Remove outdated __GLIBCXX_ checks in base::StackTrace for POSIX.

Remove outdated __GLIBCXX_ checks in base::StackTrace for POSIX.

This fixes symbol name unmangling under Mac OS X, where we have switched
from building against libstdc++, which defines __GLIBCXX__, to libc++,
which does not. All we actually care about is that the C++ library
provides the abi::__cxa_demangle() API, which both libraries do.

BUG=598886, 596760
Review-Url: https://codereview.chromium.org/2651893002
Cr-Commit-Position: refs/heads/master@{#446398}
Committed: https://chromium.googlesource.com/chromium/src/+/bf51bcfadc991bbf00161d8a28fa7fdad74b8222

[Wez, 2017-01-24 05:30:05]

The CQ bit was checked by wez@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2017-01-24 05:30:16]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[Wez, 2017-01-24 05:40:02]

Description was changed from

==========
Remove outdated __GLIBCXX_ checks.

BUG=
==========

to

==========
Remove outdated __GLIBCXX_ checks in base::StackTrace for POSIX.

Under Mac OS X we use the POSIX StackTrace implementation, and build it to use
its own name de-mangler rather than the Symbolize library that we use under
Linux.

The StackTrace de-mangler was built conditional on __GLIBCXX__, to check for the
necessary demangling support in the ABI, but recent Clang versions do not define
__GLIBCXX__ any more.

BUG=598886
==========

[Wez, 2017-01-24 05:40:02]

wez@chromium.org changed reviewers:
+ esprehn@chromium.org

[commit-bot: I haz the power, 2017-01-24 06:32:18]

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

[commit-bot: I haz the power, 2017-01-24 06:32:19]

Dry run: This issue passed the CQ dry run.

[Wez, 2017-01-24 07:53:24]

wez@chromium.org changed reviewers:
+ thakis@chromium.org
- esprehn@chromium.org

[Wez, 2017-01-24 07:53:24]

Nico, do you know if there's any remaining reason for USE_SYMBOLIZE? Could we
switch things to be purely conditional on the target platform (i.e. Linux
(Symbolize) vs Mac OS vs Android (no demangling))?

[Nico, 2017-01-25 15:51:27]

> recent Clang versions do not define __GLIBCXX__

clang doesn't define __GLIBCXX__; that define identifies libstdc++ and is
defined by its headers. It's not defined when libc++ is used instead, but that's
independent of the compiler. libc++ also has a __cxa_demangle, but you're right
that those are the only c++ libs we use on posix, so removing that check lgtm
(with fixed CL description).

(When statically linking libc++(abi), pulling in __cxa_demangle adds quite a bit
of binary size, so we might want to consider something like
https://codereview.chromium.org/1215673002 -- at the moment we don't link libc++
statically, but we will on linux eventually, and maybe on mac again.)

USE_SYMBOLIZE is from https://codereview.chromium.org/545148/

[Wez, 2017-01-26 17:29:12]

Description was changed from

==========
Remove outdated __GLIBCXX_ checks in base::StackTrace for POSIX.

Under Mac OS X we use the POSIX StackTrace implementation, and build it to use
its own name de-mangler rather than the Symbolize library that we use under
Linux.

The StackTrace de-mangler was built conditional on __GLIBCXX__, to check for the
necessary demangling support in the ABI, but recent Clang versions do not define
__GLIBCXX__ any more.

BUG=598886
==========

to

==========
Remove outdated __GLIBCXX_ checks in base::StackTrace for POSIX.

This fixes symbol name unmangling under Mac OS X, where we have switched from
building against libstdc++, which defines __GLIBCXX__, to libc++, which does
not. All we actually care about is that the C++ library provides the
abi::__cxa_demangle() API, which both libraries do.

