[GHC] #751: ghc 6.4.2 on OS X 10.4.6 fails a compiler error building Crypto-3.0.3.

[GHC]

#751: ghc 6.4.2 on OS X 10.4.6 fails a compiler error building Crypto-3.0.3.
----------------------------------+-----------------------------------------
  Reporter:  gwright at antiope.com  |          Owner:         
      Type:  bug                  |         Status:  new    
  Priority:  normal               |      Milestone:         
 Component:  Compiler             |        Version:  6.4.2  
  Severity:  normal               |     Resolution:         
  Keywords:                       |             Os:  MacOS X
Difficulty:  Unknown              |   Architecture:  powerpc
----------------------------------+-----------------------------------------
Comment (by gwright at antiope.com):

 Hmm.  This is strange.  I have run the build 4 times now, using {{{./Setup
 build -v > build.log 2>&1}}} and
 having edited the crypto-cabal file to invoke ghc with the {{{-v}}}
 option.

 One build ran to completion and exited with shell return value 0 (I'm
 using bash on OS X, so this is the output
 of {{{echo $?}}}).  The three subsequent runs exited with value 10.  I've
 also seen 11, but not when I was recording
 the output.  I've also not seen the weird activation record message in the
 recorded output, although I did see
 it once more this morning.

 The procedure I used:

 {{{
 crossroads-able> cd work/Crypto-3.0.3/
 crossroads-able> sudo cp ~/Desktop/crypto.cabal .
 crossroads-able> sudo ghc -o Setup Setup.hs -package Cabal
 crossroads-able> sudo ./Setup configure -v > ~/Desktop/conf4.log 2>&1
 crossroads-able> sudo ./Setup build -v > ~/Desktop/build4.log 2>&1
 crossroads-able> echo $?
 10
 }}}

 I copy in my edited crypto.cabal file to run ghc with the {{{-v}}} option.

 The recorded configuration files are identical, up to the naming of
 temporary files.

 Attached are

 1) conf.log, the output of {{{./Setup configure -v}}} ;
 2) build.log, a successful build;
 3) build[234].log, builds that gave return value 10.

 The odd thing is that the failures are inconsistent.

 Also, GHCRTS is not set to anything:

 {{{
 crossroads-able> echo $GHCRTS

 crossroads-able>
 }}}

 and the crypto.cabal file make no reference to RTS options.

-- 
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/751>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler