https://gcc.gnu.org/bugzilla/show_bug.cgi?id=23888
Jonathan Wakely <redi at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |RESOLVED
Resolution|--- |WONTFIX
--- Comment #6 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to Sylvain Pion from comment #5)
> What about providing both modes, with for example a global run-time flag to
> select the mode which is the most appropriate ? And maybe defaulting to
> assert, since it looks like users might prefer this (?).
That just seems like unnecessary complexity.
I'm going to close this. I don't find the benefits of this change compelling.
> Also, a side note : throwing an uncaught exception does not seem to destroy
> the stack,
It does if it is caught and rethrown:
void f() { throw 1; }
int main() {
try {
f();
} catch (...) {
throw;
}
}
The context of the original throw is lost.
Also, the prevalence of 'noexcept' in the std::lib in modern C++ makes it more
likely that an exception will terminate, so we'd just abort() again anyway.
