Add infrastructure to run tests on android.

Add infrastructure to run tests on android.

This CL have the following changes:
- Build mojo_system_unittests_apk on android
- Fix tests to run on android
- Fix multi-process test to run on android
- Fix mojob.py to run tests on android
- Fix mojo/tools/test_runner.py to run tests on android

R=viettrungluu@chromium.org

Committed: https://chromium.googlesource.com/external/mojo/+/f4583a88c6184faac8eb4035b5e8948b2e3b3240

[viettrungluu, 2014-11-14 16:36:32]

Sorry, I'm not going to approve any changes to the scripts (except perhaps the
most trivial) that don't have tests.

It's (partly) my fault that existing things don't have tests, but they were
written before "we" decided to make them infrastructure. Now that they are
infrastructure, they need tests.

https://codereview.chromium.org/728783003/diff/1/mojo/edk/embedder/test_embed...
File mojo/edk/embedder/test_embedder.h (right):

https://codereview.chromium.org/728783003/diff/1/mojo/edk/embedder/test_embed...
mojo/edk/embedder/test_embedder.h:19: MOJO_SYSTEM_IMPL_EXPORT bool
IsInitialized();
Nope, not going to allow this: if you don't know when/where you're initializing,
you're doing it wrong.

https://codereview.chromium.org/728783003/diff/1/mojo/tools/test_runner.py
File mojo/tools/test_runner.py (right):

https://codereview.chromium.org/728783003/diff/1/mojo/tools/test_runner.py#ne...
mojo/tools/test_runner.py:32: debug_group.add_argument('--debug', help='Debug
build (default)',
By design, this should not know about
debug/release/asan/whatever-other-config-you-might-come-up-with.

https://codereview.chromium.org/728783003/diff/1/mojo/tools/test_runner.py#ne...
mojo/tools/test_runner.py:38: os_group.add_argument("--android", help="Run tests
for android",
It probably shouldn't have knowledge of android either. You should come up with
a more general way of plumbing in what's required.

[qsr, 2014-11-14 16:55:09]

On 2014/11/14 16:36:32, viettrungluu wrote:
> Sorry, I'm not going to approve any changes to the scripts (except perhaps the
> most trivial) that don't have tests.
> 
> It's (partly) my fault that existing things don't have tests, but they were
> written before "we" decided to make them infrastructure. Now that they are
> infrastructure, they need tests.

 What kind of test are you thinking about?

[qsr, 2014-11-14 17:02:33]

https://codereview.chromium.org/728783003/diff/1/mojo/edk/embedder/test_embed...
File mojo/edk/embedder/test_embedder.h (right):

https://codereview.chromium.org/728783003/diff/1/mojo/edk/embedder/test_embed...
mojo/edk/embedder/test_embedder.h:19: MOJO_SYSTEM_IMPL_EXPORT bool
IsInitialized();
On 2014/11/14 16:36:32, viettrungluu wrote:
> Nope, not going to allow this: if you don't know when/where you're
initializing,
> you're doing it wrong.

 Well, then I'm open to suggestion. The issue is the following:
Multi process tests on android cannot use exec because there is no executable to
exec. Instead they just use fork, execute the code then exit.

 The code that is executed expect to be in a newly executed process. On the
other hand, the code doing the forking doesn't know anything about the current
context, so it doesn't know if the embedder is initialized or not, but it needs
to make sure there is no embedder before calling the forked code.

 Given this, I cannot know if the embedder is initialized or not...

https://codereview.chromium.org/728783003/diff/1/mojo/tools/test_runner.py
File mojo/tools/test_runner.py (right):

https://codereview.chromium.org/728783003/diff/1/mojo/tools/test_runner.py#ne...
mojo/tools/test_runner.py:38: os_group.add_argument("--android", help="Run tests
for android",
On 2014/11/14 16:36:32, viettrungluu wrote:
> It probably shouldn't have knowledge of android either. You should come up
with
> a more general way of plumbing in what's required.

 Hum. Running unit test on android is not the same as on other platform. You
cannot just run an executable. We have a test_runner script that handles this. I
guess I can build a script that will call the script, but I still need to know
that it is an android build to check what the apk is and if it changed.

 Also, note that this script is already doing some check on platform.system()
but we cannot do the same for android because we are cross-compiling.

[qsr, 2014-11-17 15:39:27]

I added a test that check what is the list of command run for given parameters.

