On Mon, 31 Aug 2015 17:52:36 +0200
ludo@gnu.org (Ludovic Courtès) wrote:

> Siniša Biđin <sinisa@bidin.eu> skribis:
> 
> > On 2015-08-23 22:47, Eric Bavier wrote:
> >> I had to replace "Exec=/etc/X11/Session/xmonad" with "Exec=xmonad"
> >> in this xmonad.desktop file in order to launch an xmonad session from
> >> Slim.  Does this seem right?
> >
> > Definitely!
> 
> [...]
> 
> >> Does it make sense to fetch this file
> >> from gentoo.org and then modify it? or should we rather just
> >> generate it
> >> ourselves?  What do you/others think?
> >
> > I had followed ratpoison's example (also using gentoo.org), but it does
> > feel dirty. I'm fine with generating it or having it stored in guix's
> > repo,
> > but couldn't find an example of any other package doing the same.
> 
> I think “generating” it (or having it as a string literal) in the source
> is preferable.
> 
> >>> +    (propagated-inputs
> >>> +     `(("ghc-mtl" ,ghc-mtl)
> >>> +       ("ghc-utf8-string" ,ghc-utf8-string)
> >>> +       ("ghc-extensible-exceptions" ,ghc-extensible-exceptions)
> >>> +       ("ghc-x11" ,ghc-x11)))
> >>
> >> Do these inputs need to be propagated?  I can start up an xmonad
> >> session without propagating these inputs.  Do they need to be available
> >> for xmonad's 'mod-q' (i.e. restarting xmonad/loading a new config)?
> >
> > I think so, yes. For one, since xmonad is an executable and a library,
> > if
> > we don't propagate, "ghc-pkg list" will report it as broken.
> 
> Makes sense to me.
> 
> Could you post an updated patch that addresses these two things?

BTW, attached is the current state of the patch that I have.  It
addresses the first item here (installing the xsession file).

For the second, I was toying with the idea of wrapping the xmonad
executable to provide both the ghc binary and the needed modules
(transitive propagated inputs, similar in style to what's done with
git-email currently, and for which I have an outstanding patch).  The
intent, IMO, would be to support xmonad's "reconfiguration".  I've not
used xmonad much before, but this appears to involve basically a
rebuild of the xmonad binary.  I'm not sure that just propagating the
input modules would be enough to support this.  It's also somewhat
unfortunate that the build would happen in a relatively polluted
environment.  Anyhow, sharing my current patch in case anyone else
wanted to play with it.

`~Eric