constant_value_for_optimization calls. If IN_INIT and not LVAL,
fold ARRAY_REF with STRING_CST and INTEGER_CST operands.
(c_fully_fold): Add LVAL argument, pass it through to
c_fully_fold_internal.
(decl_constant_value_for_optimization): Removed.
cp/
* cp-gimplify.c (c_fully_fold): Add LVAL argument, call
cp_fold_maybe_rvalue instead of cp_fold_rvalue and pass it !LVAL.
testsuite/
* gcc.dg/pr69960.c: New test.
* gcc.dg/pr66618.c: New test.
--- gcc/c-family/c-common.h.jj 2017-10-12 20:51:30.0 +0200
+++ gcc/c-family/c-common.h 2017-11-08 12:20:59.874481318 +0100
@@ -827,7 +827,7 @@ extern tree c_build_bitfield_integer_typ
extern enum conversion_safety unsafe_conversion_p (location_t, tree, tree,
tree,
bool);
extern bool decl_with_nonnull_addr_p (const_tree);
-extern tree c_fully_fold (tree, bool, bool *);
+extern tree c_fully_fold (tree, bool, bool *, bool = false);
extern tree c_wrap_maybe_const (tree, bool);
extern tree c_common_truthvalue_conversion (location_t, tree);
extern void c_apply_type_quals_to_decl (int, tree);
--- gcc/c/c-parser.c.jj 2017-11-08 12:29:02.0 +0100
+++ gcc/c/c-parser.c2017-11-08 13:53:39.756889500 +0100
@@ -14748,7 +14748,7 @@ c_parser_omp_atomic (location_t loc, c_p
case NOP_EXPR: /* atomic write */
v = c_parser_cast_expression (parser, NULL).value;
non_lvalue_p = !lvalue_p (v);
- v = c_fully_fold (v, false, NULL);
+ v = c_fully_fold (v, false, NULL, true);
if (v == error_mark_node)
goto saw_error;
if (non_lvalue_p)
@@ -14767,7 +14767,7 @@ c_parser_omp_atomic (location_t loc, c_p
{
lhs = c_parser_cast_expression (parser, NULL).value;
non_lvalue_p = !lvalue_p (lhs);
- lhs = c_fully_fold (lhs, false, NULL);
+ lhs = c_fully_fold (lhs, false, NULL, true);
if (lhs == error_mark_node)
goto saw_error;
if (non_lvalue_p)
@@ -14793,7 +14793,7 @@ c_parser_omp_atomic (location_t loc, c_p
{
v = c_parser_cast_expression (parser, NULL).value;
non_lvalue_p = !lvalue_p (v);
- v = c_fully_fold (v, false, NULL);
+ v = c_fully_fold (v, false, NULL, true);
if (v == error_mark_node)
goto saw_error;
if (non_lvalue_p)
@@ -14814,7 +14814,7 @@ restart:
lhs = expr.value;
expr = default_function_array_conversion (eloc, expr);
unfolded_lhs = expr.value;
- lhs = c_fully_fold (lhs, false, NULL);
+ lhs = c_full
Re: [PATCH] Add LVAL argument to c_fully_fold* and propagate it through (PR c/66618, PR c/69960)
On 11/09/2017 11:54 AM, Jakub Jelinek wrote: On Wed, Nov 08, 2017 at 06:57:55PM +0100, Jakub Jelinek wrote: On Wed, Nov 08, 2017 at 06:51:34PM +0100, Marek Polacek wrote: Ok, so like this if it passes bootstrap/regtest? Changes from the last patch: 1) false instead of lval for COMPOUND_EXPR and *COND_EXPR op1/op2 So... Oops, I've hand-edited it in the patch and then regenerated the patch after doing there more changes. Here is updated patch with that in again: FYI, bootstrapped/regtested on x86_64-linux and i686-linux successfully. This caused: PR83590 - [nvptx] openacc reduction C regressions ( https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83590 ). Thanks, - Tom
Re: [PATCH] Add LVAL argument to c_fully_fold* and propagate it through (PR c/66618, PR c/69960)
On Wed, Nov 08, 2017 at 05:22:45PM +0100, Jakub Jelinek wrote:
> Hi!
>
> Here is an attempt to fix these two PRs. The C++ FE already has an LVAL
Nice!
