mirror of
https://github.com/gcc-mirror/gcc.git
synced 2024-11-21 13:40:47 +00:00
80e5be0c7f
I've noticed alloc_align attribute is missing on the non-vector
::operator new with std::align_val_t and const std::nothrow_t&
arguments, this patch adds it. The last hunk is just
an attempt to make the line shorter.
The first hunk originally added also __alloc_size__ (1) attribute,
but seems that regresses
FAIL: g++.dg/tm/pr46270.C -std=gnu++98 (test for excess errors)
with
Excess errors:
.../libstdc++-v3/libsupc++/new:137:26: warning: new declaration 'void* operator new(std::size_t)' ambiguates built-in declaration 'void* operator new(long unsigned int)
+transaction_safe' [-Wbuiltin-declaration-mismatch]
.../libstdc++-v3/libsupc++/new:140:26: warning: new declaration 'void* operator new [](std::size_t)' ambiguates built-in declaration 'void* operator new [](long unsigned int)
+transaction_safe' [-Wbuiltin-declaration-mismatch]
I must say I have no clue why that happens only in C++98 (C++11 and
above are quiet) and why only with -fgnu-tm, tried to debug that but
am lost. It is some conflict with the predeclared ::operator new, but
those clearly do have the externally_visible attribute, and alloc_size (1)
attributes:
extvisattr = build_tree_list (get_identifier ("externally_visible"),
NULL_TREE);
newattrs = tree_cons (get_identifier ("alloc_size"),
build_tree_list (NULL_TREE, integer_one_node),
extvisattr);
newtype = cp_build_type_attribute_variant (ptr_ftype_sizetype, newattrs);
newtype = build_exception_variant (newtype, new_eh_spec);
...
tree opnew = push_cp_library_fn (NEW_EXPR, newtype, 0);
DECL_IS_MALLOC (opnew) = 1;
DECL_SET_IS_OPERATOR_NEW (opnew, true);
DECL_IS_REPLACEABLE_OPERATOR (opnew) = 1;
and at C++98 I think libstdc++ doesn't add transaction_safe attribute:
// Conditionally enable annotations for the Transactional Memory TS on C++11.
// Most of the following conditions are due to limitations in the current
// implementation.
#if __cplusplus >= 201103L && _GLIBCXX_USE_CXX11_ABI \
&& _GLIBCXX_USE_DUAL_ABI && __cpp_transactional_memory >= 201500L \
&& !_GLIBCXX_FULLY_DYNAMIC_STRING && _GLIBCXX_USE_WEAK_REF \
&& _GLIBCXX_USE_ALLOCATOR_NEW
#define _GLIBCXX_TXN_SAFE transaction_safe
#define _GLIBCXX_TXN_SAFE_DYN transaction_safe_dynamic
#else
#define _GLIBCXX_TXN_SAFE
#define _GLIBCXX_TXN_SAFE_DYN
#endif
push_cp_library_fn adds transaction_safe attribute whenever -fgnu-tm
is used, regardless of the other conditionals:
if (flag_tm)
apply_tm_attr (fn, get_identifier ("transaction_safe"));
Anyway, omitting alloc_size (1) fixes that test and given that the
predeclared operator new already has alloc_size (1) attribute, I think it
can be safely left out.
2024-11-08 Jakub Jelinek <jakub@redhat.com>
* libsupc++/new (::operator new, ::operator new[]): Add malloc
attribute where missing. Add alloc_align attribute when
std::align_val_t is present and where it was missing. Formatting fix.
|
||
|---|---|---|
| .. | ||
| config | ||
| doc | ||
| include | ||
| libsupc++ | ||
| po | ||
| python | ||
| scripts | ||
| src | ||
| testsuite | ||
| .editorconfig | ||
| acinclude.m4 | ||
| aclocal.m4 | ||
| ChangeLog | ||
| ChangeLog-1998 | ||
| ChangeLog-1999 | ||
| ChangeLog-2000 | ||
| ChangeLog-2001 | ||
| ChangeLog-2002 | ||
| ChangeLog-2003 | ||
| ChangeLog-2004 | ||
| ChangeLog-2005 | ||
| ChangeLog-2006 | ||
| ChangeLog-2007 | ||
| ChangeLog-2008 | ||
| ChangeLog-2009 | ||
| ChangeLog-2010 | ||
| ChangeLog-2011 | ||
| ChangeLog-2012 | ||
| ChangeLog-2013 | ||
| ChangeLog-2014 | ||
| ChangeLog-2015 | ||
| ChangeLog-2016 | ||
| ChangeLog-2017 | ||
| ChangeLog-2018 | ||
| ChangeLog-2019 | ||
| ChangeLog-2020 | ||
| ChangeLog-2021 | ||
| ChangeLog-2022 | ||
| ChangeLog-2023 | ||
| config.h.in | ||
| configure | ||
| configure.ac | ||
| configure.host | ||
| crossconfig.m4 | ||
| fragment.am | ||
| linkage.m4 | ||
| Makefile.am | ||
| Makefile.in | ||
| README | ||
file: libstdc++-v3/README New users may wish to point their web browsers to the file index.html in the 'doc/html' subdirectory. It contains brief building instructions and notes on how to configure the library in interesting ways.