I’m getting close on a portable-bytecode backend (where the tricky part is making it work across native bit widths and endianesses). I think will be useful for Racket source distributions and maybe Git checkouts, but I no longer think it’s relevant to CI.

Sam and I chatted after a meeting and arrived at this idea for CI: • move the cs-bootstrap package from the racket repo to the ChezScheme repo • keep cs-bootstrap working with a slowly upgraded Racket The general idea is that cs-bootstrap shouldn’t need the latest Racket, but it changes sometimes to reflect Chez Scheme implementation changes. So, CI could use a stashed, infrequently updated Racket to build Chez Scheme on some relatively fast platform, and the right cs-bootstrap is alongside the Chez Scheme code that needs it. Then, as part of the CI pipeline, that Chez Scheme build can be used to create bootfiles for other platforms, so the relatively slow cs-bootstrap process only needs to happen once.

That doesn’t speed up BC builds where they’re slow (like AArch64), but it’s a step in removing the BC build from the CS build sequence.

Thanks for the update. I have been under the weather since Monday evening and took a sick leave. Things are piling up but I will get back to it tomorrow hopefully.