
Cheers Simon, thanks for looking at this. On Mon, 2006-04-24 at 10:43 +0100, Simon Marlow wrote:
Duncan Coutts wrote:
I'm concerned that there are many different versions of Cabal "1.1.4". The version that GHC 6.4.2 is now shipping is actually a very old version of Cabal 1.1.4.
Sorry, this is my fault: I just copied the CVS version of Cabal into 6.4.2, and didn't bump the version on the mainline at the same time.
I've now bumped the mainline Cabal to 1.1.5, and I've tagged the version in 6.4.2 with the tag "shipped in GHC 6.4.2". That version is circa Jan 15, 2006. I didn't copy Cabal/test in 6.4.2, so you will see some differences in there if you 'diff -cr', BTW.
Ok.
So we have a deadzone between Jan 15 and now (last commit was Apr 11) where the mainline code claimed to be 1.1.4 and exposed various different APIs from the one shipped in 6.4.2. But there were no actual releases of that code, fortunately.
The right way to proceed, I think, is to claim that the Cabal shipped in 6.4.2 *is* 1.1.4.
That seems quite a reasonable way to proceed. The version shipped with GHC will have a wider user base and the people who downloaded the 1.1.4 version from the Cabal darcs repo knew it was a dev version.
Isaac; is that ok with you? Can you put up a tarball of that tag for the Gentoo guys to use?
That'd be great.
Is there anything we can do for the next release of GHC/Cabal so that we don't end up with this mess?
Yes, I should coordinate with Isaac more closely next time, and we should coincide the GHC release with a Cabal release. Either that, or GHC should be using a branch of the Cabal repository that corresponds to a recent stable version.
Ok Great. Thanks again. Duncan