GHC panic when compiling examples

[torgeirsh]

I ran "stack build :misc-examples" and got:
Preprocessing test suite 'misc-examples' for concat-examples-0.3.0.0...
[1 of 1] Compiling Main ( test\Examples.hs, .stack-work\dist\ca59d0ab\build\misc-examples\misc-examples-tmp\Main.o )
ghc.EXE: panic! (the 'impossible' happened)
(GHC version 8.0.2 for x86_64-unknown-mingw32):
Simplifier ticks exhausted
When trying RuleFired cat equal
To increase the limit, use -fsimpl-tick-factor=N (default 100)
If you need to do this, let GHC HQ know, and what factor you needed
To see detailed counts use -ddump-simpl-stats
Total ticks: 51001

Not sure if it's an actual GHC bug, or something to do with the plugin?

[conal]

Oops! I've been experimenting with some new rewrite rules. I see that even the most basic examples are failing before the CCC plugin even gets started. Investigating. Thanks for the report.

[conal]

I had a few changes not yet pushed. Seems to be working again. Try again?

[torgeirsh]

Thanks for looking into it! It's still not quite working for me, I get a different error:
Preprocessing library concat-inline-0.1.0.0...
[3 of 3] Compiling ConCat.Inline.SampleMethods ( src\ConCat\Inline\SampleMethods.hs, .stack-work\dist\ca59d0ab\build\ConCat\Inline\SampleMethods.o )
ghc.EXE: panic! (the 'impossible' happened)
(GHC version 8.0.2 for x86_64-unknown-mingw32):
Static flags have not been initialised!
Please call GHC.parseStaticFlags early enough.

To make sure everything is rebuilt, I deleted the .stack-work directory and the global snapshots directory. Let me know if there's anything else that could be left behind.

[conal]

Wow. I've never seen anything like that error. Have you? I don't know whether anyone else has tried using concat on Windows.

[torgeirsh]

I've never seen anything like it, but I also have never used a compiler plugin before. Maybe it's an issue with the Windows version of GHC.

[conal]

There seem to be other reports of this issue, e.g., here.

[torgeirsh]

Since they also use GHC 8.0, I tried to change the stackage resolver to nightly-2017-10-31 and see if it works better with GHC 8.2, but BuildDictionary.hs doesn't compile. I guess it's too early to update the code as there's no stackage LTS with GHC 8.2 yet.

[torgeirsh]

Now that stackage lts-10.0 is out with GHC 8.2, I updated the resolver and tried to build again. This time I get a different panic:

    [3 of 4] Compiling ConCat.Category  ( src\ConCat\Category.hs, .stack-work\dist\5c8418a7\build\ConCat\Category.o )
    ghc.EXE: panic! (the 'impossible' happened)
      (GHC version 8.2.2 for x86_64-unknown-mingw32):
        inlineClassOp failed
      ($dPointed_apHR
       `cast` (N:Pointed[0] <h_apHQ>_N
               :: (Pointed h_apHQ :: Constraint)
                  ~R#
                  ((forall a. a -> h_apHQ a) :: *)))
        @ a_apI3
      Call stack:
          CallStack (from HasCallStack):
            prettyCurrentCallStack, called at compiler\utils\Outputable.hs:1133:58 in ghc:Outputable
            callStackDoc, called at compiler\utils\Outputable.hs:1137:37 in ghc:Outputable
            pprPanic, called at src\ConCat\Inline\Plugin.hs:66:19 in concat-inline-0.1.0.0-DgOLZX7x9V0EDG7Ujj5GxP:ConCat.Inline.Plugin

[conal]

Thanks. I'm seeing the same error with lts-10.0. The issue seems to be a small change in GHC's treatment of single-method classes (Pointed here). I can accommodate this change easily enough, but then I run into more errors due to GHC API changes, which is often the case with new GHC releases.

It will take some time to sort through these changes, and I've been meaning to overhaul this part of the plugin anyway.