Hello,

Christopher Baines <mail@cbaines.net> writes:

> Some really minor description things I missed earlier:
>
> ghc-aeson-better-errors, ghc-transformers and ghc-websockets have URLs
> in the description, @url could probably be used.
>
> I'd also remove the following from the ghc-bower-json description, it
> reads more like uninformative advertising to me and isn't really
> relevant to this specific package:
>
>   Bower is a package manager for the web (see http://bower.io).

Done!

> Maybe it would be good to note why the package in Guix didn't work, like
> I tried to use ghc-happy@1.19.12, and the build phase failed with some
> type errors. That'll at least point out clearly that there's an
> incompatability. I don't think much detail is needed, the following
> would be fine.
>
>      `(("ghc-happy" ,ghc-happy-1.19.9)  ; build fails with 1.19.12

Done!

>> Ah thanks for the suggestion here. I found one that could be taken from
>> the existing package in (gnu packages haskell-xyz). The rest seemed to
>> be either pinned exactly or too strict to take from guix packages.
>
> Thanks, although my intent was more about adding something like this.
>
>   #:phases
>   (modify-phases %standard-phases
>     (add-after 'unpack 'patch
>       (lambda _
>         (substitute* "purescript.cabal"
>           (("clock .*$")
>            "clock,\n"))
>         #t)))))

With the addition of a patch I was able to relax most of the package
versions I introduced to ones we have. The final remaining required
dependency seems to be happy which I kept at 1.19.9.

> Now I'm not sure quite how risky this is, the purescript tests aren't
> being run, but given this is Haskell and it compiles, hopefully that
> provides some guarantees.
>
> What do you think?

My only caveat too is now I have not really tested this version. I had
done some work with a prior version.  I wish I could run the tests but
they seem to rely heavily on npm and bower.

Thanks again,

John