I removed the IsInitialize() parameter, but it means that the embedder has to be
initialized to allow multi process tests.

I also removed the --debug and --release parameters

For the --android parameter, I really do not see how to change that. Are you ok
with it, or do you have a suggestion?

[viettrungluu, 2014-11-17 23:56:39]

On 2014/11/17 15:39:27, qsr wrote:
> I added a test that check what is the list of command run for given
parameters.
> 
> I removed the IsInitialize() parameter, but it means that the embedder has to
be
> initialized to allow multi process tests.
> 
> I also removed the --debug and --release parameters
> 
> For the --android parameter, I really do not see how to change that. Are you
ok
> with it, or do you have a suggestion?

We should probably provide a way of injecting an execution wrapper. (We can then
reuse this to only run xvfb on a smaller subset of tests, etc.)

[viettrungluu, 2014-11-17 23:56:58]

https://codereview.chromium.org/728783003/diff/40001/mojo/edk/embedder/test_e...
File mojo/edk/embedder/test_embedder.h (right):

https://codereview.chromium.org/728783003/diff/40001/mojo/edk/embedder/test_e...
mojo/edk/embedder/test_embedder.h:19: MOJO_SYSTEM_IMPL_EXPORT bool
IsInitialized();
Please delete.

https://codereview.chromium.org/728783003/diff/40001/mojo/edk/test/multiproce...
File mojo/edk/test/multiprocess_test_helper.h (right):

https://codereview.chromium.org/728783003/diff/40001/mojo/edk/test/multiproce...
mojo/edk/test/multiprocess_test_helper.h:75: // On android, multi-process tests
cannot call exec. It means that the test must
In general, executing in a forked but not-exec-ed process seems pretty scary and
fragile. Is there any way of avoiding doing this?

[qsr, 2014-11-18 09:47:37]

On 2014/11/17 23:56:39, viettrungluu wrote:
> On 2014/11/17 15:39:27, qsr wrote:
> > I added a test that check what is the list of command run for given
> parameters.
> > 
> > I removed the IsInitialize() parameter, but it means that the embedder has
to
> be
> > initialized to allow multi process tests.
> > 
> > I also removed the --debug and --release parameters
> > 
> > For the --android parameter, I really do not see how to change that. Are you
> ok
> > with it, or do you have a suggestion?
> 
> We should probably provide a way of injecting an execution wrapper. (We can
then
> reuse this to only run xvfb on a smaller subset of tests, etc.)

Not sure how to do this, without making this thing a beast. It is not just an
execution wrapper for android case, there is no executable to run, instead we
need to run another python script.

 If we allow this to run anything, then it can't do a lot, because it cannot
just take the hash of what it is executing (right now, it recursively look for
dynamic libraries, and in the case of android, the python script is never
changing, the apk is).

 So, I don't really see how to prevent this script to get some knowledge about
what it runs.

[qsr, 2014-11-18 09:55:27]

https://codereview.chromium.org/728783003/diff/40001/mojo/edk/embedder/test_e...
File mojo/edk/embedder/test_embedder.h (right):

https://codereview.chromium.org/728783003/diff/40001/mojo/edk/embedder/test_e...
mojo/edk/embedder/test_embedder.h:19: MOJO_SYSTEM_IMPL_EXPORT bool
IsInitialized();
On 2014/11/17 23:56:58, viettrungluu wrote:
> Please delete.

Done.

https://codereview.chromium.org/728783003/diff/40001/mojo/edk/test/multiproce...
File mojo/edk/test/multiprocess_test_helper.h (right):

https://codereview.chromium.org/728783003/diff/40001/mojo/edk/test/multiproce...
mojo/edk/test/multiprocess_test_helper.h:75: // On android, multi-process tests
cannot call exec. It means that the test must
On 2014/11/17 23:56:58, viettrungluu wrote:
> In general, executing in a forked but not-exec-ed process seems pretty scary
and
> fragile. Is there any way of avoiding doing this?

 Not without a lot of infrastructure work.

 There is no executable to exec. If we wanted this to work, we will have to
modify gtest infrastructure to build application as well as the .so that is run
via jni. Then we will need to bundle the application in the app, and then exec
it. I'm not sure it will even work on a device. Also Android can decide to break
that at any time as they really do not want to have people doing this. And also,
if there is an JNI calls in the exec process, it will miserably fail because
there won't be any Dalvik on it.

