* Richard W.M. Jones:
Patches 2 and 3 seem very brutal. Isn't there a better way than recursively copying everything into the build directory?
I asked around on IRC and StackOverflow and the only halfway useful answer I got so far was <http://stackoverflow.com/questions/7032942>. It does not account for the test cases and Schwern even called it "painful". After spending too much time looking at autoconf etc. in the past few weeks I haven't really felt like learning how Ruby's idea of a build system works. (I'd rather spend the time learning about OCaml.)
I have applied the other patches (1, 4, 5, 6). I'm just doing a bit of testing, but will push them after. I only care about the srcdir == builddir case so I'm happy as long as that doesn't break.
Fair enough. The only reason I am interested in this is because it seems to be the sanest way to build Python bindings for multiple versions (currently 2.6, 2.7, 3.2 for the unstable distribution).
I have a few comments about patch 1 below.
-stamp-guestfs-recipes.pod: guestfs-recipes.pod create_disk.c inspect_vm.c - $(top_srcdir)/podwrapper.sh \ +stamp-guestfs-recipes.pod: $(builddir)/guestfs-recipes.pod create_disk.c inspect_vm.c + $(top_builddir)/podwrapper.sh \
This change appears to be wrong. guestfs-recipes.pod is be a source file, isn't it?
The source and target expressions don't need $(srcdir) or $(builddir) in most cases, I think. I'll do some more testing and then post another series patches. Cheers, -Hilko