building HEAD fails when editline is installed as a user package

Building stage1 of HEAD fails when editline is installed as a user-level package, but not as a global package, in the bootstrapping compiler. I am using ghc-6.8.3 to build ghc-6.9.20080830.

/Users/judah/Programming/dontbackup/ghc-unvalidated/libraries/cabal-bin /usr/local/bin/ghc /Users/judah/Programming/dontbackup/ghc-unvalidated/libraries/bootstrapping.conf configure --distpref dist-stage1 \
                           --flags=stage1 --flags=-ghci --flags=ncg --flags=editline --ghc-options='-#include "cutils.h"' --with-compiler=/usr/local/bin/ghc --with-hc-pkg=/usr/local/bin/ghc-pkg --package-db /Users/judah/Programming/dontbackup/ghc-unvalidated/libraries/bootstrapping.conf \
                           --prefix=/NONEXISTENT --bindir=/NONEXISTENT --libdir=/NONEXISTENT --libexecdir=/NONEXISTENT --datadir=/NONEXISTENT --docdir=/NONEXISTENT --haddockdir=/NONEXISTENT --htmldir=/NONEXISTENT \
                           --libsubdir='$pkgid' --with-gcc=gcc --with-ld=/usr/bin/ld  --with-hscolour=/Users/judah/.cabal/bin/HsColour --with-alex=/usr/local/bin/alex --with-happy=/Users/judah/.cabal/bin/happy    --configure-option='--prefix=/Users/judah/tmp/ghc-stage' --configure-option=--with-cc=gcc --with-hsc2hs=/Users/judah/Programming/dontbackup/ghc-unvalidated/utils/hsc2hs/install-inplace/bin/hsc2hs \
                           --ghc-option=-DSTAGE=1
Configuring ghc-6.9...
cabal-bin: At least the following dependencies are missing:
editline -any
make[1]: *** [boot.stage.1] Error 1
make: *** [stage1] Error 1

Ghc decides whether to use editline in stage1 based on the flag GhcHasEditline, which the configure script sets to True in this case. Since the above cabal-bin command doesn't look at the user-level package DB, Cabal doesn't see the editline package and complains.

When the build system changes (post-6.10) to rely less on Cabal, this issue may go away. But for 6.10, I think that the easiest solution is to just never build stage1 with editline. Note that we already never build ghci in stage1 (see compiler/Makefile), so this seems reasonable to me. Patch is attached.

Trac metadata

Trac field	Value
Version	6.8.3
Type	Bug
TypeOfFailure	OtherFailure
Priority	normal
Resolution	Unresolved
Component	Compiler
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system
Architecture