Record selectors behaving badly wrt optimisation

The attached program performs much worse with HEAD as of today (and presumably also 6.10.1) than 6.8.3.

> ghc-6.8.3 --make -O2 bug.hs
[1 of 1] Compiling Main             ( bug.hs, bug.o )
Linking bug ...
> time ./bug
1.17s real   1.17s user   0.01s system   99% ./bug
> ghc-6.11 --make -O2 bug.hs 
[1 of 1] Compiling Main             ( bug.hs, bug.o )
Linking bug ...
> time ./bug                    
1.58s real   1.58s user   0.01s system   99% ./bug

The problem appears to be in the code for delete, GHC has duplicated two calls to next.

Also note this is a version of the code we have in eyeball/IOList.lhs which illustrates some other problems in the current optimiser (although the other problems are not regressions, which is why I'm highlighting this one).

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