new should throw bad_alloc if it fails to allocate memory. If you catch
and rethrow exceptions properly you can do all the standard releasing of
mutexes, deallocating allocated memory, etc, that you would do in kernel C
code when you failed a malloc. I'm assuming that any implementation of C++
in the kernel would already have new/delete and other intrinsic
functionality rewritten in terms of standard kernel resource allocation
mechanisms.
The real problem is that writing good C++ code requires more discipline
than writing good C code (because the language gives you so many new and
exciting ways to shoot yourself in the foot), and putting it in the kernel
is going to add even more constraints that even good C++ programmers may
not see immediately.
There may be some good reasons enough to keep C++ out of the kernel, but
appeals to common practice with things like "using feature X induces bloat"
or "most C++ programs don't handle OOM the way the kernel does" really
don't count for much in these kinds of arguments. Kernel C++ would have
it's own style that differs from C++ just as kernel C looks different to
userland C.
The primary reason I can see to keep C++ out of the kernel is that the user
base doesn't seem to be large enough to get over the initial barrier of
making it work and supporting it yet. If it were, the necessary patches
would already be in doubters hands and none of us would be wasting
bandwidth on the subject.
nate
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/