BUG=598886,596760
==========

[Wez, 2017-01-26 17:51:40]

On 2017/01/25 15:51:27, Nico wrote:
> > recent Clang versions do not define __GLIBCXX__
> 
> clang doesn't define __GLIBCXX__; that define identifies libstdc++ and is
> defined by its headers. It's not defined when libc++ is used instead, but
that's
> independent of the compiler. libc++ also has a __cxa_demangle, but you're
right
> that those are the only c++ libs we use on posix, so removing that check lgtm
> (with fixed CL description).
> 
> (When statically linking libc++(abi), pulling in __cxa_demangle adds quite a
bit
> of binary size, so we might want to consider something like
> https://codereview.chromium.org/1215673002 -- at the moment we don't link
libc++
> statically, but we will on linux eventually, and maybe on mac again.)
> 
> USE_SYMBOLIZE is from https://codereview.chromium.org/545148/

Thanks for the correction & context, Nico.  I've update the description but
tried to keep it brief & to the point re lib[std]c++ ABI API. :)

[Wez, 2017-01-26 17:51:47]

The CQ bit was checked by wez@chromium.org

[commit-bot: I haz the power, 2017-01-26 17:52:21]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[Nico, 2017-01-26 17:58:37]

Thanks, new CL description looks great. I'd wrap it at 80ish cols though, those
long lines always look a bit strange in `git log`

[Wez, 2017-01-26 18:04:59]

Description was changed from

==========
Remove outdated __GLIBCXX_ checks in base::StackTrace for POSIX.

This fixes symbol name unmangling under Mac OS X, where we have switched from
building against libstdc++, which defines __GLIBCXX__, to libc++, which does
not. All we actually care about is that the C++ library provides the
abi::__cxa_demangle() API, which both libraries do.

BUG=598886,596760
==========

to

==========
Remove outdated __GLIBCXX_ checks in base::StackTrace for POSIX.

This fixes symbol name unmangling under Mac OS X, where we have switched
from building against libstdc++, which defines __GLIBCXX__, to libc++,
which does not. All we actually care about is that the C++ library
provides the abi::__cxa_demangle() API, which both libraries do.

BUG=598886,596760
==========

[Wez, 2017-01-26 18:05:44]

On 2017/01/26 17:58:37, Nico wrote:
> Thanks, new CL description looks great. I'd wrap it at 80ish cols though,
those
> long lines always look a bit strange in `git log`

Done; for some reason I thought CQ already did some wrapping, but I must have
imagined it...

[commit-bot: I haz the power, 2017-01-26 19:40:29]

CQ is committing da patch.

Bot data: {"patchset_id": 1, "attempt_start_ts": 1485453107185150, "parent_rev":
"873a0b98cbf3a3e6aa5f2b0cc2f1138257fdf21b", "commit_rev":
"bf51bcfadc991bbf00161d8a28fa7fdad74b8222"}

[commit-bot: I haz the power, 2017-01-26 19:41:02]

Description was changed from

==========
Remove outdated __GLIBCXX_ checks in base::StackTrace for POSIX.

This fixes symbol name unmangling under Mac OS X, where we have switched
from building against libstdc++, which defines __GLIBCXX__, to libc++,
which does not. All we actually care about is that the C++ library
provides the abi::__cxa_demangle() API, which both libraries do.

BUG=598886,596760
==========

to

==========
Remove outdated __GLIBCXX_ checks in base::StackTrace for POSIX.

This fixes symbol name unmangling under Mac OS X, where we have switched
from building against libstdc++, which defines __GLIBCXX__, to libc++,
which does not. All we actually care about is that the C++ library
provides the abi::__cxa_demangle() API, which both libraries do.

BUG=598886,596760

Review-Url: https://codereview.chromium.org/2651893002
Cr-Commit-Position: refs/heads/master@{#446398}
Committed:
https://chromium.googlesource.com/chromium/src/+/bf51bcfadc991bbf00161d8a28fa...
==========

[commit-bot: I haz the power, 2017-01-26 19:41:03]

Committed patchset #1 (id:1) as
https://chromium.googlesource.com/chromium/src/+/bf51bcfadc991bbf00161d8a28fa...