 In the end, the real choice we have is doing this (which works, all test are
passing locally), or not run any multi-process tests on Android, which I find a
little bit scary.

[qsr, 2014-11-19 11:57:51]

gentle ping?

[viettrungluu, 2014-11-20 17:32:24]

I don't think we should run multiprocess tests on Android.

It's too hacky and fragile.

https://codereview.chromium.org/728783003/diff/60001/base/test/multiprocess_t...
File base/test/multiprocess_test_android.cc (right):

https://codereview.chromium.org/728783003/diff/60001/base/test/multiprocess_t...
base/test/multiprocess_test_android.cc:1: // Copyright (c) 2012 The Chromium
Authors. All rights reserved.
Changes to this file will get clobbered in the next Chromium -> Mojo update.

https://codereview.chromium.org/728783003/diff/60001/mojo/tools/data/android_...
File mojo/tools/data/android_unittests (right):

https://codereview.chromium.org/728783003/diff/60001/mojo/tools/data/android_...
mojo/tools/data/android_unittests:1: # This file contains a list of Mojo gtest
unit tests.
Having to maintain a separate list of tests is awful.

https://codereview.chromium.org/728783003/diff/60001/mojo/tools/mopy/transiti...
File mojo/tools/mopy/transitive_hash.py (right):

https://codereview.chromium.org/728783003/diff/60001/mojo/tools/mopy/transiti...
mojo/tools/mopy/transitive_hash.py:64: def transitive_hash(filename,
transitive=True):
Move _file_hash() into its own file and use that instead.

https://codereview.chromium.org/728783003/diff/60001/mojo/tools/test_runner.py
File mojo/tools/test_runner.py (right):

https://codereview.chromium.org/728783003/diff/60001/mojo/tools/test_runner.p...
mojo/tools/test_runner.py:112: command = [ "python" ]
nit: no spaces inside [, ]

https://codereview.chromium.org/728783003/diff/60001/mojo/tools/test_runner.p...
mojo/tools/test_runner.py:113: command.append(os.path.join(paths.src_root,
Why not just:

command = ["python", command.append(...), ...]?

https://codereview.chromium.org/728783003/diff/60001/mojo/tools/test_runner.p...
mojo/tools/test_runner.py:121: command = [ "./" + gtest ]
...

[qsr, 2014-11-21 13:51:46]

Refactored test to separate multiprocess from non-multiprocess tests, and do not
compile and run multiprocess tests on android. 

I didn't update the file listing test. If you do not mind, I'd like to work on
this on a follow up CL where I will change this file to be a json file where we
will be able to annotate tests more easily, so that we can set the expected
platform, whether they can be cached or not, and anything else (for example, you
are speaking about data, we could be annotating tests with the data they need so
that the runner could be smarter).

https://codereview.chromium.org/728783003/diff/60001/base/test/multiprocess_t...
File base/test/multiprocess_test_android.cc (right):

https://codereview.chromium.org/728783003/diff/60001/base/test/multiprocess_t...
base/test/multiprocess_test_android.cc:1: // Copyright (c) 2012 The Chromium
Authors. All rights reserved.
On 2014/11/20 17:32:24, viettrungluu wrote:
> Changes to this file will get clobbered in the next Chromium -> Mojo update.

 This had already landed on Chromium (and in fact, already been pushed to our
own repo.)

https://codereview.chromium.org/728783003/diff/60001/mojo/tools/data/android_...
File mojo/tools/data/android_unittests (right):

https://codereview.chromium.org/728783003/diff/60001/mojo/tools/data/android_...
mojo/tools/data/android_unittests:1: # This file contains a list of Mojo gtest
unit tests.
On 2014/11/20 17:32:24, viettrungluu wrote:
> Having to maintain a separate list of tests is awful.

 Agreed. I was planning to update the format of this file to allow to specify
more information. Do you want this to be done immediately, or can this wait for
a follow up CL.

 I'd really like to go to the point where tests are run on android. Right now,
there is nothing.

https://codereview.chromium.org/728783003/diff/60001/mojo/tools/mopy/transiti...
File mojo/tools/mopy/transitive_hash.py (right):

https://codereview.chromium.org/728783003/diff/60001/mojo/tools/mopy/transiti...
mojo/tools/mopy/transitive_hash.py:64: def transitive_hash(filename,
transitive=True):
On 2014/11/20 17:32:24, viettrungluu wrote:
> Move _file_hash() into its own file and use that instead.

