simonmar at microsoft.com
Tue Jun 15 07:19:15 EDT 2004
On 15 June 2004 10:37, Ford Prefect wrote:
> I added a new option to ghc (6.2.1 release) -Of?le,
> wh?ch read opt?ons from a f?le us?ng unsafeperform?oç
> When I try to comp?le cop?s of the same f?le w?th
> d?fferent opt?ons (same d?rectory, but -o to d?fferent
> f?le) somet?mes I get error message attached.
> Th?s does not happen w?th unmod?f?ed ghc-6.2. Is
> anyth?ng speca?l about unsafeperform?o to cause th?s
If you re-ordered the core-to-core passes, it is possible that you violated an invariant somewhere. Not all orderings make sense. Some invariants will be documented (in the code), others may not. This amounts to modifying GHC, so we'll help where possible, but you have to be prepared to get stuck into debugging these problems yourself.
BTW, your email was all messed up (see quote above).
More information about the Glasgow-haskell-bugs