On Sun, Jul 10, 2016 at 10:51 AM, Rob Landley <[email protected]> wrote:
> People keep using github to try to communicate. This is > https://github.com/landley/toybox/issues/35 which I'm replying to on the > mailing list... > > > ...like Git, Ruby, or Python? > > git yes eventually (although starting with a read-only version capable > of driving a repo checkout of the AOSP source for a build), but that's > probably post-1.0. > > Ruby and Python: no. They're way out of scope. > > > I'd love to use Android as a real programming device someday. > > I just gave a talk about that at txlf, which was (as seems traditional > for my txlf talks) far less coherent than I wanted it to be because my > machine with the carefully loaded set of web pages I wanted to walk > people through (some with sections mouse-highlighted, which is > apparently preserved when you change tabs in chrome) couldn't talk to > the darn projector (the magic resolution I didn't guess was 1280x1024, > dunno why ubuntu 12.04 could autodetect this but 14.04 can't), and > showing up early didn't help because the keynote before mine ran long so > my talk started several minutes late anyway, and then fiddling with the > monitor vs screen export stuff I accidentally hit function-f7 which > disables the touchpad (there's a thing to do that?) and couldn't figure > out how to switch it off again quickly, so I just borrowed a windows > laptop with a web browser and improvised, pulling web pages up on the > fly... meaning my talk wound up a bit like this paragraph, which I > intentionally didn't edit to give an example of my main communications > failure mode. :) > > > Sure, it may seem nonsensical to some, but I think making quick Git > > commits or Ruby code on-the-go would be fantastic. > > Ruby is out of scope. Python is out of scope. > > Python isn't even one language anymore, they have one of those flag day > changes from 2.0->3.0 the same way gtk breaks its API every few years > and loses half its' developers each time. This time the retention > failure is so bad python developers are blogging comparing the 2->3 > transition to the kubler-ross stages of grief. And no, I didn't make > that up: > > https://lwn.net/Articles/669768/ > > > Side note: nobody on Earth seems to care about cross-compiling > > utilities to iOS/Android. > > I don't know what you mean by that, but my goal is to make native > compling of arbitrary code on android work fine, possibly in some kind > of posix container/chroot. > yeah, i think they're confused --- certainly anyone who works on platform or third-party app native code would be very surprised to learn they haven't been cross-compiling all this time :-) assuming they mean the opposite, that no one cares about building on-device, yeah, that's largely true. doesn't mean folks haven't done it though, but it's mainly just for fun. i've built vim and gdb, and a guy in the next office has built clang. i don't personally know of anyone who's built make, but i'd imagine it's trivial given the things we have built. > > The process is rather difficult though, and Stack Overflow is of no > > help. Perhaps someone on this repository has a better understanding > > of GCC/Clang than I do? > > I'm told Android is already using the llvm backend to convert java > bytecode to native code (that's the "optimizing your apps 1/xxx" thing > that pops up when you uprade), so adding the clang frontend to that > isn't a huge step. The toolchain's already halfway there. > no, that was never really true, and i personally removed that failed experiment over a year ago. renderscript uses llvm, but renderscript's a lot more C-like. > But it isn't _useful_ until i get a lot more toybox stuff in. > > My http://landley.net/aboriginal/about.html project was a proof of > concept about what's the smallest/simplest build environment that can > not only rebuild itself from soruce code, but build linux from scratch > under the result. > > I'm currently grinding through the "what's next" section of that page: > > http://landley.net/aboriginal/about.html#next > > That boils down to cleaning out the GPL code by replacing > uclibc/busybox/glibc with llvm/toybox/musl (or a sufficiently upgraded > bionic), and fixing the problem that distro builds always expect a very > specific build environment and are full of assumptions. (I was poking at > debian and fedora, but now I'm focusing on AOSP.) > > If I can get AOSP to build in an aboriginal chroot based on llvm, > toybox, and musl (mostly by building its prerequisite packages first and > setting up the environment it expects, but there will probably be some > patches to AOSP itself), then getting AOSP to build under android is > just a question of reproducing that known environment in the new context. > ...and waiting several days :-) > Python and Ruby aren't involved in this. Git is, due to AOSP's > integration with repo. > although we do have several :-/ python prebuilts in AOSP, the build doesn't actually use any of them right now. but it should, and then the python problem goes away. > Rob > _______________________________________________ > Toybox mailing list > [email protected] > http://lists.landley.net/listinfo.cgi/toybox-landley.net > -- Elliott Hughes - http://who/enh - http://jessies.org/~enh/ Android native code/tools questions? Mail me/drop by/add me as a reviewer.
_______________________________________________ Toybox mailing list [email protected] http://lists.landley.net/listinfo.cgi/toybox-landley.net
