Robert Collins

the (default) in `rustup toolchain list` makes that harder to work with. I wonder if dropping that default, making the list more API-like, adding 'rustup toolchain default' or something similar,...

TBH the current stalled situation is worse than you publishing a spec based on the discord and google groups discussions.

> Having briefly skimmed over some parts of this: what limits us from just using `#[tokio::main]` and making things natively async, rather than all the callback stuff and `block_on()` type...

Once done the aggregate diff should be precisely the changes needed, and I can rebase squash it all away. We did this with the removal of `failure`. Many many small...

Thats also valid to do; lets assess when this is at the next inflection point (which will be async native with no thunk, but no refactorings for better behaviour)

This is nearly complete, I think its mostly mechanical now, except for the closures in the test suite - there's some lifetime challenges to address there

This now complete modulo any refactoring of the commit series we might want to do. The key notes are: - we can't use tokio::main because it doesn't provide the per-worker-thread...

Using the netrc file would be good, or using the new bazel auth helper would be nice too; however right now it is definitely broken; putting literal passwords in repository...

Its a fairly straightforward patch, just another attribute on the tag, a logical_or over definitions of the same repo, and finally pass it into the existing coursier_fetch rule.

One decent flexible way would be a error transformer to allow setting policy via code. E.g. some errors you might ignore, others you might retry, and still others you'd propogate.