Hi Dominick, An "ExitFailure 1" is definitely not helpful and we should generally try to figure out what went wrong and print something more specific. One example where this happens if I recall correctly is if GHC runs out of RAM and exits, something that we could detect. I think the best way forward right now is to file bugs for concrete cases where the error reporting is bad. Include what (you think) went wrong and what kind of error information would have been useful to you. The bug tracker is at https://github.com/haskell/cabal/issues On Mon, Mar 24, 2014 at 3:40 PM, Dominick Samperi wrote:

I have seen the dreaded "ExitFailure 1" many times when a package fails to install (under Linux or Windows), and I wonder why something more helpful cannot be printed? Sometimes one gets a variant of this where the message is "this system is not compatible with this package," but the user must study the foo.cabal file to determine what is missing.

Also, it seems that the only way to determine if a package will install under Windows is to try and hope for the best (often getting "ExitFailure 1" if it fails). Wouldn't it make more sense to tag packages with the OS and other dependencies and fail with a helpful message like "Windows not supported" if appropriate?

It appears that some packages (like lens-4.1) have version dependencies that cannot be satisfied (at least not on my Linux and Windows boxes), so it appears that the Hackage hosting process suffers from a variant of the well-known "dll hell" problem.

I tried the new (Cabal 1.18.0) sandbox feature but this did not help in my case.

Any comments on these issues would be appreciated.

Thanks, Dominick _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe