2edc311a33
Merge pull request 'revise startup, address urls pointing to same sha256 and support mirrors (upstream and in opam file)' (#24) from startup into main
still remaining is an issue that git-kv restore takes ~30 minutes for opam-repository on my laptop... I'd hope we can get a 100x speedup somehow ;)
We keep the url in the map as key, and just carry around a set of other urls for the same artifact.
Not sure whether it is worth to have the url as key in the map anymore, though we discovered…
Indeed, currently it is a List.fold_left of "HM.bindings hm". So yes, this could be nicer code. :)
And please feel free to push the documentation updates directly to the branch.
I think we need a rethink about failure behaviour and (recoverable) errors...
There's not only the swapfs, but then we have mimic errors "no connection found" (which I guess we should retry to…
actually, if I'm not misguided, we first go to the source, and only thereafter to mirror(s). Is this a good semantics?
not doing upstream csums here since we did that for the previous entry already. downside is if there's an artifact that has some hashes in some file, and more hashes in another file, we won't get the other hashes.
this code, using a global hashtable, is ugly and should be revised.