| Message ID | 554AD940.3000800@verizon.net |
|---|---|
| State | New |
| Headers | show |
On Wed, May 06, 2015 at 11:17:20PM -0400, Ed Smith-Rowland wrote: > In addition to a PR this is 1/2 of a C=+17 feature. (The other half - really > a separate thing - is attributes on namespaces). Ah, nice, I wasn't aware. For the record, this is <http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n4266.html>. > I wonder if we should pedwarn for < C++17? > Or it could be just an extension for < C++17 - I guess that would match with > clang. Yeah, it is meant as a GNU extension. (clang supports this extension for several years already.) I'd rather let Jason decide what to do wrt C++17. > @@ -3651,11 +3651,6 @@ finish_id_expression (tree id_expression, > } > } > > - /* Handle references (c++/56130). */ > - tree t = REFERENCE_REF_P (decl) ? TREE_OPERAND (decl, 0) : decl; > - if (TREE_DEPRECATED (t)) > - warn_deprecated_use (t, NULL_TREE); > - > return decl; > } > > Why did this bit get removed? This hunk got added in r201906 to address c++/56130 - we didn't warn for deprecated references: int g_nn; int& g_n __attribute__((deprecated)) = g_nn; int main() { g_n = 1; } But then Jason added warn_deprecated_use to mark_used in r217677 and we warned twice. So I figured the warning in finish_id_expression isn't needed anymore. > Do we handle enums in template specializations? Not sure, could you provide a testcase? Thanks, Marek
On 05/07/2015 09:59 AM, Marek Polacek wrote: > On Wed, May 06, 2015 at 11:17:20PM -0400, Ed Smith-Rowland wrote: >> In addition to a PR this is 1/2 of a C=+17 feature. (The other half - really >> a separate thing - is attributes on namespaces). > Ah, nice, I wasn't aware. For the record, this is > <http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n4266.html>. > >> I wonder if we should pedwarn for < C++17? >> Or it could be just an extension for < C++17 - I guess that would match with >> clang. > > Yeah, it is meant as a GNU extension. (clang supports this extension for > several years already.) I'd rather let Jason decide what to do wrt C++17. > >> @@ -3651,11 +3651,6 @@ finish_id_expression (tree id_expression, >> } >> } >> >> - /* Handle references (c++/56130). */ >> - tree t = REFERENCE_REF_P (decl) ? TREE_OPERAND (decl, 0) : decl; >> - if (TREE_DEPRECATED (t)) >> - warn_deprecated_use (t, NULL_TREE); >> - >> return decl; >> } >> >> Why did this bit get removed? > This hunk got added in r201906 to address c++/56130 - we didn't warn for > deprecated references: > > int g_nn; > int& g_n __attribute__((deprecated)) = g_nn; > > int main() > { > g_n = 1; > } > > But then Jason added warn_deprecated_use to mark_used in r217677 and we > warned twice. So I figured the warning in finish_id_expression isn't > needed anymore. > >> Do we handle enums in template specializations? > Not sure, could you provide a testcase? Thanks, > > Marek > Instead of NULL_TREE in pt.c I grabbed the attrs. /* Actually build the enumerator itself. */ build_enumerator (DECL_NAME (decl), value, newtag, DECL_ATTRIBUTES (decl), DECL_SOURCE_LOCATION (decl)); Seems to work. Also, I haven't tested the testcase in terms of the pattern matching of the error. Tweak it if necessary. Thanks for this. Ed
diff --git gcc/cp/semantics.c gcc/cp/semantics.c index 701a8eb..b46c6fc 100644 --- gcc/cp/semantics.c +++ gcc/cp/semantics.c @@ -3651,11 +3651,6 @@ finish_id_expression (tree id_expression, } } - /* Handle references (c++/56130). */ - tree t = REFERENCE_REF_P (decl) ? TREE_OPERAND (decl, 0) : decl; - if (TREE_DEPRECATED (t)) - warn_deprecated_use (t, NULL_TREE); - return decl; }
In addition to a PR this is 1/2 of a C=+17 feature. (The other half - really a separate thing - is attributes on namespaces). I wonder if we should pedwarn for < C++17? Or it could be just an extension for < C++17 - I guess that would match with clang. if (SCOPED_ENUM_P (newtag)) Why did this bit get removed? Do we handle enums in template specializations? Ed