Add a format rejections function

[philderbeast]

Fixes #9559. Helps with #4251.

Template Α: This PR modifies cabal behaviour

Include the following checklist in your PR:

• Patches conform to the coding conventions.
• Any changes that could be relevant to users have been recorded in the changelog.
• The documentation has been updated, if necessary.
• Manual QA notes have been included.
• Tests have been added.

To test, introduce a version bound (or conflict) that is going to reject some versions of a package. Even if this creates an unsolvable combination of versions, that should suffice to trigger the rejection message by the solver for manual testing.

I found no examples of solver rejection messages in the docs.

[philderbeast]

@yvan-sraka I would like to get this in before some formatting changes for showing "constraint from project". This is a subset of your #9541 changes.

[philderbeast]

@Mikolaj I had to rerun failed jobs three times to get all green ticks:

Failures were:

• Validate ubuntu-latest ghc-8.4.4
  PackageTests/NewUpdate/RejectFutureIndexStates/cabal.test.hs

  -----BEGIN CABAL OUTPUT-----
  Error: cabal: '/usr/bin/curl' exited with an error:
  curl: (7) Failed to connect to localhost port 8000 after 0 ms: Connection refused
  -----END CABAL OUTPUT-----
  

• Validate ubuntu-latest ghc-8.10.7
  PackageTests/NewUpdate/RejectFutureIndexStates/cabal.test.hs

  -----BEGIN CABAL OUTPUT-----
  Error: cabal: '/usr/bin/curl' exited with an error:
  curl: (7) Failed to connect to localhost port 8000 after 0 ms: Connection refused
  -----END CABAL OUTPUT-----
  

• Validate macos-latest ghc-9.4.8
  PackageTests/NewUpdate/UpdateIndexState/update-index-state.test.hs

  -----BEGIN CABAL OUTPUT-----
  Error: cabal: '/usr/local/opt/curl/bin/curl' exited with an error:
  curl: (7) Failed to connect to localhost port 8000 after 0 ms: Couldn't connect to server
  -----END CABAL OUTPUT-----
  

[ulysses4ever]

@philderbeast thanks for the report. The work to prevent this bug is under way in #9540 If you have energy, your help there would be greatly appreciated because I'm running out of ideas.

[philderbeast]

Label merge+no rebase is necessary when the pull request is from an organisation.

[ulysses4ever]

Could we get @grayjay's opinion on it but any chance? She's our solver code owner effectively. Would be great to get her approval on any change to it.

[sebright]

@philderbeast Thanks for working on the solver error messages! Please add a note about starting with @yvan-sraka's change to the commit message.

[philderbeast]

Please add a note about starting with @yvan-sraka's change to the commit message.

@yvan-sraka I did not start from your change and sorry I didn't see your work on #4251 earlier. I was working on #9510 and the verbosity of the rejection messages started to annoy me so I raised #9559, saw a quick fix improvement and went for it, found your work and realized I hadn't accounted for hyphenated package names.

The commit history is lost after squashing but these were the 3 initial commit messages:

• Format solver rejections
• Take account of hyphenated package names
• Don't shorten a singe rejection

[sebright]

@yvan-sraka I did not start from your change and sorry I didn't see your work on #4251 earlier. I was working on #9510 and the verbosity of the rejection messages started to annoy me so I raised #9559, saw a quick fix improvement and went for it, found your work and realized I hadn't accounted for hyphenated package names.

@philderbeast Sorry, I had misunderstood your comment above. I see that the changes are independent.

[philderbeast]

I've left recent commits as-is for review. Please let me know if you'd like some of these squashed.

[sebright]

@philderbeast Thanks for the updates and for also handling skipped versions.

I've left recent commits as-is for review. Please let me know if you'd like some of these squashed.

I would prefer squashing to remove the large test build files before you merge.

[philderbeast]

@ulysses4ever this is ready for your approval. It already has the merge delay passed label attached.

[ulysses4ever]

Perfect, thanks!

[philderbeast]

Label merge+no rebase is necessary when the pull request is from an organisation.

[ulysses4ever]

Please, try not to forget squashing before merging.

[philderbeast]

@ulysses4ever a momentary lapse of diligence on my part, sorry about that.