GitLab

This is just a temporary solution so we can run the tests. We'll want
to find a cleaner solution for when we actually go to ship (as
described in the bug).

Bug: http://b/25231709
Change-Id: Iede46ac2c3f7bd6dfbe376d1e7bb303b5a6bdc22

Bug: http://b/23362228
Change-Id: Ifc36687ffa5509c97f775456a94ffaa580f7cdab

Bug: http://b/23362228
Change-Id: I6877bf9cb6cd3ec8d2ce674364c2f2689cae500c

The Android.mk is required for providing ndk-build with information on
how to use gnustl.

Bug: http://b/24992609
Change-Id: Ide285e7a6dc01842b98f2a392e8c12c589e4d6ef

This is the libc++ and stlport equivalent of the same commit that was
done for gnustl. Package a single libc++.tar.bz2 instead of a tarball
for each architecture's libs and a separate one for the headers.

Bug: http://b/24992609
Change-Id: I7939632eb7637d5f5b5ab520225ab938aba159c4

Works on my machine. Looking at the log, I see that the host tag isn't
being included in the python package name. IIRC I saw this in testing
and it went away after I nuked /tmp/ndk-$USER >:(

Going to revert for now and sort it out later.

This reverts commit 6924c735.

Change-Id: Ia051f246b900b509061da62a734e65e3c0a243d9

GCC doesn't build mips64r2 by default any more, and it isn't an ABI we
need to support, so just drop it.

Change-Id: I9a24a367c0132f25f63af6ec547c173fab6a144b

There's no need to install multiple OS's versions of the NDK to the
same directory, so we don't need to keep the host tag in the path for
the host tools. This lets us drop find-win-host.cmd, which was just
for figuring out if we were 32-bit or 64-bit Windows.

Change-Id: I38d459ca68425b6ad20cac542cb60b273b0a2f8f

The new gcc is finally one that is built from the same source as the
platform rather than with NDK specific patches. One of those patches
was to put libs in the 4.9 directory rather than 4.9.x-google. Now
that we don't have that, fix the directory name.

Bug: http://b/24911020
Change-Id: I7f191e81c37759d0fdb908ad647a6322c17cd269

* changes:
  Package the tests for distribution.
  Remove tests gitignore.

Bug: http://b/23362228
Change-Id: I69059295f118610ac1bcacad55db2778cdab2987

Any files generated to this directory indicates a bug in the test
runner or build scripts.

Change-Id: I4b999521c60f0070f09805ac875a7d60815ad1fe

Instead of separately packaging the headers and libs for each ABI, put
all of gnustl into a single package.

Bug: http://b/24992609
Change-Id: If9cacde42cb542ab170349046a37bec7f94d9790

Bug: http://b/24990121
Change-Id: I2d8e35ab8a7ff4f8d79b04f8ead4d5144b62d6bc

Bug: http://b/24990121
Change-Id: Ia427e426b8e59dfaabdc7d1d6a769353430ce097

Bug: http://b/24990121
Change-Id: I93e4fd75634fa92b389240615bee737f2ed1d3dd

Windows binaries end with .exe, and Windows doesn't have gold.

Bug: http://b/22387400
Change-Id: I65edab9869c09164f05b029a1347ccddc072aceb

In a modular NDK world, we might have clang without GCC being present.
In this case, we still need libgcc.

Bug: http://b/24958112
Change-Id: I10a30b60aa6026c050064efcc1f076ad98a836ea

Bug: http://b/24950607
Change-Id: I136c45a64fd44476ca0443fabc94d4326bc14cbc

Rather than conquer the toolchain/build hellscape, just copy the
binutils files out of the prebuilt gcc into their own package.

This removes one of dependencies clang has on gcc being present. The
other is the target libraries provided by GCC.

Bug: http://b/22387400
Change-Id: I08a3985f2e46bee56cdbc7be58183d699310a75f

This is a part of gcc, not binutils, so clang won't necessarily have
it available.

Change-Id: I661f72b458cb999157de36cb0bf1c54091422977

Change-Id: Idaab9312cc76b09a91bbf09b0498e0ed5512bd95