Good work David. I used to do something like this for Cabal regression testing but the method I used didn't scale well as hackage grew. I'll look into using your tool next time for testing a major Cabal / cabal-install release. Duncan On 23 April 2012 21:37, David Terei wrote:

Hi all,

I've updated the old hackage-test tool and renamed to hackager.

http://hackage.haskell.org/package/hackager

Hackager is a tool to automate the compiling of all packages on Hackage. It builds each package on hackage in isolation and records the results. The purpose being to catch regressions caused by changes to GHC (and Cabal although this was not the motivation). Two runs of Hackager can be compared, so the first run is done with a known version of GHC and the next run with a new, experimental version of GHC... ect.

The improvements to Hackager over hackage-test are: * Parallelized the build process. Can now specify how many packages to build in parallel, which cuts total run time down greatly (e.g 2 days -> 5 hours) * hackage-test and hackage-report are now one tool, 'hackager' that works as a mutli-command tool. * Proper option handling * Fixed some stability issues

The new homepage for development can be found here: https://github.com/dterei/Hackager

Cheers, David

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