Yes, there is a free SDK that compiles Objective-C for Android, and it's called "Android NDK" (which includes clang since, I think, r8c or something like that). Have you tried building GNUstep with it? :-)
To be actually productive, you need more than just what clang provides you. What can you do without having even NSDictionary, NSString, etc?
GNUstep build process for Android needs to be documented, GNUstep needs to be (slightly, in just a few places) patches, et al. I've done some work on it back in May, and there have been people who have submitted interesting, but incomplete or not fully appropriate patches.
Perhaps you were thinking of non-free SDKs -- and yes, on the mailing list we are aware of several of them. Every one of those companies is doing amazing things. If I were asked about it, I'd point someone doing commercial work on porting a video game toward those companies, because actually supporting game porting from iOS to Android requires much more than just gnustep-base; you need to provide people with (at least) chunks of UIKit, Core Animation, OpenAL, etc. and if you're porting an existing game, even if gnustep-base were available you'd save a LOT of time (meaning money) by going to one of these companies.
We could and should, however, provide at least basic upstream support for gnustep-base libraries.
If you're interested, here is a set of scripts intended to be used on stock Ubuntu:
You can use it to bootstrap the environment on a vanilla Ubuntu 12.04, or to study what needs to be done to get GNUstep working on Android. And no, it doesn't work completely; if I had a working .apk that I could launch on an Android device, I'd certainly announce it and write about it more than just throw the URL around.
The scripts still contain some non-upstreamed patches. Ideally we'd get that down to zero.
Feel free to play with it!
There is an SDK that compiles ObjC for Android already. :)
And yeah, I dont like using a VM just to test GNUstep builds. I am just waiting for the fixes of libobjc2 to be published so they can be compiled using clang. I tried a build with GCC - but it didn’t work out as expected due to incompatibilities with ABI. :/
I really want GNUstep to be more aknowledged and noticed - because its awesome. But it has issues it needs to fix - like deployment under Windows, or a working OS X build - that i am doing - and other things. o.o
I meant "developers participating in GNUstep project, with commit rights" :-)
And I by no means meant "this is unnecessary"; I meant "there is a lack of motivation for someone to sit down and go through it properly".
As David said, most core developers people use VMs. I personally sometimes reboot, as some OpenGL-related things are broken in VirtualBox. But noone has an actual day-to-day pressing need to make it work under OS X. I'd like it (to avoid the aforementioned reboots).
I don't think anyone's that dismissive about supporting GNUstep-directly-under-OS X; but, it would take work. I personally would think focusing on getting upstream gnustep-base to be usable under Android is more interesting :-)
Actually, he just said developers, which can mean "GNUstep maintainers" or "developers who use GNUstep in their projects". True, there may not be much need among the 1st category, but there is in the 2nd. That's what I referring to when mentioning the dismissed requests for help. I guess we could chalk it up as a misunderstanding on both ends :)
On Sun, Dec 22, 2013 at 3:22 PM, David Chisnall
<David.Chisnall@cl.cam.ac.uk> wrote:
On 22 Dec 2013, at 17:56, Jamie Ramone <
sancombru@gmail.com> wrote:
> I don't think there's a lack of need. Just look at how many people have been asking for help on this issue lately. Every time I see them the reply is usually "why would you or anyone want to do that?" Testing some software they're developing in Cocoa on GNUstep without the need of a VM as you said is just one example. They could want to do that to make sure it'll work properly on non-Mac systems. Or to move away from it. Or because of an issue with proprietary software, whether ideologically or practically inclined (e.g. licensing issue of some kind). In any case all those replies DO suggests reluctance, which is why I said what I said.
I definitely agree that there's a need to gave GNUstep working on OS X, however, to ease porting. I'd love it if we could ship and XCode plugin that would let people test their code with OS X in XQuartz and then just recompile on FreeBSD[1]. It would also be nice if we could bundle WINE and provide an environment for testing Windows builds. Now that Apple is shipping a recent clang, it's easy to cross compile, you just need a sysroot with the relevant libraries / headers and a GNU ld for the target platform.
Obviously, for real deployment, you're going to want to set up a VM (or a real machine) with the target platform and do QA there, but having to sync the code between the Mac and the VM seems to be too much of a barrier for some people.
David
[1] On Linux they'll need to also port to glibc most likely.