Hello Cafe, I noticed on my package's hackage page there is a build failure message: http://hackage.haskell.org/package/txt-sushi I don't know if these are real errors or not (I don't experience them on OS X and it's pure Haskell code) but I did poke around and noticed that some popular packages on hackage also have build failures: http://hackage.haskell.org/package/xmonad http://hackage.haskell.org/package/haddock http://hackage.haskell.org/package/pandoc http://hackage.haskell.org/package/alex http://hackage.haskell.org/package/hlint ... so what should I make of these errors? Are they useful in some way or just a problem with the build environment? (If that's the case I think they should probably be removed since they're confusing for potential users) Best Keith -- keithsheppard.name
I noticed on my package's hackage page there is a build failure message: http://hackage.haskell.org/package/txt-sushi
I don't know if these are real errors or not (I don't experience them on OS X and it's pure Haskell code) but I did poke around and noticed that some popular packages on hackage also have build failures:
Just as an extra data point, I've found that some of the build errors on Hackage are due to the machine's build state and not due to any intrinsic failure in the package in question. Furthermore, Hackage does not generate Haddock for these packages even though generating Haddock would still be 1) possible and 2) useful. I think it would be better to divorce the build process from the Haddock generation, since they aren't necessarily related (and Haddock builds will fail if the package *is* trivially broken.) -- Jonathan Daugherty
On Mon, 2010-01-11 at 19:59 -0500, Keith Sheppard wrote:
so what should I make of these errors? Are they useful in some way or just a problem with the build environment?
The latter.
(If that's the case I think they should probably be removed since they're confusing for potential users)
It is confusing. On the other hand if we remove the build results for all packages then there's no useful info at all. Is partial info better than none? Personally I rather agree with you, I argued for not displaying any results until it was of sufficient quality. On the other hand, if I'd had my way then perhaps there still wouldn't be any build info at all! Duncan
participants (3)
-
Duncan Coutts -
Jonathan Daugherty -
Keith Sheppard