> bool that it propagates through constexpr.c functions, or in
> cp-gimplify.c through calling cp_fold_{maybe_,}rvalue where appropriate.
> The C c_fully_fold was instead just calling
> decl_constant_value_for_optimization
> in some spots where we've previously called c_fully_fold{_internal,} on
> some rvalue.
> decl_constant_value_for_optimization wasn't doing anything when -O0,
> so we got different behavior between -O0 and -O1+ on what is accepted
> in static initializers. Furthermore, we needed the hack to ignore
> vars with ARRAY_TYPE or BLKmode, so that we actually wouldn't fold lvalues
> to rvalues.
>
> This patch adds LVAL to c_fully_fold{_internal,} and does what
> decl_constant_value_for_optimization did inside of it (without the hacks,
> furthermore, and additionally for -O0 when in initializers for consistency),
> of course only if LVAL is false. Additionally, I've added folding of
> "foo"[2] into 'o'. We have it in gimple-fold.c or so, so that one
> isn't performed when not in_init.
>
> Not sure about the COND_EXPR/VEC_COND_EXPR cases, right now I'm passing
> false as LVAL for the first operand (condition) and lval as LVAL for the
> other two (i.e. if called with lval == true on the whole *_COND_EXPR
> decl_constant_value_for_optimization etc. isn't performed on op1/op2, while
> without it it is). Can one take address of the whole COND_EXPR, or
> have it on LHS of anything in C?
I don't think so, the ?: operator does not yield an lvalue.
> @@ -218,12 +235,51 @@ c_fully_fold_internal (tree expr, bool i
>op2 = TREE_OPERAND (expr, 2);
>op3 = TREE_OPERAND (expr, 3);
>op0 = c_fully_fold_internal (op0, in_init, maybe_const_operands,
> -maybe_const_itself, for_int_const);
> +maybe_const_itself, for_int_const, lval);
>STRIP_TYPE_NOPS (op0);
>op1 = c_fully_fold_internal (op1, in_init, maybe_const_operands,
> -maybe_const_itself, for_int_const);
> +maybe_const_itself, for_int_const, false);
>STRIP_TYPE_NOPS (op1);
> - op1 = decl_constant_value_for_optimization (op1);
> + /* Fold "foo"[2] in initializers. */
> + if (!lval
> + && in_init
> + && TREE_CODE (op0) == STRING_CST
> + && TREE_CODE (op1) == INTEGER_CST
> + && TREE_CODE (TREE_TYPE (op0)) == ARRAY_TYPE
Does this also handle 2["foobar"]?
> + && tree_fits_uhwi_p (op1))
> + {
> + tree ary = op0;
> + tree index = op1;
> + unsigned len = 0;
> + tree elem_type = TREE_TYPE (TREE_TYPE (ary));
> + unsigned elem_nchars = (TYPE_PRECISION (elem_type)
> + / TYPE_PRECISION (char_type_node));
> + len = (unsigned) TREE_STRING_LENGTH (ary) / elem_nchars;
> +
> + tree nelts = array_type_nelts (TREE_TYPE (ary));
> + bool dummy1 = true, dummy2 = true;
> + nelts = c_fully_fold_internal (nelts, in_init, &dummy1, &dummy2,
> + for_int_const, false);
> + unsigned HOST_WIDE_INT i = tree_to_uhwi (index);
> + if (tree_int_cst_le (index, nelts)
> + && i < len
> + && i + elem_nchars <= len)
> + {
> + tree type = TREE_TYPE (expr);
> + if (elem_nchars == 1)
> + ret = build_int_cst (type,
> + TREE_STRING_POINTER (ary)[i]);
> + else
> + {
> + const unsigned char *ptr
> + = ((const unsigned char *)TREE_STRING_POINTER (ary)
> ++ i * elem_nchars);
> + ret = native_interpret_expr (type, ptr, elem_nchars);
> + }
> + if (ret)
> + goto out;
> + }
> + }
So this code comes from gimple-fold.c? I would probably move it to a new
function.
