https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57111
Martin Sebor <msebor at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Keywords| |patch
Assignee|unassigned at gcc dot gnu.org |msebor at gcc dot
gnu.org
CC| |msebor at gcc dot gnu.org
See Also| |https://gcc.gnu.org/bugzill
| |a/show_bug.cgi?id=90629
--- Comment #11 from Martin Sebor <msebor at gcc dot gnu.org> ---
The patch I submitted for pr90629 implements this enhancement:
https://gcc.gnu.org/pipermail/gcc-patches/2020-November/557987.html
It detects the bug in the test case in comment #0 but only with optimization
(to see through inlined calls) and with -Wsystem-headers. Just like all late
warnings to date, -Wfree-nonheap-object isn't without false positives. pr54202
is one that even the exceedingly simplistic -Wfree-nonheap-object is
susceptible to. The patch above doesn't change things.
In file included from
/build/gcc-trunk/x86_64-pc-linux-gnu/libstdc++-v3/include/memory:76,
from t.C:2:
In member function ‘typename std::enable_if<std::is_convertible<_Up (*)[], _Tp
(*)[]>::value>::type std::default_delete<_Tp []>::operator()(_Up*) const [with
_Up = int; _Tp = int]’,
inlined from ‘std::unique_ptr<_Tp [], _Dp>::~unique_ptr() [with _Tp = int;
_Dp = std::default_delete<int []>]’ at
/build/gcc-trunk/x86_64-pc-linux-gnu/libstdc++-v3/include/bits/unique_ptr.h:612:17,
inlined from ‘int main()’ at t.C:6:32:
/build/gcc-trunk/x86_64-pc-linux-gnu/libstdc++-v3/include/bits/unique_ptr.h:120:11:
warning: ‘void operator delete [](void*)’ called on unallocated object ‘arr’
[-Wfree-nonheap-object]
120 | delete [] __ptr;
| ^~~~~~~~~~~~~~~
t.C: In function ‘int main()’:
t.C:5:7: note: declared here
5 | int arr[]={1,2};
| ^~~
