David Chisnall

Draft release here: https://github.com/gnustep/libobjc2/releases/edit/untagged-debddff89ebe8e9e6160 Feel free to hit the publish button if you're happy with it!

I think this code looks correct, it would be nice to have a test. I’m still not quite sure how code can get to this path though.

I don't think that's an ideal approach, since it will complicate toolchains for little benefit. We were converging on a place where CHERIoT could be a strict addition to RV32YE....

And many of the things that CHERIoT wants (e.g. the reduced AUIPCC shift and a different capability format) are also important for some accelerator use cases. So now they also...

> It uses an entirely different capability format. As will many other things. Some accelerators will not want the current encoding, they will want something that can be packed into...

> I think the real point is that, if CHERIoT and any similar cases were to be deemed extensions to RVxY, that renders RVxY meaningless by itself. You can't generate...

> I guess most of us will be at the summit in Paris in May I'm not sure who the 'us' is there, I don't think any SCI or Microsoft...

> hybrid alters loads/stores in the same way as a PMP This is true if you consider them as applying to an I base, but on a Y base they...

I am quite nervous about building something into the base specification with zero practical experience of how software would use it. Zcherilevels is a research project.

I agree. I'm less convinced that we want `Zstid` in CHERIoT, but the idea of having a thread pointer register is a pretty well-understood thing and the change to make...