Marek
Re: [PATCH] Add LVAL argument to c_fully_fold* and propagate it through (PR c/66618, PR c/69960)
On Wed, Nov 08, 2017 at 05:42:07PM +0100, Marek Polacek wrote:
> > Not sure about the COND_EXPR/VEC_COND_EXPR cases, right now I'm passing
> > false as LVAL for the first operand (condition) and lval as LVAL for the
> > other two (i.e. if called with lval == true on the whole *_COND_EXPR
> > decl_constant_value_for_optimization etc. isn't performed on op1/op2, while
> > without it it is). Can one take address of the whole COND_EXPR, or
> > have it on LHS of anything in C?
>
> I don't think so, the ?: operator does not yield an lvalue.
Ok, I'll pass false then.
> > @@ -218,12 +235,51 @@ c_fully_fold_internal (tree expr, bool i
> >op2 = TREE_OPERAND (expr, 2);
> >op3 = TREE_OPERAND (expr, 3);
> >op0 = c_fully_fold_internal (op0, in_init, maybe_const_operands,
> > - maybe_const_itself, for_int_const);
> > + maybe_const_itself, for_int_const, lval);
> >STRIP_TYPE_NOPS (op0);
> >op1 = c_fully_fold_internal (op1, in_init, maybe_const_operands,
> > - maybe_const_itself, for_int_const);
> > + maybe_const_itself, for_int_const, false);
> >STRIP_TYPE_NOPS (op1);
> > - op1 = decl_constant_value_for_optimization (op1);
> > + /* Fold "foo"[2] in initializers. */
> > + if (!lval
> > + && in_init
> > + && TREE_CODE (op0) == STRING_CST
> > + && TREE_CODE (op1) == INTEGER_CST
> > + && TREE_CODE (TREE_TYPE (op0)) == ARRAY_TYPE
>
> Does this also handle 2["foobar"]?
Yes, because build_array_ref has:
2607 if (TREE_CODE (TREE_TYPE (array)) != ARRAY_TYPE
2608 && TREE_CODE (TREE_TYPE (array)) != POINTER_TYPE
2609 /* Allow vector[index] but not index[vector]. */
2610 && !VECTOR_TYPE_P (TREE_TYPE (array)))
2611{
2612 if (TREE_CODE (TREE_TYPE (index)) != ARRAY_TYPE
2613 && TREE_CODE (TREE_TYPE (index)) != POINTER_TYPE)
2614{
2615 error_at (loc,
2616"subscripted value is neither array nor pointer nor
vector");
2617
2618 return error_mark_node;
2619}
2620 std::swap (array, index);
2621 swapped = true;
2622}
> > + && tree_fits_uhwi_p (op1))
> > + {
> > + tree ary = op0;
> > + tree index = op1;
> > + unsigned len = 0;
> > + tree elem_type = TREE_TYPE (TREE_TYPE (ary));
> > + unsigned elem_nchars = (TYPE_PRECISION (elem_type)
> > + / TYPE_PRECISION (char_type_node));
> > + len = (unsigned) TREE_STRING_LENGTH (ary) / elem_nchars;
> > +
> > + tree nelts = array_type_nelts (TREE_TYPE (ary));
> > + bool dummy1 = true, dummy2 = true;
> > + nelts = c_fully_fold_internal (nelts, in_init, &dummy1, &dummy2,
> > +for_int_const, false);
> > + unsigned HOST_WIDE_INT i = tree_to_uhwi (index);
> > + if (tree_int_cst_le (index, nelts)
> > + && i < len
> > + && i + elem_nchars <= len)
> > + {
> > + tree type = TREE_TYPE (expr);
> > + if (elem_nchars == 1)
> > + ret = build_int_cst (type,
> > +TREE_STRING_POINTER (ary)[i]);
> > + else
> > + {
> > + const unsigned char *ptr
> > + = ((const unsigned char *)TREE_STRING_POINTER (ary)
> > + + i * elem_nchars);
> > + ret = native_interpret_expr (type, ptr, elem_nchars);
> > + }
> > + if (ret)
> > + goto out;
> > + }
> > + }
>
> So this code comes from gimple-fold.c? I would probably move it to a new
> function.
No, this comes from very simplified cp/constexpr.c (extract_string_elt,
cxx_eval_array_reference). Can outline it into a separate function, sure.
