| Message ID | 20160304203711.GM3017@tucnak.redhat.com |
|---|---|
| State | New |
| Headers | show |
On 03/04/2016 01:37 PM, Jakub Jelinek wrote: > Hi! > > As mentioned in the PR, on the following testcase we emit invalid > error on s390* in the va_arg ((ap), int) case, va_arg (ap, int) is fine > (and the former is fine in -std=c++11 or -std=c++98 too). > The bug only triggers in constructors (or destructors), and happens because > build_va_arg creates ADDR_EXPR that has slightly different type, and has > INDIRECT_REF inside of it. copy_tree_body_r notices this and cancels the > two, but unlike e.g. build_fold_addr_expr_with_type_loc > if (TREE_CODE (t) == INDIRECT_REF) > { > t = TREE_OPERAND (t, 0); > > if (TREE_TYPE (t) != ptrtype) > t = build1_loc (loc, NOP_EXPR, ptrtype, t); <<<<<<---------- this > } > it doesn't attempt to handle the case where ADDR_EXPR contained different > type from the type of INDIRECT_REF's operand. > > Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux, > bootstrap/regtest on s390{,x}-linux pending. Ok for trunk if it passes > even there? > > 2016-03-04 Jakub Jelinek <jakub@redhat.com> > > PR c++/70084 > * tree-inline.c (copy_tree_body_r): When cancelling ADDR_EXPR > of INDIRECT_REF and ADDR_EXPR changed type, fold_convert it > to the right type. > > * g++.dg/expr/stdarg3.C: New test. OK. jeff
--- gcc/tree-inline.c.jj 2016-02-12 00:50:55.000000000 +0100 +++ gcc/tree-inline.c 2016-03-04 18:35:28.079458603 +0100 @@ -1266,7 +1266,12 @@ copy_tree_body_r (tree *tp, int *walk_su /* Handle the case where we substituted an INDIRECT_REF into the operand of the ADDR_EXPR. */ if (TREE_CODE (TREE_OPERAND (*tp, 0)) == INDIRECT_REF) - *tp = TREE_OPERAND (TREE_OPERAND (*tp, 0), 0); + { + tree t = TREE_OPERAND (TREE_OPERAND (*tp, 0), 0); + if (TREE_TYPE (t) != TREE_TYPE (*tp)) + t = fold_convert (remap_type (TREE_TYPE (*tp), id), t); + *tp = t; + } else recompute_tree_invariant_for_addr_expr (*tp); --- gcc/testsuite/g++.dg/expr/stdarg3.C.jj 2016-03-04 18:08:22.029669394 +0100 +++ gcc/testsuite/g++.dg/expr/stdarg3.C 2016-03-04 18:06:50.000000000 +0100 @@ -0,0 +1,18 @@ +// PR c++/70084 +// { dg-do compile } + +#include <stdarg.h> + +struct A +{ + A (const char *f, ...); +}; + +A::A (const char *f, ...) +{ + va_list ap; + va_start (ap, f); + int i = va_arg (ap, int); // { dg-bogus "first argument to 'va_arg' not of type 'va_list'" } + int j = va_arg ((ap), int); // { dg-bogus "first argument to 'va_arg' not of type 'va_list'" } + va_end (ap); +}
Hi! As mentioned in the PR, on the following testcase we emit invalid error on s390* in the va_arg ((ap), int) case, va_arg (ap, int) is fine (and the former is fine in -std=c++11 or -std=c++98 too). The bug only triggers in constructors (or destructors), and happens because build_va_arg creates ADDR_EXPR that has slightly different type, and has INDIRECT_REF inside of it. copy_tree_body_r notices this and cancels the two, but unlike e.g. build_fold_addr_expr_with_type_loc if (TREE_CODE (t) == INDIRECT_REF) { t = TREE_OPERAND (t, 0); if (TREE_TYPE (t) != ptrtype) t = build1_loc (loc, NOP_EXPR, ptrtype, t); <<<<<<---------- this } it doesn't attempt to handle the case where ADDR_EXPR contained different type from the type of INDIRECT_REF's operand. Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux, bootstrap/regtest on s390{,x}-linux pending. Ok for trunk if it passes even there? 2016-03-04 Jakub Jelinek <jakub@redhat.com> PR c++/70084 * tree-inline.c (copy_tree_body_r): When cancelling ADDR_EXPR of INDIRECT_REF and ADDR_EXPR changed type, fold_convert it to the right type. * g++.dg/expr/stdarg3.C: New test. Jakub