Attached are tarballs with the new code and patches for existing
makefiles and configure.in stuff.
I've bound GConfClient & GConfValue. I've not bound GConfEntry or
GConfSchema. I may bind these; opinions on how useful they might be are
Like the other gtk2hs extensions, I've put it in a separate directory
and it gets its own package "gconf". The other tarball is a demo (which
contains instructions on how to make use of it).
* gconfClientNotifyAdd is not yet bound. This is a priority.
* I just noticed, it doesn't yet deal with getting unset keys.
GconfClient returns null in this case which makes the
marshaling routines segfault. Which would be better: change the
type signature to Maybe or raise an exception? (Or allow either
using overloading via the GConfValueClass - if you expect a
Maybe you'll receive Nothing; if you expect a raw value, you'll
get an exception)
* error checking, currently passing nullPtr for the GError
return. GConfClient's default behaviour in this case is to print
warning messages (eg incorrect type / key does not exist).
I'm using the signal generator on the gconfmarchal.list. I would be able
to use the existing Signal.chs-boot* files if the HookGenerator were a
little less flexible in naming the module. At the moment, the module
name is expected to be in Signal.chs-boot1. But the name must always be
the filename (less the .chs extension) so the HookGenerator could just
do that. I've attached a quick ugly hack to show what I mean.