Add regression test for #5782, which requires extending test harness to v2-install

[Mikolaj]

This adds a regression test for #5782.

I've extended cabal-testsuite harness to exes taken from v2-install (it assumed in-place install, as in v2-build, so AFAICT no amount of extra flags would help). And there is a separate issue that for cabal <= 3.2 my test requires a dummy --installdir option, but I hope we don't run regtression tests with old cabals? Otherwise it would fail anyway, because the test fails in cabal 3.2, being a regression test for an issue fixed (for good) in 3.4

---

Please 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.

Please also shortly describe how you tested your change. Bonus points for added tests!

[Mikolaj]

@fgaz: I bet I missed some thing one needs to do with a new test, like disable it for too old cabals, etc. Any hints appreciated, especially as checks fail...

[Mikolaj]

This is now ready for review.

The only failing CI test is continuous-integration/appveyor/pr where I'm consistently getting 'heap exhauted' failures, e.g., https://ci.appveyor.com/project/23Skidoo/cabal/builds/39255946 but the PR can't be causing that, so it's unrelated.

BTW, I don't have admin access to the AppVeyor CI and so can't, e.g., restart it; I wonder if it should be somehow granted based on github roles.

BTW2, the 'Checks' menu of github doesn't know about AppVeyor --- are any other tests not present there? I guess this can't be easily fixed (no clue about github CI and UI).

[Mikolaj]

@jneira: would you add a quick review, given that this affair is based in part on your findings?

[jneira]

Hi, thanks for the mention (and the invitation! 😄 )
The test looke totally reasonable, to have a case of a diamond dependency seems to be good even if no there was not real bug before.

[jneira]

lgtm, thanks for take care of the possible regression

[Mikolaj]

@jneira: thank you for the review and your reports and troubleshooting.

[emilypi]

LGTM

[Mikolaj]

I've rebased and the CI failure about heap exhausted vanished, which proves it was caused by the state of master branch at that time (rummaging in CI history seems to confirm that). Right now I'm getting MacOS failures, in turn, but surely not my fault either --- probably an outage of wherever we are taking the MacOS archives from.

@fgaz, @phadej: I will merge as soon as you say 'go' or I get bored waiting. :)

[emilypi]

@Mikolaj go

[Mikolaj]

I go.