Stackage: Revision of semigroupoids-5.0.0.4 broke lts-3.x

[unclechu]

Revision have been made 6 January:
https://hackage.haskell.org/package/semigroupoids-5.0.0.4/revisions/

Even latest lts-3.22 have tagged-0.8.2 but minimal acceptable version in new revision is 0.8.5, so, build that depends on lts-3.12 fails with this error:

Error: While constructing the build plan, the following exceptions were encountered:
In the dependencies for semigroupoids-5.0.0.4:
    tagged-0.8.1 from stack configuration does not match >=0.8.5 && <1 (latest matching version is 0.8.5)
needed due to %OUR_PACKAGE_NAME%-1.0 -> semigroupoids-5.0.0.4
Some potential ways to resolve this:
  * Recommended action: try adding the following to your extra-deps in /foo/bar/baz/stack.yaml:
- tagged-0.8.5
  * Set 'allow-newer: true' to ignore all version constraints and build anyway.
  * You may also want to try using the 'stack solver' command.
Plan construction failed.

Is there any solution to use original revisions to prevent this mess?

[unclechu]

@RyanGlScott Maybe you could provide some more information?

[dzhus]

In the grand scheme of things, is this a Stackage issue («due to revisions, snapshots are not really snapshots») or a Hackage issue («tightening of lower bounds is allowed and one can do something other than https://github.com/haskell-infra/hackage-trustees/blob/master/cookbook.md#best-practice-for-managing-meta-data »)?

[kadoban]

@dzhus:

https://github.com/haskell-infra/hackage-trustees/blob/master/revisions-information.md seems to suggest that you can do whatever you want to bounds on dependencies via revisions.

So, assuming revisions exist and they aren't going anywhere (which seems like a good assumption), it must be something that has to be dealt with by stackage or by stack. Any package author being able to break any old snapshot at any time doesn't sound good.

p.s. I'm assuming this bug report is accurate, it looks like it is but I didn't actually try it myself yet.

[gbaz]

relevant: https://github.com/ekmett/semigroupoids/issues/62

[kadoban]

@unclechu: I can't seem to replicate this. Can you give more information about what you did to get that error? Also what is your version of stack? I'm on v1.6.3

I tried creating a new stack project, setting the resolver to lts-3.22 and adding the semigroupoids package as a dependency. It appears to build fine with no warning or error.

It's possible I screwed something up, but I'm not seeing what.

[mgsloan]

Any lts-6 and onward, and nightlies roughly around 2016-05-25 and onward all specify hackage revisions, so that they are not affected by these mutations to cabal metadata.

@kadoban I can reproduce it by running stack build --resolver lts-3.22 semigroupoids.

We should consider a stack modification that makes older snapshots more reliable. Perhaps by matching revisions by the latest revision at the time of the snapshot release. Hopefully revisions come with date info.

[kadoban]

@mgsloan That's odd, I tried that too and couldn't reproduce with that command either.

[mgsloan]

Huh! You might already have that semigroupoids built, or some other form of caching. Curious!

[unclechu]

@kadoban Maybe something is cached? I also couldn't reproduce it in CI builds until I remove cache of .stack.

[kadoban]

Oh, that's interesting. Maybe I already had 'tagged' built or something? I noticed it did actually build semigroupoids when I tried, so I figured that was enough, but I didn't consider any caching beyond that really. I'll try later with a clean .stack and etc.

[jberryman]

I think my experience meshes with @unclechu . We ran into this same issue (with semigroupoids revision bump), which only manifested on CI. Could not repro locally after a stack clean --full && stack update. We only fixed the dependency and made sure the build worked on CI, but didn't try to understand the issue further.

[unclechu]

@jberryman stack clean --full removes just .stack-work of a current project, but look also for ~/.stack. You could try to build with overwritten $HOME, like this:

$ env HOME=/tmp/foobar stack build

[unclechu]

Related reddit thread: https://www.reddit.com/r/haskell/comments/7pb5tl/til_stackage_can_be_broken_retroactively_due_to/

[unclechu]

I found a temporary solution, I just added explicit revision of semigroups package to extra-deps: section of stack.yaml: - semigroupoids-5.0.0.4@rev:0 (but also realized that explicit revision isn't working for stack 1.4.0 but working with 1.6.3, not sure in which stack version this feature was invented).

[unclechu]

Thanks to @snoyberg for the hint in his PR so I can solve it right now.