> So, the only downside seems to be performance related. > > See: https://bugzilla.mozilla.org/show_bug.cgi?id=1215479 > > <<iOS requires it (because it doesn't allow allocating RWX memory), but it's > not > enabled elsewhere because it's a performance hit (we should remeasure how much > though). This is because we have to toggle pages from executable to writable > and back whenever we patch JIT code>> > > What about enabling it in our port and see how it will end up? > We can always re-disable it at any time if people complain...
We could block OpenBSD from allocating RWX memory tomorrow. The blocker against us doing this, is a very small set of programs like firefox. This circular reference is really bullshit. We need to absolutely move forward, get this tested, give them feedback. If we can accept the cost etc, and provide feedback upstream, which results in those 'performance' issues being resolved, then maybe everyone can run safer software later down the road. We (as a userbase) are exactly the type of people to help them with this.
