Impossible happened

[iampure@…]

ghc-6.7.20070702: panic! (the 'impossible' happened)

(GHC version 6.7.20070702 for i386-unknown-linux):

nameModule $dMonad{v a1BV5}

Please report this as a GHC bug: http://www.haskell.org/ghc/reportabug

Unfortunately, I have no test case, but it happened during the evaluation of something of type IO ([(Int, Int)]) which uses undecidable instances, functional dependencies and the no monomorphism restriction.

[iampure@…]

It appears that merely evaluating

foo = putStrLn "hi"

triggers the bug, but only when inside a module that uses those extensions in combination with my code.

[guest]

During compilation (ghc --make) the following happened:

ghc-6.7.20070702: panic! (the 'impossible' happened)

(GHC version 6.7.20070702 for i386-unknown-linux):

initC: srt

It might be the case that my compiled version of GHC is simply broken, but since I didn't do anything odd and just followed build instructions, I report it.

This is another issue from the one reported above, but they are probably related.

[simonpj]

Dear iampure, (I wonder what your real name is!)

I'm sorry that GHC crashes on you, but it's extremely rare for to be able to fix a bug without being able to reproduce it. If you want us to look into this, there's not alternative to tar'ing up your files and attaching them to this report, adding repro instructions. Or email us direct if it's secret. Without that there is literally nothing we can do to help you.

Meanwhile, make sure you compile everything with -dcore-lint. That often shows up a bug much earlier than otherwise.

Simon

[iampure@…]

Dear Simon,

On  http://hackage.haskell.org/trac/ghc/ticket/1494 you can find a reproducible <500 bytes test case. #1489 is most likely fixed when #1494 is. Thank you for the -dcore-lint hint. As for my real name, I'd rather not have <insert search engine> turn up lots of results when someone searches for my name. If someone would spend some time finding out my name, they would probably succeed however. If you can send me an email address that you will actually read and respond too, I will give you my name.

[simonpj]

OK I think #1494 is fixed; can you confirm, and close this one?

Simon

[iampure@…]

Confirmed