Remove redundant OPTIONS_GHC pragmas from templates

[sol]

There are two things that I lack the context and background to comprehend.

1. For what reason would these pragmas be needed in the first place?
2. What was the rational for changing these in f47768a? Was it simply an oversight, or are we at the point where "a subjective urge for tidiness" takes precedences over providing value to the user and backwards compatibility?

• squash commits (I did this through GitHub UI from mobile..)

[sol]

Just to be clear, if this was an honest mistake, then that's OK with me. Nobody's perfect and mistakes do happen. That's normal.

But if this was intentional, then I don't have much sympathy for it.

[Bodigrim]

1. For what reason would these pragmas be needed in the first place?

If a package is compiled with, say, -Werror -Weverything, Paths_*.hs must forcefully disable some warnings so that this autogenerated file does not cause the entire build to fail.

It could probably bluntly set -Wwarn -Wno-everything, but what's exactly the issue with the current approach?

[sol]

1. For what reason would these pragmas be needed in the first place?

If a package is compiled with, say, -Werror -Weverything, Paths_*.hs must forcefully disable some warnings so that this autogenerated file does not cause the entire build to fail.

My naïve assumption was that the -w that's already there is enough to achieve this. But maybe I am mistaken.

what's exactly the issue with the current approach?

It needlessly breaks GHC 7.10.3.

https://github.com/hspec/setup-haskell/actions/runs/19809279822/job/56748489581#step:6:25

Now, if you would get anything in return for making things less compatible than that could be a worthwhile trade-off.

But changing

{-# OPTIONS_GHC -fno-warn-missing-import-lists #-}
{-# OPTIONS_GHC -w #-}

to

{-# OPTIONS_GHC -Wno-missing-import-lists #-}
{-# OPTIONS_GHC -w #-}

?

That's strictly less compatible, no? Is there any justification to do that?

[Bodigrim]

It needlessly breaks GHC 7.10.3.

Cabal stopped testing compatibility with GHC 7.10 quite some time ago (modern Linux distributives can no longer run old GHC binaries and macOS on ARM even less so), so that's why no one noticed it. But sure, -fno-warn instead of -Wno should be fine. In the interest to get things done without a prolonged discussion, I'd suggest to update the PR to do just that.

[sol]

In the interest to get things done without a prolonged discussion, I'd suggest to update the PR to do just that.

I'm not a big fan of cargo culting.

Is there a reason -w alone isn't sufficient?

[Bodigrim]

Dunno, I can see #7352 introducing -w with a stated goal to quash -Wno-prepositive-qualified-module (but still retaining -fno-warn-missing-import-lists!), and then #8336 a year later, which added -Wno-prepositive-qualified-module explicitly.

[sol]

then #8336 a year later, which added -Wno-prepositive-qualified-module explicitly

Even adding a test that doesn't seem to fail without the corresponding change…

[Bodigrim]

Given that the tests still pass and that #7352 / #8336 does not mention any specific reason why -w would be insufficient to be used instead of any other -Wno / -fno-warn, I suppose this is good to go.

[ulysses4ever]

It looks like a good cleanup of a somewhat messy history of drive-by edits... Thanks!

[Bodigrim]

@sol would you like to add "merge me" label?

[Bodigrim]

@sol could you please rebase and push again to make CI fully green?