Bump time upper bound to acccomodate 1.14

[bgamari]

Just as it says on the tin.

[geekosaur]

Just did this in xmonad a couple days ago too. Note that, since it's a bootlib, you'll need a --constraint to test it.

[andreasabel]

Yes this is untested, CI passing says nothing here.
@bgamari Have you tested the build locally on your machine, with -c 'time >= 1.14' ?

[geekosaur]

I'm running validate.sh locally with a constraint: in cabal.project.validate. Note that this also requires allow-newer: time because not all of our dependencies have been updated for time 1.14 yet.

ETA: validate passed

[bgamari]

@bgamari Have you tested the build locally on your machine, with -c 'time >= 1.14' ?

Yes, I tested it in my GHC tree.

[geekosaur]

Hm, actually this just updates Cabal and Cabal-syntax whereas I just validated the whole thing. Is a follow-up PR a good idea?

[bgamari]

I only bumped Cabal and Cabal-syntax as this is what I need to bump the time submodule in GHC.

[geekosaur]

That's fine, it's what I figured. But from a Cabal dev POV, I've already done everything necessary to bump cabal-install as well, so I'm asking if I should go for it. After all, we'll have to deal with it eventually anyway.

[andreasabel]

I wish we could get started to test cabal with GHC 9.10 in general, but we will run into hashable not being ready:

• hashable doesn't support filepath-1.5, ghc-9.10 will bump filepath to 1.5 haskell-unordered-containers/hashable#292
• https://gitlab.haskell.org/ghc/head.hackage/-/issues/100

@bgamari Is there something to be done about that? The maintainer does not seem to respond atm. Maybe the GHC team is aware of hashable blocking the road and has a plan...?

[bgamari]

@andreasabel unfortunately I don't have any particular plan here. It would be very nice if we cabal-install provided a finer-grained variant of --allow-newer (e.g. #7896) since it would then at least be more likely that one could work around the issue with more specific constraints.

@phadej, perhaps you can comment here as hashable's maintainer of record?

[phadej]

@bgamari comment on what?

hashable has specifically bitten by releasing / doing compatibility work for unreleased GHC and then redoing it, so I'm no in interest to support first alpha. If head.hackage doesn't work for you, it's not my problem. Isn't head.hackage there so you don't need to bother maintainers for GHC head / alpha testing?

[andreasabel]

@phadej wrote:

hashable has specifically bitten by releasing / doing compatibility work for unreleased GHC and then redoing it, so I'm no in interest to support first alpha. If head.hackage doesn't work for you, it's not my problem. Isn't head.hackage there so you don't need to bother maintainers for GHC head / alpha testing?

Well, if I understand the problem correctly, it is primarily the incompatibility of hashable with filepath-1.5 which causes the problem. filepath-1.5 is already released. Not supporting it yet is just delaying the inevitable. And it makes evaluating filepath-1.5 much harder (so that most won't bother supporting it yet either). Now it also makes evaluating GHC 9.10 much harder.
hashable has 1000 users on Stackage alone, should they all do their own source-package-repository workaround?

[Bodigrim]

hashable has 1000 users on Stackage alone, should they all do their own source-package-repository workaround?

I know two simple workarounds:

• Use head.hackage and force -c 'hashable +os-string'.
• Or don't use head.hackage and force -c 'filepath < 1.5'.

[Mikolaj]

Let me try to verify if the CI failure is a fluke or not.

[Mikolaj]

@mergify rebase

[mergify]

rebase

✅ Branch has been successfully rebased

[Mikolaj]

@bgamari: do I understand right that this should be backported to 3.12? I guess, you already apply this patch to your checkout of Cabal 3.12 in GHC 9.10?

Edit: and, if so, are any other such PRs pending? It would be unfortunate to release 3.12.0.0 without them...

Edit2: maybe this should have a checkbox in #9729?

[Mikolaj]

@mergify backport 3.12

[mergify]

backport 3.12

✅ Backports have been created

• #9869 Bump time upper bound to acccomodate 1.14 (backport #9848) has been created for branch 3.12