Jakub
Re: [PATCH] Add LVAL argument to c_fully_fold* and propagate it through (PR c/66618, PR c/69960)
On Wed, 8 Nov 2017, Jakub Jelinek wrote: > of course only if LVAL is false. Additionally, I've added folding of > "foo"[2] into 'o'. We have it in gimple-fold.c or so, so that one Note that if the 2 there comes from an overflowing expression that's not valid as an extension to constant expressions in initializers (that is, static char c = "foo"[INT_MAX * -2]; should result in an error with -pedantic-errors because of the overflow, just as INT_MAX * -2 by itself wouldn't be a valid initializer in that case). > Not sure about the COND_EXPR/VEC_COND_EXPR cases, right now I'm passing > false as LVAL for the first operand (condition) and lval as LVAL for the > other two (i.e. if called with lval == true on the whole *_COND_EXPR > decl_constant_value_for_optimization etc. isn't performed on op1/op2, while > without it it is). Can one take address of the whole COND_EXPR, or > have it on LHS of anything in C? If not, perhaps I could just pass false > in all the 3 calls. Though, then likely it would be called with lval == false > anyway. Conditional and compound expressions are never valid lvalues in C. -- Joseph S. Myers jos...@codesourcery.com
Re: [PATCH] Add LVAL argument to c_fully_fold* and propagate it through (PR c/66618, PR c/69960)
On Wed, Nov 08, 2017 at 05:17:50PM +, Joseph Myers wrote:
> On Wed, 8 Nov 2017, Jakub Jelinek wrote:
>
> > of course only if LVAL is false. Additionally, I've added folding of
> > "foo"[2] into 'o'. We have it in gimple-fold.c or so, so that one
>
> Note that if the 2 there comes from an overflowing expression that's not
> valid as an extension to constant expressions in initializers (that is,
>
> static char c = "foo"[INT_MAX * -2];
>
> should result in an error with -pedantic-errors because of the overflow,
> just as INT_MAX * -2 by itself wouldn't be a valid initializer in that
> case).
>
> > Not sure about the COND_EXPR/VEC_COND_EXPR cases, right now I'm passing
> > false as LVAL for the first operand (condition) and lval as LVAL for the
> > other two (i.e. if called with lval == true on the whole *_COND_EXPR
> > decl_constant_value_for_optimization etc. isn't performed on op1/op2, while
> > without it it is). Can one take address of the whole COND_EXPR, or
> > have it on LHS of anything in C? If not, perhaps I could just pass false
> > in all the 3 calls. Though, then likely it would be called with lval ==
> > false
> > anyway.
>
> Conditional and compound expressions are never valid lvalues in C.
Ok, so like this if it passes bootstrap/regtest?
Changes from the last patch:
1) false instead of lval for COMPOUND_EXPR and *COND_EXPR op1/op2
2) outlined the "foo"[2] folding into c_fold_array_ref
3) added !TREE_OVERFLOW check there
4) added another testcase
2017-11-08 Jakub Jelinek
PR c/66618
PR c/69960
c-family/
* c-common.h (c_fully_fold): Add LVAL argument defaulted to false.
c/
* c-parser.c (c_parser_omp_atomic): Pass true as LVAL to c_fully_fold
where needed.
* c-typeck.c (build_unary_op, build_modify_expr, build_asm_expr,
handle_omp_array_sections): Likewise.
(digest_init): Don't call decl_constant_value_for_optimization.
* c-tree.h (decl_constant_value_for_optimization): Removed.
* c-fold.c (c_fold_array_ref): New function.
(c_fully_fold_internal): Add LVAL argument, propagate it through
recursive calls. For VAR_P call decl_constant_value and
unshare if not LVAL and either optimizing or IN_INIT. Remove
decl_constant_value_for_optimization calls. If IN_INIT and not LVAL,
fold ARRAY_REF with STRING_CST and INTEGER_CST operands.
(c_fully_fold): Add LVAL argument, pass it through to
c_fully_fold_internal.
(decl_constant_value_for_optimization): Removed.
cp/
* cp-gimplify.c (c_fully_fold): Add LVAL argument, call
cp_fold_maybe_rvalue instead of cp_fold_rvalue and pass it !LVAL.
testsuite/
* gcc.dg/pr69960.c: New test.
* gcc.dg/pr66618.c: New test.
* gcc.dg/pr66618-2.c: New test.
