Alexander Bluhm <[email protected]> wrote: > On Sun, Jun 23, 2019 at 04:18:29PM +0100, David CARLIER wrote: > > Here the changeling > > https://botan.randombit.net/news.html#version-2-10-0-2019-03-30 > > The problem with this update is that they added pledge(2) in a wrong > way. So I hesitated to update the port. > > They ignore pledge "error" instead of dying, so they do not see the > bugs. > > They use mlock()/munlock() which is not allowed with pledge. In a > TLS server error condition they call munlock() which is denied. > > This has not much impact as pledge(2) is only used very little in > their command line tool and not in the library. But this has to > be discussed with upstream. I just did not find time to do it. > > > which underwent a certain amount of internal code rewrite, justifying > > the shared libs bump. > > This is C++, there the ABI is almost never stable. I bump it every > time.
I suggest deleting incorrect pledges when they are spotted with #if 0. Eventually they'll notice it is useless, and delete them or fix them. It is true, "error" was added for a different use case. And I hesitated, since I knew it would be abused.
