This CL leaves the Ganesh world in a very odd place. In particular it still:
has pendingIO refs on GrSurfaces
forwards the proxy refs on to the backing GrSurface
Removing everything at once only ends in a mess thus, this goofball CL.
Change-Id: If112ff311bcef2e8d65a36c3b53b0ded4041c24e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/210040
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
This makes GrCFResource a template class with similar semantics to sk_sp.
Change-Id: I9ae9988dac6b39477b16d65591ef6fff44903c36
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/218376
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Reviewed-by: Mike Reed <reed@google.com>
We want the non-color, non-pixel-data version of createBackendTexture to truly create uninitialized textures.
Change-Id: I08867508ea181b7ba3685638cc7a3ea11d527a24
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/218396
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Add retain() instead (this will be added to sk_cf_obj as well).
The issue is that Chrome is storing their objects in a unique_ptr<>
style class which wishes to retain ownership, so we can't use reset(),
which transfers ownership.
Change-Id: I84fd453128963c11c6c8b6e950be8a0851e03066
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/219523
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Change-Id: I6655c2458a4a9b007cb3d5e100f5368697a0f6bb
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/218547
Commit-Queue: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Auto-Submit: Greg Daniel <egdaniel@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Chrome is using operator= to set the Metal texture resource, but
sk_cf_obj doesn't support that for the bare object. This will allow
me to transition them to reset() instead.
Change-Id: Iadfb9624fac2ee6292e3d3f16057a19c683923e9
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/218537
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
This is modeled on how we coordinate VkImage layout changes between
GrContext and clients.
A type GrGLTextureParameters is used to track the current parameter
state.
When a client creates a GrBackendTexture in order to wrap a resource
they created a new GrGLTextureParameters is created and the wrapped
GrGLTexture will share ownership.
When GrContext creates a non-wrapped GrGLTexture, the GrGLTexture
creates a new GrGLTextureParameters and any GrBackendTextures created
from that GrGLTexture will share ownership.
Clients indicate parameter changes by calling
GrBackendTexture::glTextureParametersModified().
We still assume all texture parameters may have changed after a call
to GrContext::resetContext() (for now). The "timestamp" that is used
to implement this has been moved from GrGpu to GrGLGpu as there were
no other use cases.
Change-Id: Ic24e00488fad254a29d5eec6890278b67df6efae
Bug: skia:7966
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/217385
Auto-Submit: Brian Salomon <bsalomon@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
The texture resource in MtlTextureInfo is a CoreFoundation object,
so we have to manage the refcounting ourselves. GrCFResource is a
wrapper class which will automatically take care of this for us on
creation, assignment, and destruction.
Bug: chromium:964498
Change-Id: I9a3768833744d4aaaec2142200283f0e7eaad165
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/216351
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Reviewed-by: Christopher Cameron <ccameron@chromium.org>
Reviewed-by: Mike Reed <reed@google.com>
This reverts commit 186a295111.
Reason for revert: Metal bots failing
Original change's description:
> Change Metal to not take ownership of objects
>
> Prior to this change, Skia/Metal interfaces take ownership of the Metal
> objects passed in (that is, the caller should count passing the object
> to Skia as "freeing" the object).
>
> Change this behavior so that Skia/Metal retains its own separate
> ownership of the Metal objects.
>
> Make GrBackendTexture and GrBackendRenderTarget maintain their own
> references to the underlying MTLTexture by using the CFRetain/CFRelease
> interfaces. Do this by adding a private GrMtlBackendSurfaceInfo.
>
> Move GrMtlBackendSurfaceInfo (formerly GrMtlTextureInfo) out of the
> union in GrBackendTexture and GrBackendRenderTarget because unions
> cannot have nontrivial constructors and destructors (how fVkInfo isn't
> causing a compile error is unclear).
