On Thu, 16 Apr 2026 at 22:58, Andrew Ball <[email protected]> wrote:
> Hello Michael, > > CM> Given that 11RC3 is now being tested, I wondered if it made > > sense for The NetBSD Foundation to join Project Glasswing, > > for the express purpose of submitting the codebase to LLM > > security audit, before officially releasing 11.0 ? > > I can't speak for the Foundation or NetBSD developers but as a > fellow user, I wonder whether there is a danger of LLM-based testing > distracting developers who already have plenty of code to review and > human-curated PRs to look at. > You mean, the human-curated PRs where randos manually submit LLM slop for review anyways? I think a more official LLM review would be far better. This wouldn't be the first, either: * In the old day, we all ran Static Analysis on the source code; I've personally found and verified multiple bugs with the help of LLVM/Clang back in the day, before the wider adoption. * Coverity used to offer this as a service to some OSS projects as well. C.
