David Allsopp

From dev meeting this morning: there's a slightly different slant on the same idea, which should work! Rather than atomic update at the end - which we can't do easily...

It's not frequently a problem, but it comes up with some I/O functions (e.g. the Registry) - in this case passing a UCS-2 buffer with a length. In this case,...

> @dra27 i'm really curious how we both ran into this days apart :) Spooky! I'm was (finally) updating opam to support Windows Unicode properly (the C stubs I contributed...

There's an implementation of the first two in one of my relocatable branches (https://github.com/dra27/ocaml/commit/8fe97fc299c9eadf1baf0316e5182d7beb8cbc9f) - it can either end up going in with that or can be extracted to go...

I was apparently a bit lazy when I wrote the original function (I guess I was writing a lot of C to get the port working!). It would be better...

I don't think we need to care about packages in the invariant which don't have the `compiler` flag any more - that made a certain amount of sense in 2.0...

This definitely needs to be fixed, but we'll address this for opam 2.2 as part of a wider check of the checksums!

Just for belt-and-braces, as it's a build system change, this is running through [precheck#700](https://ci.inria.fr/ocaml/job/precheck/700/).

Sorry, I rushed that one somewhat at the weekend, as @sadiqj had been bitten by it! The only CI worker which matters for this is the Ubuntu worker on GitHub...

> My main development machines are still running Ubuntu 20.04 LTS and have Autoconf 2.69. Does this mean I'll be in trouble until the upgrade to 22.04 LTS is offered...