Move the call to CRYPTO_set_NEON_capable up.

Move the call to CRYPTO_set_NEON_capable up.

BoringSSL needs to probe for NEON support via SIGILL in cases where
getauxval isn't provided and the application doesn't do explicit
initialisation.

However, Chromium might have gone multithreaded by the time that we
initialise BoringSSL and, although it doesn't look like we'll race the
disposition of SIGILL with anything, it's best not to test that hope.

So this change causes CRYPTO_set_NEON_capable to always be called, and
to be called before SSL_library_init. BoringSSL will take that as a
signal that probing for NEON support isn't needed.

BUG=none

Committed: https://crrev.com/fa9063829e638aeb72ede79e5d0396a81a0211d1
Cr-Commit-Position: refs/heads/master@{#314201}

[agl, 2015-02-02 19:20:20]

agl@chromium.org changed reviewers:
+ davidben@chromium.org

[davidben, 2015-02-02 19:47:16]

lgtm

https://codereview.chromium.org/899463002/diff/1/crypto/openssl_util.cc
File crypto/openssl_util.cc (right):

https://codereview.chromium.org/899463002/diff/1/crypto/openssl_util.cc#newco...
crypto/openssl_util.cc:56: // that getauxval isn't present.
Fun fact: this also avoids a deadlock were CRYPTO_set_NEON_functional to take a
lock. I tried to reorder the locking callback functions and we actually assume
that nothing in this ctor calls the locking callbacks, because those go back and
query the singleton. (We should probably fix this at some point...)

[agl, 2015-02-02 19:51:45]

The CQ bit was checked by agl@chromium.org

[commit-bot: I haz the power, 2015-02-02 19:52:38]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/899463002/1

[commit-bot: I haz the power, 2015-02-02 22:31:40]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2015-02-02 22:32:54]

Patchset 1 (id:??) landed as
https://crrev.com/fa9063829e638aeb72ede79e5d0396a81a0211d1
Cr-Commit-Position: refs/heads/master@{#314201}

[Noel Gordon, 2015-02-03 09:39:09]

Various encrypted video media test began failing on ChromeOS Test (1), ChromeOS
Test (1) Debug, example

http://build.chromium.org/p/chromium.webkit/builders/Linux%20ChromiumOS%20Tes...

after this patch landed ...

[Noel Gordon, 2015-02-03 09:49:10]

Also on
http://build.chromium.org/p/chromium.webkit/builders/Mac10.8%20Tests/builds/1...

Will attempt a revert for now to see if the tests come green.

[Noel Gordon, 2015-02-03 09:54:25]

A revert of this CL (patchset #1 id:1) has been created in
https://codereview.chromium.org/895013002/ by noel@chromium.org.

The reason for reverting is: Revert: suspect this change upset the webkit
ChromeOS Test 1, Mac 10.8, etc encrypted media tests. Reverting to see if these
builders turn green.

http://build.chromium.org/p/chromium.webkit/builders/Linux%20ChromiumOS%20Tes....

[davidben, 2015-02-03 18:22:06]

On 2015/02/03 09:54:25, noel gordon wrote:
> A revert of this CL (patchset #1 id:1) has been created in
> https://codereview.chromium.org/895013002/ by mailto:noel@chromium.org.
> 
> The reason for reverting is: Revert: suspect this change upset the webkit
> ChromeOS Test 1, Mac 10.8, etc encrypted media tests. Reverting to see if
these
> builders turn green.
> 
>
http://build.chromium.org/p/chromium.webkit/builders/Linux%20ChromiumOS%20Tes....

Did that fix it? As far as I'm aware, this code shouldn't even be running on
ChromeOS...

[agl, 2015-02-03 18:40:56]

On Tue, Feb 3, 2015 at 10:22 AM,  <davidben@chromium.org> wrote:
> Did that fix it? As far as I'm aware, this code shouldn't even be running on
> ChromeOS...

No, the builder is still broken. It was a fair suspicion, but that
breakage wasn't my fault at least :)

Will reland today.


Cheers

AGL

To unsubscribe from this group and stop receiving emails from it, send an email
to chromium-reviews+unsubscribe@chromium.org.

[Noel Gordon, 2015-02-04 00:08:48]

On 2015/02/03 18:40:56, agl wrote:
> On Tue, Feb 3, 2015 at 10:22 AM,  <mailto:davidben@chromium.org> wrote:
> > Did that fix it? As far as I'm aware, this code shouldn't even be running on
> > ChromeOS...
>
> No, the builder is still broken. It was a fair suspicion, but that
> breakage wasn't my fault at least :)

Agree, and apologies for the revert.

> Will reland today.

SGTM.