--- gcc/c-family/c-common.h.jj 2017-11-08 16:19:30.309488815 +0100
+++ gcc/c-family/c-common.h 2017-11-08 17:49:00.548178246 +0100
@@ -827,7 +827,7 @@ extern tree c_build_bitfield_integer_typ
extern enum conversion_safety unsafe_conversion_p (location_t, tree, tree,
tree,
bool);
extern bool decl_with_nonnull_addr_p (const_tree);
-extern tree c_fully_fold (tree, bool, bool *);
+extern tree c_fully_fold (tree, bool, bool *, bool = false);
extern tree c_wrap_maybe_const (tree, bool);
extern tree c_common_truthvalue_conversion (location_t, tree);
extern void c_apply_type_quals_to_decl (int, tree);
--- gcc/c/c-parser.c.jj 2017-11-08 16:19:30.569485632 +0100
+++ gcc/c/c-parser.c2017-11-08 17:49:00.552178196 +0100
@@ -14748,7 +14748,7 @@ c_parser_omp_atomic (location_t loc, c_p
case NOP_EXPR: /* atomic write */
v = c_parser_cast_expression (parser, NULL).value;
non_lvalue_p = !lvalue_p (v);
- v = c_fully_fold (v, false, NULL);
+ v = c_fully_fold (v, false, NULL, true);
if (v == error_mark_node)
goto saw_error;
if (non_lvalue_p)
@@ -14767,7 +14767,7 @@ c_parser_omp_atomic (location_t loc, c_p
{
lhs = c_parser_cast_expression (parser, NULL).value;
non_lvalue_p = !lvalue_p (lhs);
- lhs = c_fully_fold (lhs, false, NULL);
+ lhs = c_fully_fold (lhs, false, NULL, true);
if (lhs == error_mark_node)
goto saw_error;
if (non_lvalue_p)
@@ -14793,7 +14793,7 @@ c_parser_omp_atomic (location_t loc, c_p
{
v = c_parser_cast_expression (parser, NULL).value;
non_lvalue_p = !lvalue_p (v);
- v = c_fully_fold (v, false, NULL);
+ v = c_fully_fold (v, false, NULL, true);
if (v == error_mark_node)
goto saw_error;
if (non_lvalue_p)
@@ -14814,7 +14814,7 @@ restart:
lhs = expr.value;
expr = default_function_array_conversion (eloc, expr);
unfolded_lhs = expr.value;
- lhs = c_fully_fold (lhs, false, NULL);
+ lhs = c_fully_fold (lhs, false, NULL, true);
orig_lhs = lhs;
switch
Re: [PATCH] Add LVAL argument to c_fully_fold* and propagate it through (PR c/66618, PR c/69960)
On Wed, Nov 08, 2017 at 06:38:21PM +0100, Jakub Jelinek wrote: > On Wed, Nov 08, 2017 at 05:17:50PM +, Joseph Myers wrote: > > On Wed, 8 Nov 2017, Jakub Jelinek wrote: > > > > > of course only if LVAL is false. Additionally, I've added folding of > > > "foo"[2] into 'o'. We have it in gimple-fold.c or so, so that one > > > > Note that if the 2 there comes from an overflowing expression that's not > > valid as an extension to constant expressions in initializers (that is, > > > > static char c = "foo"[INT_MAX * -2]; > > > > should result in an error with -pedantic-errors because of the overflow, > > just as INT_MAX * -2 by itself wouldn't be a valid initializer in that > > case). > > > > > Not sure about the COND_EXPR/VEC_COND_EXPR cases, right now I'm passing > > > false as LVAL for the first operand (condition) and lval as LVAL for the > > > other two (i.e. if called with lval == true on the whole *_COND_EXPR > > > decl_constant_value_for_optimization etc. isn't performed on op1/op2, > > > while > > > without it it is). Can