Workaround removal of __system_property_get in Android NDK.

Workaround removal of __system_property_get in Android NDK.

The 'L' release of Android will remove __system_property_get from the NDK.
Until we can replace all calls of __system_property_get, this CL adds a
temporary work-around for builds compiled against the 'L' API.  The workaround
involves creating a stub __system_property_get, which uses dlsym to dynamically
load the address of the real __system_property_get symbol in libc.so (which is
still available as a hidden symbol on the devices libc.so dispite having been
removed from the NDK).

BUG=392191, 393903
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=284012

[rmcilroy, 2014-07-15 17:49:05]

Nico, please take a look, thanks.

[Nico, 2014-07-16 19:57:33]

How much work is it to remove all calls? This here looks pretty iffy.

[rmcilroy, 2014-07-16 20:32:07]

On 2014/07/16 19:57:33, Nico (away) wrote:
> How much work is it to remove all calls? This here looks pretty iffy.

Quite a lot of work - see crbug.com/392196 and crbug.com/392201 for two of the
places this will need updated and don't have easy solutions.  We would need to
update base and breakpad as well.  We should do this eventually, however we
won't be able to do it all for M37, which is where we will need the changes,
hence this temporary work-around.

[Nico, 2014-07-17 16:55:50]

https://code.google.com/p/chromium/codesearch#search/&q=__system_property_get...
<- are these all hits? That doesn't look like that many.

If each of these calls require a lot of work to remove, can you make a
spreadsheet that lists for each of these calls:
* The intended replacement api
* An owner to do the replacement by the post-l mstone
?

https://codereview.chromium.org/393923002/diff/20001/base/sys_info_android.cc
File base/sys_info_android.cc (right):

https://codereview.chromium.org/393923002/diff/20001/base/sys_info_android.cc...
base/sys_info_android.cc:27: static int (*__real_system_property_get)(const
char*, char*) = NULL;
Is this always called from the same thread? (I think we build with
-fno-threadsafe-statics.)

[rmcilroy, 2014-07-17 18:41:52]

>
https://code.google.com/p/chromium/codesearch#search/&q=__system_property_get...
> <- are these all hits? That doesn't look like that many.
> 
> If each of these calls require a lot of work to remove,  can you make a
> spreadsheet that lists for each of these calls:
> * The intended replacement api
> * An owner to do the replacement by the post-l mstone
> ?

I've assigned bugs and owners for all of these (the bugs blocking
crbug.com/392191).  I'll do everything other than skia and net which have
seperate bugs.  Is this OK or do you want a spreadsheet too?

https://codereview.chromium.org/393923002/diff/20001/base/sys_info_android.cc
File base/sys_info_android.cc (right):

https://codereview.chromium.org/393923002/diff/20001/base/sys_info_android.cc...
base/sys_info_android.cc:27: static int (*__real_system_property_get)(const
char*, char*) = NULL;
On 2014/07/17 16:55:50, Nico (away) wrote:
> Is this always called from the same thread? (I think we build with
> -fno-threadsafe-statics.)

Good point.  Updated to use a LazyInstance instead.

[Nico, 2014-07-17 23:32:25]

lgtm :-(

[rmcilroy, 2014-07-18 00:26:53]

On 2014/07/17 23:32:25, Nico (away) wrote:
> lgtm :-(

Thanks Niko.

[rmcilroy, 2014-07-18 00:26:54]

The CQ bit was checked by rmcilroy@chromium.org

[rmcilroy, 2014-07-18 00:26:59]

The CQ bit was checked by rmcilroy@chromium.org

[commit-bot: I haz the power, 2014-07-18 00:28:27]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/rmcilroy@chromium.org/393923002/60001

[commit-bot: I haz the power, 2014-07-18 05:42:00]

Change committed as 284012