ghc-6.6: panic! (the 'impossible' happened) mkWWcpr: not a product GHC-Brian-6.5.1:IdInfo.IdInfo{tc rfD}

[briansmith]

I receive this message:

ghc-6.6: panic! (the 'impossible' happened)
  (GHC version 6.6 for i386-unknown-linux):
        mkWWcpr: not a product GHC-Brian-6.5.1:IdInfo.IdInfo{tc rfD}

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

I am using GHC from the 6.6 branch (pulled & built today).

I am building the GHC API using the Cabal file that is attached. To use it, put the Cabal file and Setup.lhs in ghc/compiler/ along and run:

runhaskell Setup.lhs configure --with-hc=/path/to/ghc-6.6/compiler/stage2/ghc-inplace --with-hc-pkg=/path/to/ghc-6.6/utils/ghc-pkg/ghc-pkg-inplace
runhaskell Setup.lhs build

It takes a long time because TidyPgm? is module 215 of 274. When the "-O2" option is removed from the Cabal file, it builds and works fine.

This was on Ubuntu 6.06.

[briansmith]

Cabal file

[briansmith]

Setup.lhs

[briansmith]

If you remove the -O2 to compile TidyPgm?, then it happens again in TcIface?:

iface/TcIface.lhs:63:29:
    Warning: Imported from `Util' but not used: `equalLength'
ghc-6.6: panic! (the 'impossible' happened)
  (GHC version 6.6 for i386-unknown-linux):
        mkWWcpr: not a product GHC-Brian-6.5.1:IdInfo.IdInfo{tc rfD}

[simonpj]

This is an annoying, known, but awkward-to-fix problem concerning mutually-recursive modules. The problem is that when compiling the modules *inside* a loop, a data type that is only defined at the top of the loop looks opaque; but after the loop is done, the structure of the data type becomes apparent.

The difficulty is then that two different bits of code have different notions of what the data type looks like.

So far this bug has bitten only GHC itself and one other person. I need to think carefully about how to fix it. Thanks for reporting it.

Unsatisfactory workaround: compile without -O.

[briansmith]

Another workaround seems to be to use a makefile and "ghc -c" instead of "ghc --make." I compile GHC using its makefile all the time with -O2 without incident.

[simonpj]

Yes, I confirm that's correct. The problem only shows up with --make.

Simon

[simonpj]

See also #1072

[igloo]

This affects relatively few packages, but for those it does, it's a pain when you want to build with Cabal.

[Isaac Dupree]

It would be convenient for GHC to be able to compile itself with --make, to help working/testing on #1405 (possibly involving Cabal...). Alternatively #1409 could work too.

[igloo]

Won't happen in time for 6.8.

[simonmar]

Fixed:

Mon Sep 10 11:38:30 BST 2007  Simon Marlow <simonmar@microsoft.com>
  * FIX #903: mkWWcpr: not a product

[Isaac Dupree]

I guess it is intended to merge to 6.8? (Asking because it doesn't look like that's happened yet.)

[Isaac Dupree]

Okay, I see that is now done.