Hi Martin,

many thanks for your much cleaner solution. I feel a bit blameworthy
making you doing so much effort.

Probably, I've failed to communicate the out-of-source tree is not on
the critical path of Tobias' work. I've just tried to help him a bit by
doing some cosmetics, like integrating our run-script environment in his
port, and taking the kernel, and syscall-bindings build out of the
source tree.

The motivation was to build for e.g. 32-bit and 64-bit x86 without the
need of a duplicated source tree. I've only send the patch to the
mailing-list, because I thougt it might be of interest to you too,
because you've to build for different platforms all the time.
Nevertheless, I wasn't demanding to integrate it in the mainline kernel,
and giving you more work than you've already.
Despite that, your technical criticism was fully valid. In this respect
your reworked solution is much better. So thank you again.

Btw. Tobias and me did a bit pair-programming today, and I felt excited
while exploring HelenOS internals!

Best regards
Stefan

On 08.06.2012 15:18, Martin Decky wrote:
> Hi Stefan (and Tobias),
> 
> there is a new out-of-source-tree compilation tool in the HelenOS
> mainline branch as of changeset 1519. It creates a symlink farm in the
> destination directory and builds HelenOS there without touching the
> original source tree. It is pretty easy to use, just run (from anywhere)
> the tool dest_build.py by supplying the path to the source tree as the
> first argument and the path to the build/destination tree as the second
> argument. You can also append any Make arguments as you wish.
> 
> The benefit of this solution is that it is completely non-intrusive, no
> modification of our makefiles was necessary (except one minor bug fix).
> Modifying the tool to be able to build also individual components
> (kernel, uspace, etc.) should not be an issue.
> 
> I have also spent some time investigating the use of VPATH, but this is
> a dead end. The VPATH search algorithm in GNU Make is simply not strong
> enough for our complex makefiles. Many things could be solved by
> slightly twisting our makefiles, but for example sourcing files from
> upper-level directories (e.g. requiring ../version from the makefile in
> kernel/) unfortunately didn't cooperate with VPATH at all and I really
> don't like the idea of manually prefixing the paths with some variable
> all over the place.
> 
> 
> I hope this helps Tobias, I am certainly open to any suggestions. For
> the ultimate and most elegant solution we'll have to wait for the
> complete rewrite the HelenOS build system in Waf.
> 
> 
> M.D.
> 
> _______________________________________________
> HelenOS-devel mailing list
> [email protected]
> http://lists.modry.cz/cgi-bin/listinfo/helenos-devel

-- 
Stefan Kalkowski
Genode Labs

http://www.genode-labs.com/ · http://genode.org/

Genode Labs GmbH · Amtsgericht Dresden · HRB 28424 · Sitz Dresden
Geschäftsführer: Dr.-Ing. Norman Feske, Christian Helmuth

_______________________________________________
HelenOS-devel mailing list
[email protected]
http://lists.modry.cz/cgi-bin/listinfo/helenos-devel

Reply via email to