packaging taffybar and problem with dummy package

Martin Bukatovic martin.bukatovic at gmail.com
Tue Nov 3 17:37:25 UTC 2015


On Mon, Nov 02, 2015 at 05:23:49PM -0500, Ben Boeckel wrote:
> This seems…wrong. It would probably be easiest to just patch out that
> from the .cabal file.

Thanks, that resolves it. I'm going to patch it then.

> FWIW, git-annex works fine, but it is already
> packaged and updating just for Rawhide is what I'll be doing:
>
>     http://copr.fedoraproject.org/coprs/mathstuf/git-annex/

It seems that git-annex upstream uses more friently definition of
network-uri flag which doesn't require dummy package, see:

~~~
  if flag(network-uri)
    Build-Depends: network-uri (>= 2.6), network (>= 2.6)
  else
    Build-Depends: network (< 2.6), network (>= 2.0)
~~~

So you don't have this problem with git-annex.

On the other hand, I see that network-uri flag is left in it's default True
state, so I don't understand how it could work with network-2.4 module.

Btw unrelated question: looking at this patch:

http://copr-dist-git.fedorainfracloud.org/cgit/mathstuf/git-annex/git-annex.git/tree/git-annex-5.20150812-flags.patch

wouldn't be better to use cabal-tweak-flag script instead of patch which
changes flags? Like this:

~~~
%prep
%setup -q
/usr/bin/cabal-tweak-flag S3 False
/usr/bin/cabal-tweak-flag WebDAV False
/usr/bin/cabal-tweak-flag inotify True
...
~~~

Or is there some reason I don't recognize for which patch like this is
better compared to cabal-tweak-flag script in spec file? I'm just curious - as
I said I'm new to haskell packaging.

Thank you for your reply.

--
Martin Bukatovic


More information about the haskell mailing list