Fix index-state falling back warning

[hamishmack]

I think it should only be a warning here if ts0 > isiHeadTime isi.

My understanding is that isiMaxTime isi <= isiHeadTime isi (is that correct?) and that since isiMaxTime isi is calculated on the filtered index isiMaxTime isi <= ts0.

[jneira]

Makes sense, thanks for the fix. A regression test to ensure the correct behaviour and a changlog entry would be needed imo

[Martinsos]

Thanks for adding me as a reviewer -> I do have to admit that at this point I don't understand enough context here to be able to provide assessment of the change, but it does help with learning more about how Cabal works!

I don't quite comprehend difference between head and max time -> so max time is the oldest timestamp among the timestamps of all the filtered cache packages, where filtered cache packages are all the packages from the local package cache that match the versions from the specified index state (or are just not older than it)? And head time is the time of the index state itself? OK, so that is why head time is always >= max time, that makes sense.

And then in this piece of code that is modified in this PR, we want to return package source based on specified index state (which is either a timestamp or HEAD, where HEAD means latest)? I am confused with the check of when (isiMaxTime isi /= ts0) -> I thought isiMaxTime is oldest timestamp from filtered cache packages (filtered related to specific index state timestamp), so what does this check even mean, why would we care about /= relationship here? But I must have gotten smth wrong, probably the meaning of maxTime.

[andreabedini]

I think this is part of the problems I am fixing in #7934. In my patch I changed the code around a bit and that comparison got removed.

I take this back. I was mistaken.