>
> Change-Id: Iae3719c0715825d86503d03c766e47f0f6015bdf
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/215685
> Commit-Queue: Jim Van Verth <jvanverth@google.com>
> Reviewed-by: Jim Van Verth <jvanverth@google.com>
TBR=egdaniel@google.com,jvanverth@google.com,bsalomon@google.com,ccameron@chromium.org
# Not skipping CQ checks because original CL landed > 1 day ago.
Change-Id: Ie569fe1938857706b5413876a9480ef1eb3314ea
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/216221
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Prior to this change, Skia/Metal interfaces take ownership of the Metal
objects passed in (that is, the caller should count passing the object
to Skia as "freeing" the object).
Change this behavior so that Skia/Metal retains its own separate
ownership of the Metal objects.
Make GrBackendTexture and GrBackendRenderTarget maintain their own
references to the underlying MTLTexture by using the CFRetain/CFRelease
interfaces. Do this by adding a private GrMtlBackendSurfaceInfo.
Move GrMtlBackendSurfaceInfo (formerly GrMtlTextureInfo) out of the
union in GrBackendTexture and GrBackendRenderTarget because unions
cannot have nontrivial constructors and destructors (how fVkInfo isn't
causing a compile error is unclear).
Change-Id: Iae3719c0715825d86503d03c766e47f0f6015bdf
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/215685
Commit-Queue: Jim Van Verth <jvanverth@google.com>
Reviewed-by: Jim Van Verth <jvanverth@google.com>
This initial portion of the API should be ready to go. Follow on CLs will add the other entry points.
Change-Id: Ia9c708046ba08b16f9a71558e2bf2c38279abe5d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/214680
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
This reverts commit 753836fcad.
Reason for revert: fixed
Original change's description:
> Revert "Get EGLimage functions out of GrGLInterface."
>
> This reverts commit bc233135e4.
>
> Reason for revert:
> EGLImageTest and TextureBindingTest broken on Windows ANGLE
> Build failures on Debian9, unable to link
> ld.lld: error: undefined symbol: eglCreateImageKHR
>
> Original change's description:
> > Get EGLimage functions out of GrGLInterface.
> >
> > Only used in test code which can just directly call these.
> >
> > Also removed related code in interface assemble/validate generator.
> >
> > Change-Id: I7e821ebb9328bdd2a54d8ab3af929399ca0517d5
> > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/213480
> > Commit-Queue: Robert Phillips <robertphillips@google.com>
> > Reviewed-by: Robert Phillips <robertphillips@google.com>
>
> TBR=bsalomon@google.com,robertphillips@google.com
>
> Change-Id: I622fadf22c5d79f5b1d335a8a81455978d49a86a
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/213960
> Reviewed-by: Michael Ludwig <michaelludwig@google.com>
> Commit-Queue: Michael Ludwig <michaelludwig@google.com>
TBR=bsalomon@google.com,robertphillips@google.com,michaelludwig@google.com
EXTRA_TRYBOTS=Test-Win10-Clang-AlphaR2-GPU-RadeonR9M470X-x86_64-Debug-All-ANGLE
Change-Id: I71397072fa79ae3c72f3835c5e991b9ef2465a9a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/214040
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
This reverts commit bc233135e4.
Reason for revert:
EGLImageTest and TextureBindingTest broken on Windows ANGLE
Build failures on Debian9, unable to link
ld.lld: error: undefined symbol: eglCreateImageKHR
Original change's description:
> Get EGLimage functions out of GrGLInterface.
>
> Only used in test code which can just directly call these.
>
> Also removed related code in interface assemble/validate generator.
>
> Change-Id: I7e821ebb9328bdd2a54d8ab3af929399ca0517d5
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/213480
> Commit-Queue: Robert Phillips <robertphillips@google.com>
> Reviewed-by: Robert Phillips <robertphillips@google.com>
TBR=bsalomon@google.com,robertphillips@google.com
Change-Id: I622fadf22c5d79f5b1d335a8a81455978d49a86a
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/213960
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
Only used in test code which can just directly call these.
Also removed related code in interface assemble/validate generator.
Change-Id: I7e821ebb9328bdd2a54d8ab3af929399ca0517d5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/213480
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
This intended to bring this API more into line with the proposed GrBackendObject API with an eye towards replacing the former with the latter.
