https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58159
--- Comment #6 from Geoff Romer gromer at google dot com ---
A Chromium maintainer privately pointed out a use case that would be thwarted
by a check like this: basically, unique_ptr is used to hold pointers from a
legacy API, using a custom
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58159
Jonathan Wakely redi at gcc dot gnu.org changed:
What|Removed |Added
Status|UNCONFIRMED |NEW
Last
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58159
Geoff Romer gromer at google dot com changed:
What|Removed |Added
CC||gromer at google
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58159
--- Comment #1 from Jonathan Wakely redi at gcc dot gnu.org ---
What if the deleter doesn't actually destroy the object, and doing self-reset
is used as a crazy way to trigger the deleter to do something with the pointer,
but not to alter the
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58159
--- Comment #2 from Jonathan Wakely redi at gcc dot gnu.org ---
I'm also a little concerned that doing a self-reset followed by release() is
indeed valid ... but probably rare enough that we can still assert anyway at
the time of the self-reset.
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58159
--- Comment #3 from Geoff Romer gromer at google dot com ---
What's the standard of review here? If we can only assert on undefined
behavior, even in debug mode, then this just can't be done (although maybe we
should make this undefined in the
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58159
--- Comment #4 from Jonathan Wakely redi at gcc dot gnu.org ---
I think all existing Debug Mode checks only trigger for genuine undefined
behaviour
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58159
Paul Pluzhnikov ppluzhnikov at google dot com changed:
What|Removed |Added
CC|
