Maxime Chevalier-Boisvert

> Multiple calls could return to the same block. A possiblity would be to always store the latest call instruction in the RetEntry, but I am sure that one could...

@krypt-n I made it so `RetEntry` objects are unique to a given call instruction, and the `RetEntry` struct has a `callInstr` field: ad68e9909c70b7355ca04ffddbac4efec46a15f3

Plush asserts now use exceptions. I removed the `abort` opcode: https://github.com/zetavm/zetavm/commit/eac17b2f78e23e9b93eda91e77fb7bc17181b885

Hmm. I suppose we could. The minor complication here is that Zeta exception objects have to be converted into `RunError` objects when they are passing through the VM code. Then...

Hi @krypt-n. Welcome back. That seems like good progress. More than halfway there. Is there anything you're stuck on, need help with?

> Nothing in particular. Is hostCall the only place that needs to convert from RunError to zeta exception? Yes, I believe so. > And do we need an explicit conversion...

> top-level is just the name that plush codegen gives the top-level function. The exception handling doesn't differentiate between it and normal functions Seems good to me. > In the...

Three thoughts: 1. This would also be fixed by a mechanism to throw away executable code when we hit the limit. 2. We could technically monitor how many times a...

What you wrote makes sense. The main question I have is: can we count on the `__bp__` field in `rb_control_frame_struct` always being present? Does MJIT currently make use of this?...

So we'd keep `sp_offset`, compute the BP when entering JITted code, update the SP when leaving or doing a fallback, update BP on method calls. I'm not sure what the...