]> git.ipfire.org Git - thirdparty/gcc.git/commit
c++: Reject pack expansion of assume attribute [PR109756]
authorJakub Jelinek <jakub@redhat.com>
Tue, 9 May 2023 14:05:22 +0000 (16:05 +0200)
committerJakub Jelinek <jakub@redhat.com>
Tue, 9 May 2023 14:07:14 +0000 (16:07 +0200)
commit2499540e9abb55079b5f7b7ccdac97fbc63d9ab4
tree72134a6c1280539667880bc43a5167e0da2a1259
parent204303c81e82ddd01e7dc5a5a63719d476f9043c
c++: Reject pack expansion of assume attribute [PR109756]

http://eel.is/c++draft/dcl.attr#grammar-4 says
"In an attribute-list, an ellipsis may appear only if that attribute's
specification permits it."
and doesn't explicitly permit it on any standard attribute.
The https://wg21.link/p1774r8 paper which introduced assume attribute says
"We could therefore hypothetically permit the assume attribute to directly
support pack expansion:
template <int... args>
void f() {
[[assume(args >= 0)...]];
}
However, we do not propose this. It would require substantial additional work
for a very rare use case. Note that this can instead be expressed with a fold
expression, which is equivalent to the above and works out of the box without
any extra effort:
template <int... args>
void f() {
[[assume(((args >= 0) && ...))]];
}
", but as the testcase shows, GCC 13+ ICEs on assume attribute followed by
... if it contains packs.
The following patch rejects those instead of ICE and for C++17 or later
suggests using fold expressions instead (it doesn't make sense to suggest
it for C++14 and earlier when we'd error on the fold expressions).

2023-05-09  Jakub Jelinek  <jakub@redhat.com>

PR c++/109756
* cp-gimplify.cc (process_stmt_assume_attribute): Diagnose pack
expansion of assume attribute.

* g++.dg/cpp23/attr-assume11.C: New test.
gcc/cp/cp-gimplify.cc
gcc/testsuite/g++.dg/cpp23/attr-assume11.C [new file with mode: 0644]