Roll Clang 233105-2:236765-2

Roll Clang 233105-2:236765-2

(Abandoned due to a bug found by the trybots)

Note to sheriffs: we're expecting a ~400 kB size regression due to
a bug fix that causes more loop unrolling in this Clang version.

BUG=483026
NOPRESUBMIT=true

[hans, 2015-05-14 21:09:03]

hans@chromium.org changed reviewers:
+ thakis@chromium.org

[hans, 2015-05-14 21:09:03]

Please take a look.

[Nico, 2015-05-14 21:11:09]

you probably need to do the .gyp change in some .gn file too

https://codereview.chromium.org/1135023004/diff/1/build/common.gypi
File build/common.gypi (right):

https://codereview.chromium.org/1135023004/diff/1/build/common.gypi#newcode3797
build/common.gypi:3797: # less effective.
Maybe link to the thread or put in a TODO to remove this if the upstream thread
gets resolved with a revert or something

https://codereview.chromium.org/1135023004/diff/1/tools/clang/blink_gc_plugin...
File tools/clang/blink_gc_plugin/JsonWriter.h (left):

https://codereview.chromium.org/1135023004/diff/1/tools/clang/blink_gc_plugin...
tools/clang/blink_gc_plugin/JsonWriter.h:14: #define JSON_WRITER_STREAM
llvm::raw_fd_ostream*
\o/

[hans, 2015-05-14 21:25:20]

New patch uploaded.

> you probably need to do the .gyp change in some .gn file too

Thanks! Done.

https://codereview.chromium.org/1135023004/diff/1/build/common.gypi
File build/common.gypi (right):

https://codereview.chromium.org/1135023004/diff/1/build/common.gypi#newcode3797
build/common.gypi:3797: # less effective.
On 2015/05/14 21:11:09, Nico wrote:
> Maybe link to the thread or put in a TODO to remove this if the upstream
thread
> gets resolved with a revert or something

Done.

[Nico, 2015-05-14 21:31:22]

lgtm

[hans, 2015-05-15 22:07:39]

Uploaded new attempt that doesn't build the ubsan run-time on Mac.

Will start tryjobs as soon as goma is ready.

[Nico, 2015-05-15 22:09:00]

diff lgtm

[hans, 2015-05-15 22:10:05]

Adding NOPRESUBMIT=true due to long lines in update.sh.

[hans, 2015-05-15 23:49:19]

Tweaking the gn flags to fix NaCl build.

[hans, 2015-05-16 19:18:38]

hans@chromium.org changed reviewers:
+ brettw@chromium.org

[hans, 2015-05-16 19:18:39]

brettw: can you take an OWNER's look at the .gn file?

[brettw, 2015-05-17 00:11:32]

In GYP this change only affects release builds, and in GN it affects all. If you
want it to be release-only, add as !is_debug check to the condition in the GN
build. If you need it in debug only, find a shared place in the file for the
condition in gyp.

[hans, 2015-05-18 16:06:28]

> In GYP this change only affects release builds, and in GN it affects all. If
you
want it to be release-only, add as !is_debug check to the condition in the GN
build. If you need it in debug only, find a shared place in the file for the
condition in gyp.

Thanks for catching this! New patch uploaded, please take a look.

[brettw, 2015-05-18 19:09:55]

lgtm

[hans, 2015-05-18 20:00:35]

Thanks!

The only remaining issue is linux_chromium_chromeos_dbg_ng. It seems to fail
fairly consistently with this patch, but not with a whitespace change
(https://codereview.chromium.org/1131653006).

Something is dodgy though, as it fails in the "without patch" step as well.

I tried the test locally on Linux, and it fails as on the bot, both with and
without the roll. That, and two open bugs about this test (crbug.com/345666 and
crbug.com/346278) makes me think the problem is with the test, not the roll.

[hans, 2015-05-18 20:00:53]

The CQ bit was checked by hans@chromium.org

[hans, 2015-05-18 20:00:53]

The patchset sent to the CQ was uploaded after l-g-t-m from thakis@chromium.org
Link to the patchset: https://codereview.chromium.org/1135023004/#ps70001
(title: "gn: pass -funique-section-names in release builds only")

[commit-bot: I haz the power, 2015-05-18 20:08:21]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1135023004/70001

[commit-bot: I haz the power, 2015-05-18 20:57:12]

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

[commit-bot: I haz the power, 2015-05-18 20:57:14]

Try jobs failed on following builders:
  ios_rel_device_ninja on tryserver.chromium.mac (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.mac/builders/ios_rel_device_ni...)

[hans, 2015-05-18 20:59:54]

On 2015/05/18 20:57:14, I haz the power (commit-bot) wrote:
> Try jobs failed on following builders:
>   ios_rel_device_ninja on tryserver.chromium.mac (JOB_FAILED,
>
http://build.chromium.org/p/tryserver.chromium.mac/builders/ios_rel_device_ni...)

:-( That's http://llvm.org/PR23486, which is supposedly fixed in r237211. I
guess we need a newer Clang.