On Fri, Oct 07, 2016 at 07:34:26PM +0000, Clint Adams wrote: > Things look a lot better now:
So I actually think this is viable, with two main blockers: 1) Every -doc package generated by ghc 8.0.1-3 has lintian errors because of <script> tags linking to cdn.mathjax.org. ghc 8.0.0.20160111-1's changelog mentions this exact issue, and ghc-doc still has a libjs-mathjax dependency, so someone will have to figure out what went wrong and fix it. 2) esqueleto needs to be patched to work with persistent 2.6. This will break git-annex and so git-annex will have to be patched as well. It would be better if this could be fixed upstream so we don't end up breaking API compatibility, but that doesn't seem to be happening. #1 is currently blocking the building and upload of a handful of necessary packages to NEW. Otherwise it's a matter of patching or removing (and all the reverse dependencies of) things like haskell-csv-conduit, anything with "haskelldb" in the name, haskell-hsh, haskell-iso8601-time, haskell-glib, haskell-polynomial, haskell-http-link-header, hasktags, haskell-cabal-helper, haskell-css-text, and dealing with any arch-specific failures. Anyone want to do these things?