[haskell-llvm] Upcoming release, high level binding state

[Henning Thielemann]

On Fri, 26 Apr 2013, Bryan O'Sullivan wrote:

>       Is there any release procedure beyond the obvious tag-and-upload-to-hackage I should be aware
>       of?
> 
> 
> Not really. When I was actively maintaining the package, I'd do roughly the following:
>  *  cabal sdist
>  *  Build the package and run the tests from the resulting tarball (not from the git tree), since it's easy
>     to miss important files in the tarball

I have put the packages cabal-scripts and darcs-scripts to Hackage that do 
e.g. this testing from a tarball.

http://hackage.haskell.org/package/cabal-scripts
http://hackage.haskell.org/package/darcs-scripts

>  *  Try with all supported versions of LLVM (whatever those are today)
>  *  If everything is good, tag, upload, and push

> My convention has been to number the Haskell package based on whatever current version of LLVM it is
> targeted at, so LLVM 3.2 -> llvm-base 3.2.0.0 and so on.

You may use a slightly modified PVP where the first three numbers (not 
only the first two ones) constitute the major version. This way you can 
make API changes during one version of LLVM. I also find it an advantage 
to increase the package version number in the repository right after a 
Hackage release. This way I can work with versions of the llvm package 
both from Hackage and from git.