[xmonad] xmonad on OLPC laptop?

[Spencer Janssen]

On Sat, Jan 19, 2008 at 02:45:31AM -0500, Nick Cabatoff wrote:
> Hi folks,
> 
> I've been using xmonad since 0.3 and loving it.  Thanks to all who've
> contributed for their hard work; the WM is so good that you managed to infect
> me with the Haskell bug just so I could customize it, though I've yet to
> actually write any original code in it.
> 
> Anyway, we finally got our OLPC XO-1s up here in Canada this week and now I'm
> trying to get xmonad running on it like I've seen in the pictures.  This is on
> debian etch, installed via olpc-update debian-big.  After a bit of frustration
> with ghc6.6, which has worked well for me on Ubuntu feisty and gutsy, I moved
> my sources.list to unstable and upgraded to 6.8.
> 
> I had good experiences building xmonad 0.5 and xmobar on gutsy using
> cabal-install, so I tried to go that route this time too.  I got all the
> dependancies built, but then when it came to the final link stage I ran out of
> memory.  Added swap, but then it got very slow once ld crept around the 200MB
> mark (there's 237604k available total according to top, excluding swap) and
> vmstat showed that the system was spending virtually all of its time waiting.
> I suppose I could connect a USB hard disk rather than building off SD card, but
> it was easier to abandon cabal-install and go with the packages Joachim
> Breitner has announced on this list.

There is an issue with split objects and some versions of binutils -- you're
probably observing that.  You can either rebuild the libraries with split-objs
off or try to upgrade binutils.

> I installed the .deb files from ~nomeata and did apt-get install -f to get the
> missing dependancies.  I had to delete some gnome stuff to make it all fit but
> that won't be missed.  Everything works like a charm if I just run startx and
> put xmonad by itself in my .xinitrc.  Very nice work Joachim.  However, I'd
> ideally like to have the same setup as on my desktop, which means xmobar.
> There, I have in my .xsession:
> 
>     PIPE=$HOME/.xmonad.log                                                                  
>     rm -f $PIPE                                                                             
>     PATH=${PATH}:/sbin mkfifo -m 600 $PIPE                                                  
>     [ -p $PIPE ] || exit                                                                    
>     xmobar &                                                                                
>     xmonad > $PIPE    
> 
> xmobar was easy enough to build too.  When I run startx however, xmobar
> launches normally but xmonad quickly dies.  Or that's what I thought at first.
> I tried changing the line to
> 
>     strace -o /tmp/s -f xmonad > $PIPE    
> 
> and the strace output file wasn't created.  So it seems like the shell running
> .xinitrc is getting stuck trying to open $PIPE for xmonad's stdout, and never
> getting around to launching xmonad.
> 
> I thought maybe it might be a JFFS2 thing, so I tried changing $PIPE to use an
> ext3 volume, but that didn't help either.  Anyone else have any ideas?
> 
> To be honest, I'm not sure I'm really going to be willing to sacrifice any of
> my precious screen real estate on this machine to xmobar, but I'd like to give
> it a try before dismissing it.

Rather than using a FIFO to communicate with xmonad, you can use xmobar's
PipeReader plugin.  Then the much simpler .xinitrc should suffice:

    xmonad | xmobar


Cheers,
Spencer Janssen