On Mon, Sep 6, 2010 at 4:52 PM, Angus Lees <g....@inodes.org> wrote:
On Tue, Sep 7, 2010 at 05:35, Tim Mensch <tim....@gmail.com> wrote:
On 9/6/2010 6:18 AM, Peter wrote:
Well, I tested it and now I'm confused, because without any flags
defined this _doesn't_ fail:
bla i = 5;
I can verify this; as a result of the earlier messages in this thread, I
changed to __ANDROID__ for my #define flag, but I had to change my -DANDROID
to -D__ANDROID__ to get it to work, as Peter described above.
To be fair, I'm using the Crystax build of the NDK, which may potentially
be missing __ANDROID__ ?
Oh, sorry for the misinformation :(
I've been using custom-built toolchains for so long I have very little idea
of what the NDK gcc actually does... I can confirm that the new gcc-4.5
"arm-none-linux-androideabi" target declares __ANDROID__ internally (see
http://gcc.gnu.org/onlinedocs/gcc/GNU_002fLinux-Options.html for example).
Wow, that's stunning news that Android support has reached the gcc
mainline. That should mean Android support in the CodeSourcery
toolchain can't be far behind.
Mind you, it's too bad that we're scrambling around trying to find
alternative toolchains that support the features we really need.