Done.

https://codereview.chromium.org/728783003/diff/60001/mojo/tools/test_runner.py
File mojo/tools/test_runner.py (right):

https://codereview.chromium.org/728783003/diff/60001/mojo/tools/test_runner.p...
mojo/tools/test_runner.py:112: command = [ "python" ]
On 2014/11/20 17:32:24, viettrungluu wrote:
> nit: no spaces inside [, ]

Done.

https://codereview.chromium.org/728783003/diff/60001/mojo/tools/test_runner.p...
mojo/tools/test_runner.py:113: command.append(os.path.join(paths.src_root,
On 2014/11/20 17:32:24, viettrungluu wrote:
> Why not just:
> 
> command = ["python", command.append(...), ...]?

Done.

https://codereview.chromium.org/728783003/diff/60001/mojo/tools/test_runner.p...
mojo/tools/test_runner.py:121: command = [ "./" + gtest ]
On 2014/11/20 17:32:24, viettrungluu wrote:
> ...

Done.

[viettrungluu, 2014-11-21 22:18:41]

https://codereview.chromium.org/728783003/diff/120001/mojo/edk/embedder/embed...
File mojo/edk/embedder/embedder_multiprocess_unittest.cc (right):

https://codereview.chromium.org/728783003/diff/120001/mojo/edk/embedder/embed...
mojo/edk/embedder/embedder_multiprocess_unittest.cc:1: // Copyright 2014 The
Chromium Authors. All rights reserved.
Can't you just disable individual tests (using suitable ifdefs[*]) instead?

[*] Adding DISABLED_ as usual, if that works; ifdef-ing out entire tests if it
doesn't.

https://codereview.chromium.org/728783003/diff/120001/mojo/edk/system/BUILD.gn
File mojo/edk/system/BUILD.gn (right):

https://codereview.chromium.org/728783003/diff/120001/mojo/edk/system/BUILD.g...
mojo/edk/system/BUILD.gn:139: "../test/multiprocess_test_helper_unittest.cc",
Are there compilation problems?

I'd rather not do this sort of thing.

https://codereview.chromium.org/728783003/diff/120001/mojo/tools/check_mojob.py
File mojo/tools/check_mojob.py (right):

https://codereview.chromium.org/728783003/diff/120001/mojo/tools/check_mojob....
mojo/tools/check_mojob.py:1: #!/usr/bin/env python
I don't think golden files are a good way to test mojob -- the cost (of
maintaining a million golden files) generally seems to outweigh the benefits.

[qsr, 2014-11-24 12:00:48]

https://codereview.chromium.org/728783003/diff/120001/mojo/edk/embedder/embed...
File mojo/edk/embedder/embedder_multiprocess_unittest.cc (right):

https://codereview.chromium.org/728783003/diff/120001/mojo/edk/embedder/embed...
mojo/edk/embedder/embedder_multiprocess_unittest.cc:1: // Copyright 2014 The
Chromium Authors. All rights reserved.
On 2014/11/21 22:18:41, viettrungluu wrote:
> Can't you just disable individual tests (using suitable ifdefs[*]) instead?
> 
> [*] Adding DISABLED_ as usual, if that works; ifdef-ing out entire tests if it
> doesn't.

 I can, and will as you seem to prefer it, but doing it this way makes it
impossible to compile a multi-process test on android, which prevent trying to
understand why it is failing (or even have it work, because it mostly does)

https://codereview.chromium.org/728783003/diff/120001/mojo/edk/system/BUILD.gn
File mojo/edk/system/BUILD.gn (right):

https://codereview.chromium.org/728783003/diff/120001/mojo/edk/system/BUILD.g...
mojo/edk/system/BUILD.gn:139: "../test/multiprocess_test_helper_unittest.cc",
On 2014/11/21 22:18:41, viettrungluu wrote:
> Are there compilation problems?
> 
> I'd rather not do this sort of thing.

Done.

https://codereview.chromium.org/728783003/diff/120001/mojo/tools/check_mojob.py
File mojo/tools/check_mojob.py (right):

https://codereview.chromium.org/728783003/diff/120001/mojo/tools/check_mojob....
mojo/tools/check_mojob.py:1: #!/usr/bin/env python
On 2014/11/21 22:18:41, viettrungluu wrote:
> I don't think golden files are a good way to test mojob -- the cost (of
> maintaining a million golden files) generally seems to outweigh the benefits.