one take address of the whole COND_EXPR, or > > > have it on LHS of anything in C? If not, perhaps I could just pass false > > > in all the 3 calls. Though, then likely it would be called with lval == > > > false > > > anyway. > > > > Conditional and compound expressions are never valid lvalues in C. > > Ok, so like this if it passes bootstrap/regtest? > > Changes from the last patch: > 1) false instead of lval for COMPOUND_EXPR and *COND_EXPR op1/op2 So... > @@ -278,21 +339,16 @@ c_fully_fold_internal (tree expr, bool i >orig_op0 = op0 = TREE_OPERAND (expr, 0); >orig_op1 = op1 = TREE_OPERAND (expr, 1); >op0 = c_fully_fold_internal (op0, in_init, maybe_const_operands, > -maybe_const_itself, for_int_const); > +maybe_const_itself, for_int_const, > +op0_lval); >STRIP_TYPE_NOPS (op0); > - if (code != MODIFY_EXPR > - && code != PREDECREMENT_EXPR > - && code != PREINCREMENT_EXPR > - && code != POSTDECREMENT_EXPR > - && code != POSTINCREMENT_EXPR) > - op0 = decl_constant_value_for_optimization (op0); >/* The RHS of a MODIFY_EXPR was fully folded when building that >expression for the sake of conversion warnings. */ >if (code != MODIFY_EXPR) > op1 = c_fully_fold_internal (op1, in_init, maybe_const_operands, > - maybe_const_itself, for_int_const); > + maybe_const_itself, for_int_const, > + code == COMPOUND_EXPR ? lval : false); ...shouldn't we be passing just false here? Marek
Re: [PATCH] Add LVAL argument to c_fully_fold* and propagate it through (PR c/66618, PR c/69960)
On Wed, Nov 08, 2017 at 06:51:34PM +0100, Marek Polacek wrote:
> > Ok, so like this if it passes bootstrap/regtest?
> >
> > Changes from the last patch:
> > 1) false instead of lval for COMPOUND_EXPR and *COND_EXPR op1/op2
>
> So...
Oops, I've hand-edited it in the patch and then regenerated the patch
after doing there more changes.
Here is updated patch with that in again:
2017-11-08 Jakub Jelinek
PR c/66618
PR c/69960
c-family/
* c-common.h (c_fully_fold): Add LVAL argument defaulted to false.
c/
* c-parser.c (c_parser_omp_atomic): Pass true as LVAL to c_fully_fold
where needed.
* c-typeck.c (build_unary_op, build_modify_expr, build_asm_expr,
handle_omp_array_sections): Likewise.
(digest_init): Don't call decl_constant_value_for_optimization.
* c-tree.h (decl_constant_value_for_optimization): Removed.
* c-fold.c (c_fold_array_ref): New function.
(c_fully_fold_internal): Add LVAL argument, propagate it through
recursive calls. For VAR_P call decl_constant_value and
unshare if not LVAL and either optimizing or IN_INIT. Remove
decl_constant_value_for_optimization calls. If IN_INIT and not LVAL,
fold ARRAY_REF with STRING_CST and INTEGER_CST operands.
(c_fully_fold): Add LVAL argument, pass it through to
c_fully_fold_internal.
(decl_constant_value_for_optimization): Removed.
cp/
* cp-gimplify.c (c_fully_fold): Add LVAL argument, call
cp_fold_maybe_rvalue instead of cp_fold_rvalue and pass it !LVAL.
testsuite/
* gcc.dg/pr69960.c: New test.
* gcc.dg/pr66618.c: New test.
* gcc.dg/pr66618-2.c: New test.
