Stop if cannot attain requested index-state, #8076.

[philderbeast]

As this is a change of behaviour, we will need a changelog entry won't we?

I intend to supplement this by changing "this command" with the actual command name.

The rendered output:

> cabal build --enable-tests
Error: cabal: Stopping this command as the requested
index-state=3022-03-22T14:16:26Z is newer than (2022-03-31T17:42:23Z), the
most recent state of 'hackage.haskell.org'. You could try 'cabal update' to
bring down a later state or request an earlier timestamp for index-state.

[Mikolaj]

Yes, a tiny changelog file and a test would be necessary. Let me also repeat the questions about how breaking change this is and how can be mitigate the breakage, hoping that others would join the discussion:

Anyway, is anybody opposed to this change? Can we think of a workflow it breaks? Should it be an option turned on by a flag somewhere?

Edit: the question was answered in the issue ticket. Some mitigation, though, e.g., an instruction how to get the old behaviour, would be great.

[gbaz]

We discussed this at the cabal advisors meeting and think this makes sense. Just needs a changelog and then we can merge.

[jneira]

After the further issue discussion i am for throwing the error, but i would ask to suggest --index-state=HEAD as the more direct way to override the requested index state and ignore it.
Also maybe we should wait for #8086 or make the message taking in account.

[jneira]

taking in account #8086 maybe we could change it with something like:

Suggested change

	++ "), the most recent state of '" ++ unRepoName rname
	++ "), the local state of '" ++ unRepoName rname

maybe HEAD instead of local? maybe it would be less clear for newcomers

[jneira]

Suggested change

	++ "'. You could try 'cabal update' to bring down a later state or request an earlier timestamp for index-state."
	++ "'. You could try 'cabal update' to bring down a later state or force the use of current local state with `--index-state=HEAD`."

you have to use a index-sate <= HEAD and it seems to me than is more frequent use HEAD than a older one

[Mikolaj]

Thank you very much for the PR. I'm entering this early review to block this on changelog, on some mitigations/warnings about the breaking change (e.g., some ideas from #8076 (comment)) and, ideally, a test.

[ulysses4ever]

@philderbeast are you interested in finishing this off? That’d be super cool!

[philderbeast]

Yes I would like to finish this.

[Kleidukos]

Marking this PR as draft for now 🙂

[andreabedini]

#8944 is a more recent take on the same problem.
@philderbeast I suggest you leave a comment there if you believe that PR is missing anything you had done here.