> kbuild 2.5 handles this by constructing a set of symlinks then invoking
> the configure system under those symlinks, followed by copying any
> results to their destination. The symlink tree completely isolates
> the config system from any knowledge of where its inpuuts and outputs
> really are, everything looks local.
I noticed this 'magic' in kbuild-2.5.
As you see a lot of cruft is required to circumvent the fact that the
current config tools are hardcoded to read from current directory
and hardcoded to write to current directory.
> You have a 750+ line patch to imbed tree knowledge into configure, that
> knowledge will have to be duplicated for any new CML tools. kbuild 2.5
> does it in a few lines of scripts/Makefile-2.5 which automatically
> works for any new CML code.
Did you look in the patch?
It basically teach the config tools that the SRC are no longer
in current directory but pointed out by $srctree, that output files
are pointed out by $objtree, and temporary files the same place.
No nasty tricks with sym-lnks required, no copy files around before
or after the config tools are used. No specific directories that
needs to be created beforehand.
Indeed my approach is a number of lines - but that on the other hand
simplify the usage of the config tools.
But I see your point that we should avoid hardcoding too much
knowledge in the config tools, and I may change the patch to
use command-line parameters to specify SRC and OBJ dirs.
Sam
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/