android: Add 4 new linker tests.

android: Add 4 new linker tests.

This extracts the build/android/ changes for Patch Set 32 of
https://codereview.chromium.org/23717023/

This adds 4 new linker tests to check the library load addresses
and their proper randomization when simulating low-memory and
regular devices.

Since there are now 6 linker tests, also add support for
 -f / --gtest_filter options to filter the tests to run,
e.g. to run all low-memory device tests, one can use:

  build/android/test_runner.py linker --gtest_filter='*LowMemory*'

BUG=287739
R=bulach@chromium.org, frankf@chromium.org, yfriedman@chromium.org

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=227418

[digit1, 2013-10-07 13:50:24]

By the way, I've removed the use of FlagChanger, because it failed _silently_ on
a production device to modify any file. I've filed
https://code.google.com/p/chromium/issues/detail?id=304832 to address this
separately.

[bulach, 2013-10-07 16:49:44]

lgtm, just small nits and suggestions

https://codereview.chromium.org/26251003/diff/1/build/android/pylib/linker/se...
File build/android/pylib/linker/setup.py (right):

https://codereview.chromium.org/26251003/diff/1/build/android/pylib/linker/se...
build/android/pylib/linker/setup.py:33:
test_case.LinkerRandomizationTest(is_low_memory=False),
probably less readable, but:
test_cases = [
  test_case.LinkerLibraryAddressTest,
  test_case.LinkerSharedRelroTest,
  test_case.LinkerRandomizationTest
] 
low_memory = [False, True]
all_tests = [
    lambda x,y:x(is_low_memory=y) for x, y in
    zip(test_cases * len(low_memory), low_memory * len(test_cases)
]

https://codereview.chromium.org/26251003/diff/1/build/android/pylib/linker/te...
File build/android/pylib/linker/test_case.py (right):

https://codereview.chromium.org/26251003/diff/1/build/android/pylib/linker/te...
build/android/pylib/linker/test_case.py:73: not work at all with production
devices."""
would this work on directory that's not writable by the shell user? if it does,
would it make sense to make FlagChanger to do the same, and use FlagChanger
instead? :)
is the problem the lack of "su" or the lack of permissions?

https://codereview.chromium.org/26251003/diff/1/build/android/pylib/linker/te...
build/android/pylib/linker/test_case.py:231: logging.info('%s library load map:
%s' % (process_type, lib_addr_map))
nit: logging can take a format with vararg, so rather than % (tuple0, tuple1),
just use , tuple0, tuple1 directly..

https://codereview.chromium.org/26251003/diff/1/build/android/pylib/linker/te...
build/android/pylib/linker/test_case.py:377: logging.info('Renderer libraries:
%s' % renderer_libs)
nit: as above, prefer , rather than % and some places like 383 below..

[digit1, 2013-10-07 20:31:58]

https://codereview.chromium.org/26251003/diff/1/build/android/pylib/linker/se...
File build/android/pylib/linker/setup.py (right):

https://codereview.chromium.org/26251003/diff/1/build/android/pylib/linker/se...
build/android/pylib/linker/setup.py:33:
test_case.LinkerRandomizationTest(is_low_memory=False),
I see, good idea, but I think I've found something a little simpler though:
  all_tests = [ t(is_low_memory=m) for t in test_cases for m in [True, False]]

https://codereview.chromium.org/26251003/diff/1/build/android/pylib/linker/te...
File build/android/pylib/linker/test_case.py (right):

https://codereview.chromium.org/26251003/diff/1/build/android/pylib/linker/te...
build/android/pylib/linker/test_case.py:73: not work at all with production
devices."""
No, the main issue is that FlaChanger really runs "su" directly, which doesn't
exit on production devices.
I didn't want to modify FlagChanger in this patch, because fixing this correctly
is not completely trivial (also why I filed a bug about this).

[digit1, 2013-10-07 20:58:21]

the clang builder is out of disk space. I'm going to dcommit this manually.

[craigdh, 2013-10-07 21:05:54]

On 2013/10/07 20:58:21, digit1 wrote:
> the clang builder is out of disk space. I'm going to dcommit this manually.

I'll have a patch up in the next hour to fix the issues with flag_changer,
please wait for that.

[commit-bot: I haz the power, 2013-10-07 21:18:31]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/digit@chromium.org/26251003/7001

[digit1, 2013-10-07 21:18:36]

On 2013/10/07 21:05:54, craigdh wrote:
> On 2013/10/07 20:58:21, digit1 wrote:
> > the clang builder is out of disk space. I'm going to dcommit this manually.
> 
> I'll have a patch up in the next hour to fix the issues with flag_changer,
> please wait for that.

Actually, I'd rather not. This is part of a series of patches that we'd wish to
merge into M31 for Android. It's already late and any extra dependency like that
would be a drag.

[digit1, 2013-10-07 21:18:37]

On 2013/10/07 21:05:54, craigdh wrote:
> On 2013/10/07 20:58:21, digit1 wrote:
> > the clang builder is out of disk space. I'm going to dcommit this manually.
> 
> I'll have a patch up in the next hour to fix the issues with flag_changer,
> please wait for that.

Actually, I'd rather not. This is part of a series of patches that we'd wish to
merge into M31 for Android. It's already late and any extra dependency like that
would be a drag.

[digit1, 2013-10-07 21:19:50]

Looks like I can't dcommit when connected remotely (something to do with
credentials / gnome-keyring  / whatever). So I've push the commit button here
instead. Let's wait.

[commit-bot: I haz the power, 2013-10-08 01:19:10]

Change committed as 227418