ext Peter Naulls wrote:
I'm a bit concerned by the impression given by the whole scratchbox idea, especially its rationale for existing. I'm not saying it doesn't have its place, but, as many people will atest to, the majority of basic unix utilities can and do cross compile, using well defined (is most cases, anyway) mechanisms provided by configure.
I wonder what you include in your "basic unix utilities", if you are mostly talking about non-GUI components then yes, I do agree with you that they can be cross-compiled in the normal way (we've done it), but as you move higher up in the software stack, things get ugly. I think we can take the openzaurus.org or GPE projects as an example. They do extensive autoconf-hacking to get their stuff cross-compiled. I'm fully aware of what they do (we've done that too) and that was the original motivation for Scratchbox.
Scratchbox aims at providing an environment where we can cross-compile a substantial subset of desktop applications for ARM with as few modifications as possible. This is needed to facilitate easy upgrade of the components, we don't want to get stuck with our hacked-up version of gtk. The effort to tweak an individual component may seem small, but this tweaking needs to be done over and over again as new versions come out.
Yes, it would be possible to fix the problems in the components themselves, but as openssh has repeatedly demonstrated, cross-compilation gets broken very often and very easily. We need the ability to cross-compile in order to be able to build all of the software in a timely manner.
Eventually we'd like to be able to cross-compile the "interesting" parts of Debian from scratch. There are other issues that would need to be solved first however. First of all Debian would need to be buildable from scratch natively before starting to think about doing the same by cross-compiling.
regards, Lauri