 Do you have a suggestion? You asked me to test this, but all this is doing is
taking a bunch of argument and outputting a list of command. I do not really
know how I can test this any other way.

[viettrungluu, 2014-11-25 16:44:29]

On 2014/11/24 12:00:48, qsr wrote:
>
https://codereview.chromium.org/728783003/diff/120001/mojo/edk/embedder/embed...
> File mojo/edk/embedder/embedder_multiprocess_unittest.cc (right):
> 
>
https://codereview.chromium.org/728783003/diff/120001/mojo/edk/embedder/embed...
> mojo/edk/embedder/embedder_multiprocess_unittest.cc:1: // Copyright 2014 The
> Chromium Authors. All rights reserved.
> On 2014/11/21 22:18:41, viettrungluu wrote:
> > Can't you just disable individual tests (using suitable ifdefs[*]) instead?
> > 
> > [*] Adding DISABLED_ as usual, if that works; ifdef-ing out entire tests if
it
> > doesn't.
> 
>  I can, and will as you seem to prefer it, but doing it this way makes it
> impossible to compile a multi-process test on android, which prevent trying to
> understand why it is failing (or even have it work, because it mostly does)

Won't it still compile?

> 
> https://codereview.chromium.org/728783003/diff/120001/mojo/edk/system/BUILD.gn
> File mojo/edk/system/BUILD.gn (right):
> 
>
https://codereview.chromium.org/728783003/diff/120001/mojo/edk/system/BUILD.g...
> mojo/edk/system/BUILD.gn:139: "../test/multiprocess_test_helper_unittest.cc",
> On 2014/11/21 22:18:41, viettrungluu wrote:
> > Are there compilation problems?
> > 
> > I'd rather not do this sort of thing.
> 
> Done.
> 
>
https://codereview.chromium.org/728783003/diff/120001/mojo/tools/check_mojob.py
> File mojo/tools/check_mojob.py (right):
> 
>
https://codereview.chromium.org/728783003/diff/120001/mojo/tools/check_mojob....
> mojo/tools/check_mojob.py:1: #!/usr/bin/env python
> On 2014/11/21 22:18:41, viettrungluu wrote:
> > I don't think golden files are a good way to test mojob -- the cost (of
> > maintaining a million golden files) generally seems to outweigh the
benefits.
> 
>  Do you have a suggestion? You asked me to test this, but all this is doing is
> taking a bunch of argument and outputting a list of command. I do not really
> know how I can test this any other way.

Yeah, probably we should refactor it in a more sensible way, so let's put this
off for now. Sorry for the trouble.

[viettrungluu, 2014-11-25 16:45:09]

LGTM w/the (minor) comments below resolved.

https://codereview.chromium.org/728783003/diff/140001/mojo/edk/embedder/embed...
File mojo/edk/embedder/embedder_unittest.cc (right):

https://codereview.chromium.org/728783003/diff/140001/mojo/edk/embedder/embed...
mojo/edk/embedder/embedder_unittest.cc:286: // Android multi-process tests are
not executing the new process. This is flaky.
Maybe just say "On Android we can't exec(), so we don't support multiprocess
tests."?

(Would it be flaky? Or just not work? Is what I said accurate?)

Maybe also file a bug and reference it here. (We can dream of this being
resolved someday.)

https://codereview.chromium.org/728783003/diff/140001/mojo/edk/system/message...
File mojo/edk/system/message_pipe_perftest.cc (right):

https://codereview.chromium.org/728783003/diff/140001/mojo/edk/system/message...
mojo/edk/system/message_pipe_perftest.cc:144: // Android multi-process tests are
not executing the new process. This is flaky.
"

https://codereview.chromium.org/728783003/diff/140001/mojo/edk/system/multipr...
File mojo/edk/system/multiprocess_message_pipe_unittest.cc (right):

https://codereview.chromium.org/728783003/diff/140001/mojo/edk/system/multipr...
mojo/edk/system/multiprocess_message_pipe_unittest.cc:100: // Android
multi-process tests are not executing the new process. This is flaky.
"

https://codereview.chromium.org/728783003/diff/140001/mojo/edk/test/multiproc...
File mojo/edk/test/multiprocess_test_helper_unittest.cc (right):

https://codereview.chromium.org/728783003/diff/140001/mojo/edk/test/multiproc...
mojo/edk/test/multiprocess_test_helper_unittest.cc:46: // Android multi-process
tests are not executing the new process. This is flaky.
"

https://codereview.chromium.org/728783003/diff/140001/mojo/tools/mopy/file_ha...
File mojo/tools/mopy/file_hash.py (right):

https://codereview.chromium.org/728783003/diff/140001/mojo/tools/mopy/file_ha...
mojo/tools/mopy/file_hash.py:10: # pylint: enable=E0611
I think you can re-enable this after the hashlib import, no?

