Link binaries targeting iOS simulator to the appropriate ASan dynamic runtime.

Link binaries targeting iOS simulator to the appropriate ASan dynamic runtime.

BUG=344836, 235466
R=thakis@chromium.org

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=262372

[Alexander Potapenko, 2014-04-01 18:41:32]

PTAL
This should fix the runhooks step on the iossim ASan buildbot (see issue 235466)

[Nico, 2014-04-02 02:46:04]

https://codereview.chromium.org/218613015/diff/1/build/mac/asan.gyp
File build/mac/asan.gyp (right):

https://codereview.chromium.org/218613015/diff/1/build/mac/asan.gyp#newcode22
build/mac/asan.gyp:22: 'postbuilds': [
Why does mac require a postbuild now?

[Alexander Potapenko, 2014-04-02 14:58:31]

On 2014/04/02 02:46:04, Nico wrote:
> https://codereview.chromium.org/218613015/diff/1/build/mac/asan.gyp
> File build/mac/asan.gyp (right):
> 
> https://codereview.chromium.org/218613015/diff/1/build/mac/asan.gyp#newcode22
> build/mac/asan.gyp:22: 'postbuilds': [
> Why does mac require a postbuild now?

This is needed to run install_name_tool when building Chromium or tests, not
when packaging Clang (see
https://code.google.com/p/chromium/issues/detail?id=344836)
I've updated the CL description.

[Nico, 2014-04-02 15:07:52]

I'm somewhat confused. The code looks like something you'd want to inject into
every dependent target, but it's not injected everywhere, it's just in a regular
target that everything depends on.

I'm probably terribly misreading this CL :-/

https://codereview.chromium.org/218613015/diff/20001/build/mac/asan.gyp
File build/mac/asan.gyp (right):

https://codereview.chromium.org/218613015/diff/20001/build/mac/asan.gyp#newco...
build/mac/asan.gyp:27: # performs proper relativization during dict merging.
Is this necessary? asan_dynamic_runtime is a regular target in a gyp file,
nothing in a gypi, so all paths in it should be relative to the folder
containing the gyp file (build/mac). So if you use the correct relative path,
you shouldn't need a variable.

https://codereview.chromium.org/218613015/diff/20001/build/mac/asan.gyp#newco...
build/mac/asan.gyp:29: 'mac/copy_asan_runtime_dylib.sh',
This script checks if "${BUILT_PRODUCTS_DIR}/${EXECUTABLE_PATH}" exists and
exits if not. Since asan_dynamic_runtime is a type none target, this path will
never exit, and the postbuild won't do anything, as far as I understand.

What am I missing?

[Alexander Potapenko, 2014-04-02 15:35:42]

I may be misunderstanding how relative paths work.
Anyway, let us drop the postbuild action which isn't needed to fix the iOS bot.
I'll go through the install_name_tool stuff again and make a proper fix in a
separate CL.

[Alexander Potapenko, 2014-04-03 08:51:17]

Removed the postbuild and reverted testing/gtest.gyp, PTAL

[Alexander Potapenko, 2014-04-05 12:10:22]

On 2014/04/03 08:51:17, Alexander Potapenko wrote:
> Removed the postbuild and reverted testing/gtest.gyp, PTAL

Ping?

[Nico, 2014-04-05 18:40:18]

lgtm, sorry.

Thanks for splitting this up.

https://codereview.chromium.org/218613015/diff/60001/build/mac/asan.gyp
File build/mac/asan.gyp (right):

https://codereview.chromium.org/218613015/diff/60001/build/mac/asan.gyp#newco...
build/mac/asan.gyp:28: '<!(/bin/ls <(asan_osx_dynamic))',
Out of curiosity, why do you need the call to ls?

https://codereview.chromium.org/218613015/diff/60001/build/mac/asan.gyp#newco...
build/mac/asan.gyp:36: 'target_conditions': [
nit: I think you can move target_conditions inside the copies block, like

  'copies': [
    {
      'destination': '<(PRODUCT_DIR)'
      'target_conditions': [
        ['_toolset_=="host";, { 'files': [ 'asan_osx_dynamic'] }],
        ['_toolset_=="target";, { 'files': [ 'asan_iossim_dynamic'] }],
      ]
    }
  ]

Haven't tried it, though.

[Alexander Potapenko, 2014-04-08 08:31:10]

https://codereview.chromium.org/218613015/diff/60001/build/mac/asan.gyp
File build/mac/asan.gyp (right):

https://codereview.chromium.org/218613015/diff/60001/build/mac/asan.gyp#newco...
build/mac/asan.gyp:28: '<!(/bin/ls <(asan_osx_dynamic))',
On 2014/04/05 18:40:18, Nico wrote:
> Out of curiosity, why do you need the call to ls?

IIUC 'copies' doesn't support wildcards, does it?

https://codereview.chromium.org/218613015/diff/60001/build/mac/asan.gyp#newco...
build/mac/asan.gyp:36: 'target_conditions': [
On 2014/04/05 18:40:18, Nico wrote:
> nit: I think you can move target_conditions inside the copies block, like
> 
>   'copies': [
>     {
>       'destination': '<(PRODUCT_DIR)'
>       'target_conditions': [
>         ['_toolset_=="host";, { 'files': [ 'asan_osx_dynamic'] }],
>         ['_toolset_=="target";, { 'files': [ 'asan_iossim_dynamic'] }],
>       ]
>     }
>   ]
> 
> Haven't tried it, though.

Done.

[Alexander Potapenko, 2014-04-08 08:33:12]

The CQ bit was checked by glider@chromium.org

[commit-bot: I haz the power, 2014-04-08 08:33:22]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/glider@chromium.org/218613015/70001

[commit-bot: I haz the power, 2014-04-08 09:08:07]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2014-04-08 09:08:08]

Try jobs failed on following builders:
  tryserver.chromium on mac_chromium_rel

[Alexander Potapenko, 2014-04-08 09:18:59]

The CQ bit was checked by glider@chromium.org

[commit-bot: I haz the power, 2014-04-08 09:19:19]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/glider@chromium.org/218613015/70001

[commit-bot: I haz the power, 2014-04-08 11:12:00]

Change committed as 262372