Set the default visibility for __asan_default_options.

Set the default visibility for __asan_default_options.

BUG=368351
R=thakis@chromium.org

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=268161

[Alexander Potapenko, 2014-04-30 13:35:40]

PTAL

[Nico, 2014-04-30 15:31:27]

"Chromium-packaged Clang doesn't contain the .syms file" on the bug suggests
that this is no longer needed once we include these, is that right? If so,
change the TODO text to "Remove after the next clang roll, crbug.com/$number"
(and prepare a CL for bundling the syms files, I assume they're small.)

[Nico, 2014-04-30 15:31:33]

…and lgtm

[Alexander Potapenko, 2014-04-30 15:56:20]

On 2014/04/30 15:31:27, Nico wrote:
> "Chromium-packaged Clang doesn't contain the .syms file" on the bug suggests
> that this is no longer needed once we include these, is that right? If so,
> change the TODO text to "Remove after the next clang roll, crbug.com/$number"
> (and prepare a CL for bundling the syms files, I assume they're small.)

No, that's worse.
Right now the Chromium-packaged Clang works because it doesn't contain the .syms
file, and any other Clang does not.
But having the .syms file is a right thing to do (it reduces the size of the
symbol table) despite it'll break the Chromium-packaged Clang (I think we anyway
want to allow the users use their own Clang).
This particular hack can be removed only after the problem is fixed in binutils
and Chromium switches to the appropriate gold version.

[Alexander Potapenko, 2014-04-30 15:57:45]

Added a TODO, committing.

[Alexander Potapenko, 2014-04-30 16:02:28]

The CQ bit was checked by glider@chromium.org

[commit-bot: I haz the power, 2014-04-30 16:02:57]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/glider@chromium.org/254293002/20001

[Nico, 2014-04-30 16:03:57]

On 2014/04/30 15:56:20, Alexander Potapenko wrote:
> On 2014/04/30 15:31:27, Nico wrote:
> > "Chromium-packaged Clang doesn't contain the .syms file" on the bug suggests
> > that this is no longer needed once we include these, is that right? If so,
> > change the TODO text to "Remove after the next clang roll,
crbug.com/$number"
> > (and prepare a CL for bundling the syms files, I assume they're small.)
> 
> No, that's worse.
> Right now the Chromium-packaged Clang works because it doesn't contain the
.syms
> file, and any other Clang does not.
> But having the .syms file is a right thing to do (it reduces the size of the
> symbol table) despite it'll break the Chromium-packaged Clang 

Why does it break our clang?

> (I think we anyway
> want to allow the users use their own Clang

(No, we don't want to support that.)

> This particular hack can be removed only after the problem is fixed in
binutils
> and Chromium switches to the appropriate gold version.

[commit-bot: I haz the power, 2014-04-30 16:06:00]

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

[commit-bot: I haz the power, 2014-04-30 16:06:00]

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

[Alexander Potapenko, 2014-04-30 17:07:48]

Nico, can you please take another look?
Regardless of whether there's an error in handling visibility=hidden and dynamic
lists (I've sent a letter to the gold maintainer to clarify this), the problem
can be solved by setting visibility=default for __asan_default_options().
Sorry I haven't tried that before.

[Nico, 2014-04-30 17:25:06]

lgtm

That's better, thanks!

…actually, I think we have some build step on Mac that checks the list of
exported symbols; you might have to disable that step for asan. (Maybe it
already doesn't run for asan?)

[Alexander Potapenko, 2014-04-30 17:33:17]

On 2014/04/30 17:25:06, Nico wrote:
> lgtm
> 
> That's better, thanks!
I've also added __attribute__(used) like Sergey recently did for the very same
__asan_default_options in NaCl.

> …actually, I think we have some build step on Mac that checks the list of
> exported symbols; you might have to disable that step for asan. (Maybe it
> already doesn't run for asan?)

IIRC it does not, or at least it's horribly broken.
There's been asan.saves that was passed in xcode_settings, but I haven't updated
it since the switch to Ninja happened.

[Alexander Potapenko, 2014-04-30 17:36:51]

The CQ bit was checked by glider@chromium.org

[commit-bot: I haz the power, 2014-04-30 17:37:30]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/glider@chromium.org/254293002/60001

[commit-bot: I haz the power, 2014-04-30 17:45:54]

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

[commit-bot: I haz the power, 2014-04-30 17:45:54]

Try jobs failed on following builders:
  chromium_presubmit on tryserver.chromium
  linux_chromium_rel on tryserver.chromium

[Alexander Potapenko, 2014-04-30 18:03:19]

The CQ bit was checked by glider@chromium.org

[commit-bot: I haz the power, 2014-04-30 18:05:34]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/glider@chromium.org/254293002/60001

[commit-bot: I haz the power, 2014-04-30 18:12:59]

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

[commit-bot: I haz the power, 2014-04-30 18:13:00]

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

[Alexander Potapenko, 2014-05-05 08:01:42]

The CQ bit was checked by glider@chromium.org

[commit-bot: I haz the power, 2014-05-05 08:01:46]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/glider@chromium.org/254293002/60001

[commit-bot: I haz the power, 2014-05-05 13:12:52]

Change committed as 268161