TBR=bsalomon@google.com
Change-Id: I4367f03fb10fff788749f21c4843060111a6df1c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/213220
Reviewed-by: Robert Phillips <robertphillips@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
This will allow clients to specify a set of SkImages and SkSurfaces that they
want transitioned to a state which they can use for their own external IO.
Specifically for Vulkan this will move the surface and images back to the queues
in which they were originally wrapped or created with.
Bug: skia:8802
Change-Id: I6a76c4c4a333a8e752632d349899f5fd9921329d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/210460
Commit-Queue: Greg Daniel <egdaniel@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
This will be required for async readback support.
This is implemented using sync objects when available and otherwise
calls glFinish.
Relaxes the unit test requirement that providing a callback with no
work to flush always calls the proc immediately.
Bug: skia:8962
Change-Id: Ieefcab6dccc3924e50260343f01904e7303bb12b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/212198
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
Allows clients to customize behavior when shaders fail to compile.
Added nicer shader error handling to viewer.
Change-Id: If82b48e40d64fd786f37e88c564fd623b53c7f9d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/211361
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Part of a multi-step plan:
1) Add this option (which is *not* GL specific).
2) Disable this in Chromium at context-creation.
3) Finish adding the shader error handler, with default
behavior controlled via this flag, and remove Chromium
driver checks from the corresponding code at the same time.
Change-Id: I7e50c8404a05c1f291f187cf48481d0cbb34066a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/211646
Reviewed-by: Robert Phillips <robertphillips@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This should have very little impact until we enable opList-splitting-reduction.
Change-Id: Iacc3d1b34c390b65b5ee185bcbdd118d5023aaa7
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/210630
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Current strategy: everything from the top
Things to look at first are the manual changes:
- added tools/rewrite_includes.py
- removed -Idirectives from BUILD.gn
- various compile.sh simplifications
- tweak tools/embed_resources.py
- update gn/find_headers.py to write paths from the top
- update gn/gn_to_bp.py SkUserConfig.h layout
so that #include "include/config/SkUserConfig.h" always
gets the header we want.
No-Presubmit: true
Change-Id: I73a4b181654e0e38d229bc456c0d0854bae3363e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/209706
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Hal Canary <halcanary@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Florin Malita <fmalita@chromium.org>
Repurpose GrTRecorder for storing these tasks. It's currently unused.
Reimplement on top of SkArenaAlloc and using emplace methods now that we
have C++14.
Currently it stores copy and upload tasks. In the future it will store
transfer-out commands.
Removes the optimization that reset copy-ins on clear/discard. However,
none of our existing tests exercised it.
Change-Id: I0474f77cc2d368461d542de50a7a0c5609312001
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/209643
Reviewed-by: Chris Dalton <csmartdalton@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This behaves like the SkSurface::wait call but is not tied to a specific
surface.
Change-Id: Ic572296e0f581204bf69a7178645d99e365c0692
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/209648
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
I left off the optional finished proc on the previous cl :(
Change-Id: I5f540c718bb7e12e7ea74287f1e50e35edd79700
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/209417
Commit-Queue: Greg Daniel <egdaniel@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Auto-Submit: Greg Daniel <egdaniel@google.com>
Still only works with the GL backend. For other backends,
need to add similar logic to the shader caching, and some
extra checks in Viewer to force the SkSL flag on.
But in GL, this lets you toggle the checkbox and see the
SkSL / GLSL at will (and edit in either form).
Change-Id: I6d392113aa9cbcbd6e64589b849de70d0ac3beeb
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/209165
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
GrCaps now only provides the offset alignment requirement. The row bytes
is always bpp * width.
GrGpu::transferPixelsFrom now just returns bool since row bytes value is
implicit. It now asserts offset is aligned with GrCap's provided value
in base class.
Implement caps for GL.
Bug: skia:8962
Change-Id: I3299b62efe9fe05bfe02f2a6a4c2704f647d0f8a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/206686
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This is to match the current features of SkSurface::flush and to prepare
for adding additional features to flush.
Bug: skia:8802
Change-Id: I5d68272e1277b416af357e6ffaf426841ceda943
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/207301
Commit-Queue: Greg Daniel <egdaniel@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
We don't want to just disable the "ccpr" path renderer anymore (or
whatever it should be called), because we are adding an MSAA
implementation. This new cap will eventually tell ccpr whether it can
use coverage counting, or if it should just use the MSAA impl.
