Lee Spector

I've been a slow convert to `require`, and while I still prefer `use` in many contexts with small projects, I agree that Clojush has grown too big for this. But...

> Whenever the value of `max-number-magnitude` is the default value of `1000000000000`, then the result of `(float_mult 10000000.0 10000000.0)` (or any larger `float` result) will be dropped down to the...

> No, you're looking at the underflow. If the result of `float` multiplication is very large, the result of `keep-number-reasonable` is always an integer. I still don't see it. Here's...

> So (not asking for a history, just proposing) why not simply check values pushed to the stacks? Isn't that exactly what we do? When a numeric instruction is about...

> Alternately, just use exception-handling as it's intended, and capture over- and underflows and do the truncation when it occurs. Perhaps that's a good alternative implementation, but I'm not sure....

> should have a single function with clear and explicit responsibility for making numbers the right size _when they are pushed to the stack_, I thought we were sort of...

> I wish somebody from your side had the resources to actually pair with me. Preferably you, since you are the final gatekeeper, and seem to have the most to...

> started [this project](https://github.com/Vaguery/push-in-closure) Cool. > the code _defines_ the Push `integer` type as being an integer between the bounds set by `max-number-magnitude`... and not existing outside that range. Is...

@thelmuth wrote: "My guess is that something Really Bad happened because we were redeffing an instruction without meaning to." That's my vague memory too. In the current code base there's...

@Vaguery wrote: "Is there any reason that `registered-instructions` isn't just a reference to `(keys instruction-table)`?" There may be different a historical explanation, but the only reason I can think of...