Build libclang_rt.asan-arm-android with stlport and API level 19.

Build libclang_rt.asan-arm-android with stlport and API level 19.

After https://codereview.chromium.org/951983002 ASan runtime library
is built with libc++-enabled Android toolchain. There is an apparent
unwinder bug there which is breaking ClusterFuzz.

With this change, we will build ASan runtime library with stlport,
and the rest of the code with libc++, which may sound scary, but
should work fine, since ASan has C-only interface (except for
new/delete operators).

Also, ASan is no longer supported on JellyBean, so bump the API level of the Android toolchain to 19 (KitKat).

BUG=492676

[eugenis, 2015-05-28 22:13:39]

eugenis@chromium.org changed reviewers:
+ hwennborg@google.com, thakis@chromium.org

[hans, 2015-05-28 22:33:52]

hans@chromium.org changed reviewers:
+ hans@chromium.org

[hans, 2015-05-28 22:33:53]

https://codereview.chromium.org/1161973004/diff/40001/tools/clang/scripts/upd...
File tools/clang/scripts/update.sh (right):

https://codereview.chromium.org/1161973004/diff/40001/tools/clang/scripts/upd...
tools/clang/scripts/update.sh:14: CLANG_SUB_REVISION=2
I was thinking we could land this without changing the revision to let the ToT
buildbot try it before we roll it in for real.

I'd like to hear what Nico thinks though..

https://codereview.chromium.org/1161973004/diff/40001/tools/clang/scripts/upd...
tools/clang/scripts/update.sh:684: --stl=stlport \
Hmm, didn't we just flip this in https://codereview.chromium.org/951983002. I
guess this is the scary part you refer to in the commit message :-)

[eugenis, 2015-05-28 22:37:57]

https://codereview.chromium.org/1161973004/diff/40001/tools/clang/scripts/upd...
File tools/clang/scripts/update.sh (right):

https://codereview.chromium.org/1161973004/diff/40001/tools/clang/scripts/upd...
tools/clang/scripts/update.sh:14: CLANG_SUB_REVISION=2
On 2015/05/28 22:33:53, hans wrote:
> I was thinking we could land this without changing the revision to let the ToT
> buildbot try it before we roll it in for real.
> 
> I'd like to hear what Nico thinks though..

OK, sure, reverted this chunk.

https://codereview.chromium.org/1161973004/diff/40001/tools/clang/scripts/upd...
tools/clang/scripts/update.sh:684: --stl=stlport \
On 2015/05/28 22:33:53, hans wrote:
> Hmm, didn't we just flip this in https://codereview.chromium.org/951983002. I
> guess this is the scary part you refer to in the commit message :-)

Exactly. It had no effect until the latest clang roll, and then *boom*
crbug.com/492676.

[Nico, 2015-05-29 00:56:21]

lgtm, thanks, up to you two what to do with the sub revision

https://codereview.chromium.org/1161973004/diff/40001/tools/clang/scripts/upd...
File tools/clang/scripts/update.sh (right):

https://codereview.chromium.org/1161973004/diff/40001/tools/clang/scripts/upd...
tools/clang/scripts/update.sh:14: CLANG_SUB_REVISION=2
On 2015/05/28 22:33:53, hans wrote:
> I was thinking we could land this without changing the revision to let the ToT
> buildbot try it before we roll it in for real.
> 
> I'd like to hear what Nico thinks though..

Nico is, as often, confused. Wouldn't the ToT buildbot try this with an updated
SUB_REVISION too? It seems strange to me to land this without this bump because
then running package.sh will give you a zip with the same name but different
contents.

[hans, 2015-05-29 01:19:09]

On 2015/05/29 00:56:21, Nico wrote:
> Nico is, as often, confused. Wouldn't the ToT buildbot try this with an
updated
> SUB_REVISION too? It seems strange to me to land this without this bump
because
> then running package.sh will give you a zip with the same name but different
> contents.

If we bump SUB_REVISION, we'll have to provide a package, and _all_ builds will
start using that package.

I was thinking it would be nice to get the ToT bot to test it first. But it is
strange too for the reason you mention. I guess we need ToT try-bots :-)

[hans, 2015-05-29 01:20:23]

Let's do this the usual way actually.

I'll try to get a new package built tomorrow and take it to the trybots the
usual way.

eugenis: Please ping me if I seem to forget.

[eugenis, 2015-05-29 18:06:58]

On 2015/05/29 01:20:23, hans wrote:
> Let's do this the usual way actually.
> 
> I'll try to get a new package built tomorrow and take it to the trybots the
> usual way.
> 
> eugenis: Please ping me if I seem to forget.

Now I'm confused. What's the usual way and do we bump sub_revision or not?

[hans, 2015-05-29 18:39:30]

On 2015/05/29 18:06:58, eugenis wrote:
> Now I'm confused. What's the usual way and do we bump sub_revision or not?

Sorry, I guess I was mostly talking to myself :-)

The usual way is: we do a CL like this (with the sub_revision bump), but before
landing or running tryjobs, we build the new clang packages for Linux & Mac and
push to goma. Once the package is on goma we can run tryjobs and land the patch
as any other patch.

I've built the packages, but the goma push failed. I cc'd you on that email..
I'm not sure what's going on, hopefully the goma team can fix it soon.

[Fabrice (no longer in Chrome), 2015-06-02 12:44:06]

fdegans@chromium.org changed reviewers:
+ fdegans@chromium.org

[Fabrice (no longer in Chrome), 2015-06-02 12:44:06]

I was going to change the minimum API level myself because the
ClangToTAndroidASan bot is broken.
We removed android-14 and only left android-16 and higher headers so this is no
longer working currently.

[hans, 2015-06-02 14:29:16]

hans@chromium.org changed reviewers:
- hwennborg@google.com

[hans, 2015-06-02 14:29:17]

I've pushed binaries based on this patch to goma. Sorry ti took so long.

Please bump CLANG_SUB_REVISION to 3, run the tryjobs described at
https://code.google.com/p/chromium/wiki/UpdatingClang and commit if they look
good.

[Nico, 2015-06-02 21:37:36]

Ok, CLANG_SUB_REVISION= needs to be changed to 3, and then this can land.

Eugenis, are you planning on doing that? If not, I'll do this in about one hour
(since the android/asan tot bot is red due to that ndk version thing)

[Nico, 2015-06-02 22:58:41]

Trying to land this here: https://codereview.chromium.org/1149863010/