https://codereview.chromium.org/728783003/diff/140001/mojo/tools/mopy/transit...
File mojo/tools/mopy/transitive_hash.py (right):

https://codereview.chromium.org/728783003/diff/140001/mojo/tools/mopy/transit...
mojo/tools/mopy/transitive_hash.py:12: from mopy.file_hash import file_hash
I'd list the mopy imports separately from the standard library imports
(separated by a blank line).

https://codereview.chromium.org/728783003/diff/140001/mojo/tools/mopy/transit...
mojo/tools/mopy/transitive_hash.py:15: from os.path import basename, realpath
"

https://codereview.chromium.org/728783003/diff/140001/mojo/tools/test_runner.py
File mojo/tools/test_runner.py (right):

https://codereview.chromium.org/728783003/diff/140001/mojo/tools/test_runner....
mojo/tools/test_runner.py:18: from mopy.transitive_hash import file_hash
Can probably do ... import file_hash, transitive_hash on one line, no?

[qsr, 2014-11-25 17:17:50]

https://codereview.chromium.org/728783003/diff/140001/mojo/edk/embedder/embed...
File mojo/edk/embedder/embedder_unittest.cc (right):

https://codereview.chromium.org/728783003/diff/140001/mojo/edk/embedder/embed...
mojo/edk/embedder/embedder_unittest.cc:286: // Android multi-process tests are
not executing the new process. This is flaky.
On 2014/11/25 16:45:09, viettrungluu wrote:
> Maybe just say "On Android we can't exec(), so we don't support multiprocess
> tests."?
> 
> (Would it be flaky? Or just not work? Is what I said accurate?)
> 
> Maybe also file a bug and reference it here. (We can dream of this being
> resolved someday.)

 We can exec, but with the current infrastructure for tests, there is nothing to
exec. We don't have an executable. The native code is in a .so loaded by the
java side.

https://codereview.chromium.org/728783003/diff/140001/mojo/tools/mopy/file_ha...
File mojo/tools/mopy/file_hash.py (right):

https://codereview.chromium.org/728783003/diff/140001/mojo/tools/mopy/file_ha...
mojo/tools/mopy/file_hash.py:10: # pylint: enable=E0611
On 2014/11/25 16:45:09, viettrungluu wrote:
> I think you can re-enable this after the hashlib import, no?

Done.

https://codereview.chromium.org/728783003/diff/140001/mojo/tools/mopy/transit...
File mojo/tools/mopy/transitive_hash.py (right):

https://codereview.chromium.org/728783003/diff/140001/mojo/tools/mopy/transit...
mojo/tools/mopy/transitive_hash.py:12: from mopy.file_hash import file_hash
On 2014/11/25 16:45:09, viettrungluu wrote:
> I'd list the mopy imports separately from the standard library imports
> (separated by a blank line).

Done.

https://codereview.chromium.org/728783003/diff/140001/mojo/tools/mopy/transit...
mojo/tools/mopy/transitive_hash.py:15: from os.path import basename, realpath
On 2014/11/25 16:45:09, viettrungluu wrote:
> "

Done.

https://codereview.chromium.org/728783003/diff/140001/mojo/tools/test_runner.py
File mojo/tools/test_runner.py (right):

https://codereview.chromium.org/728783003/diff/140001/mojo/tools/test_runner....
mojo/tools/test_runner.py:18: from mopy.transitive_hash import file_hash
On 2014/11/25 16:45:09, viettrungluu wrote:
> Can probably do ... import file_hash, transitive_hash on one line, no?

Done.

[qsr, 2014-11-26 09:54:44]

Committed patchset #9 (id:160001) manually as
f4583a88c6184faac8eb4035b5e8948b2e3b3240 (presubmit successful).