--- gcc/c-family/c-common.h.jj 2017-11-08 16:19:30.309488815 +0100
+++ gcc/c-family/c-common.h 2017-11-08 17:49:00.548178246 +0100
@@ -827,7 +827,7 @@ extern tree c_build_bitfield_integer_typ
extern enum conversion_safety unsafe_conversion_p (location_t, tree, tree,
tree,
bool);
extern bool decl_with_nonnull_addr_p (const_tree);
-extern tree c_fully_fold (tree, bool, bool *);
+extern tree c_fully_fold (tree, bool, bool *, bool = false);
extern tree c_wrap_maybe_const (tree, bool);
extern tree c_common_truthvalue_conversion (location_t, tree);
extern void c_apply_type_quals_to_decl (int, tree);
--- gcc/c/c-parser.c.jj 2017-11-08 16:19:30.569485632 +0100
+++ gcc/c/c-parser.c2017-11-08 17:49:00.552178196 +0100
@@ -14748,7 +14748,7 @@ c_parser_omp_atomic (location_t loc, c_p
case NOP_EXPR: /* atomic write */
v = c_parser_cast_expression (parser, NULL).value;
non_lvalue_p = !lvalue_p (v);
- v = c_fully_fold (v, false, NULL);
+ v = c_fully_fold (v, false, NULL, true);
if (v == error_mark_node)
goto saw_error;
if (non_lvalue_p)
@@ -14767,7 +14767,7 @@ c_parser_omp_atomic (location_t loc, c_p
{
lhs = c_parser_cast_expression (parser, NULL).value;
non_lvalue_p = !lvalue_p (lhs);
- lhs = c_fully_fold (lhs, false, NULL);
+ lhs = c_fully_fold (lhs, false, NULL, true);
if (lhs == error_mark_node)
goto saw_error;
if (non_lvalue_p)
@@ -14793,7 +14793,7 @@ c_parser_omp_atomic (location_t loc, c_p
{
v = c_parser_cast_expression (parser, NULL).value;
non_lvalue_p = !lvalue_p (v);
- v = c_fully_fold (v, false, NULL);
+ v = c_fully_fold (v, false, NULL, true);
if (v == error_mark_node)
goto saw_error;
if (non_lvalue_p)
@@ -14814,7 +14814,7 @@ restart:
lhs = expr.value;
expr = default_function_array_conversion (eloc, expr);
unfolded_lhs = expr.value;
- lhs = c_fully_fold (lhs, false, NULL);
+ lhs = c_fully_fold (lhs, false, NULL, true);
orig_lhs = lhs;
switch (TREE_CODE (lhs))
{
@@ -14954,15 +14954,19 @@ restart:
if (c_tree_equal (TREE_OPERAND (rhs1, 0), unfolded_lhs))
{
opcode = TREE_CODE (rhs1);
- rhs = c_fully_fold (TREE_OPERAND (rhs1, 1), false, NULL);
- rhs1 = c_fully_fold (TREE_OPERAND (rhs1, 0), false, NULL);
+ rhs = c_fully_fold (TREE_OPERAND (rhs1, 1), false, NULL,
+ true);
+ rhs1 = c_fully_fold (TREE_OPERAND (rhs1, 0), false, NULL,
+ true);
goto stmt_done;
}
if (c_tree_equal (TREE_OPERAND (rhs1, 1), unfolded_lhs))
{
opcode = TREE_CODE (rhs1);
- rhs = c_fully_fold (TREE_OPERAND (rhs1, 0), false, NULL);
- rhs1 = c_fully_fold (TREE_OPERAND (rhs1, 1), false, NULL);
+ rhs = c_fully_fold (TREE_OPERAND (rhs1, 0), false, NULL,
+ true);
+ rhs1 = c_fully_fold (TREE_OPERAND (rhs1, 1), false, NULL
Re: [PATCH] Add LVAL argument to c_fully_fold* and propagate it through (PR c/66618, PR c/69960)
On Wed, Nov 08, 2017 at 06:57:55PM +0100, Jakub Jelinek wrote: > On Wed, Nov 08, 2017 at 06:51:34PM +0100, Marek Polacek wrote: > > > Ok, so like this if it passes bootstrap/regtest? > > > > > > Changes from the last patch: > > > 1) false instead of lval for COMPOUND_EXPR and *COND_EXPR op1/op2 > > > > So... > > Oops, I've hand-edited it in the patch and then regenerated the patch > after doing there more changes. > > Here is updated patch with that in again: FYI, bootstrapped/regtested on x86_64-linux and i686-linux successfully. Jakub
Re: [PATCH] Add LVAL argument to c_fully_fold* and propagate it through (PR c/66618, PR c/69960)
On Wed, 8 Nov 2017, Jakub Jelinek wrote: > On Wed, Nov 08, 2017 at 06:51:34PM +0100, Marek Polacek wrote: > > > Ok, so like this if it passes bootstrap/regtest? > > > > > > Changes from the last patch: > > > 1) false instead of lval for COMPOUND_EXPR and *COND_EXPR op1/op2 > > > > So... > > Oops, I've hand-edited it in the patch and then regenerated the patch > after doing there more changes. > > Here is updated patch with that in again: > > 2017-11-08 Jakub Jelinek > > PR c/66618 > PR c/69960 OK. -- Joseph S. Myers jos...@codesourcery.com
