Matthew Flatt

I wonder how much difference this makes. If the check to dispatch to `guard-large-allocation` is inlined, I'd expect the safe `make-vector` to be about as fast as an unsafe version....

Ah, I see. At the Racket level, I think adding a new vector function as exported by `racket/vector` may be the right choice, so a suitable length check can be...

> Would vector-extend be implemented in rumble directly, or atop unsafe-make-vector? > I added a commit implementing it on top of unsafe-make-vector. My intent was to avoid having `unsafe-make-vector`, and...

This looks like a good direction overall. Just tests and docs to go?

@samth — I added a small commit, which you can squash after vetting.

LGTM

This in consultation with Robby and meant to address #4922, but `treelist/c` here is changed only to use the new API.

@LiberalArtist — Thanks for your analysis! I've pushed several changes in the latest attempt: * The current state can be extracted from a chaperoned treelist using a key that is...

I'm planning to merge this. It's not clear that this set of changes is right, yet, so I wouldn't say that these changes are the last word on treelist chaperones....

Thanks for the report! It appears that the build and the "README.txt" have gotten out of sync as we evolved the way that `--enable-scheme` and `--enable-racket` are used. The valid...