Somewhat off-topic, but if we require a bootstrapping Python interpreter these days, is there any reason why pegen has to be regenerated manually and checked in instead of being generated by make in the checkout itself?

is there any reason why pegen has to be regenerated manually and checked in instead of being generated by make in the checkout itself?

The original reasons were that we checkout the rest of regenerated files and having them in the tree helps reviews because even if is regenerated is very easy to review what's going on if the code is diffed. Additionally, it helps when developing because we can go back to previous versions if we messes something without having to change an in-flight edited grammar. I have also checked the file plenty of times in the GH interface to look for certain behaviours. I suppose we could live without any of these if someone feels strongly about starting to remove auto-generated files.

I have also checked the file plenty of times in the GH interface to look for certain behaviours.

Same here. This has been useful a handful of times.