Bug: skia:
Change-Id: Ie6e5ca1a637ca4408bc6bb844153afa9da26f58e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/204883
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
This reverts commit 215da624d1.
Reason for revert: Blink issues ironed out.
Original change's description:
> Revert "Remove the NullGL interface (and associated test context)"
>
> This reverts commit de206c75c2.
>
> Reason for revert: Chrome is having issues with the switch to Mock in blink tests.
>
> Original change's description:
> > Remove the NullGL interface (and associated test context)
> >
> > Bug: skia:
> > Change-Id: Ie3c9ee39fc1e0a4406de085c60d8433ffb4419df
> > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/203708
> > Reviewed-by: Brian Salomon <bsalomon@google.com>
> > Commit-Queue: Brian Osman <brianosman@google.com>
>
> TBR=bsalomon@google.com,robertphillips@google.com,brianosman@google.com
>
> Change-Id: Iff0cbf29dcea26957efc800a8c33d0ad8285de0a
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Bug: skia:
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/205343
> Reviewed-by: Brian Osman <brianosman@google.com>
> Commit-Queue: Brian Osman <brianosman@google.com>
TBR=bsalomon@google.com,robertphillips@google.com,brianosman@google.com
# Not skipping CQ checks because original CL landed > 1 day ago.
Bug: skia:
Change-Id: I1f4fbbcb00f302c5d830cb1392badd6ec7a33c69
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/205832
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
We want GrResourceCache to be able to track which resources are held by
refs (as opposed to pending IOs) so that it can track the affect of
flushing on resource purgeability. Therefore, all cases that can add the
first ref to a GrGpuResource must funnel through GrResourceCache. This
lays the groundwork by restricting initial refs.
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:8927
Change-Id: I1213c3db258d2412df6666e3222419211ceaa192
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/205482
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
Change-Id: I30a042a41991a47ff1a4e008b88431b4767453dd
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/205496
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This reverts commit de206c75c2.
Reason for revert: Chrome is having issues with the switch to Mock in blink tests.
Original change's description:
> Remove the NullGL interface (and associated test context)
>
> Bug: skia:
> Change-Id: Ie3c9ee39fc1e0a4406de085c60d8433ffb4419df
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/203708
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Commit-Queue: Brian Osman <brianosman@google.com>
TBR=bsalomon@google.com,robertphillips@google.com,brianosman@google.com
Change-Id: Iff0cbf29dcea26957efc800a8c33d0ad8285de0a
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/205343
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Bug: skia:
Change-Id: Ie3c9ee39fc1e0a4406de085c60d8433ffb4419df
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/203708
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
There is some logic in here for 2.0 as well, just as a
"as long as I was looking at the specs", but only 1.0
is really supported.
This seems to resolve the bug where some GPUs weren't
advertising correctly that they had vertex array object
support, by checking for both extension names (with and
without "GL_" prefix)
Of note, this saves about 18 Kb (5.5 Kb gzipped) of code size
by compiling out the unneeded GLES checks/functionality.
Bug: skia:8378
Change-Id: I773bf4dbf231b991051d2a9f640b8047a9010e7d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/203461
Commit-Queue: Kevin Lubick <kjlubick@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
This lets memory traces agree on how big textures are.
Bug: chromium:944846
Change-Id: I5ea8cb5e1331b4ad8d6f59f656c6e61d44290489
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/203727
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Auto-Submit: Adrienne Walker <enne@chromium.org>
Bug: skia:
Change-Id: Id403aabcae1dbf38e55381bb3b4353e85a12a745
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/203705
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Bug: skia:
Change-Id: I5d447ea40ef541682f1e855d14e07bf8e45ee89a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/203392
Auto-Submit: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
This allows us to remove certain interfaces at compile time.
This replaces most (all?) of the cases where it was
if (gl) else [implicit gles] to be explicitly
if (gl) else if (gles)
in preparation for adding a WebGL standard.
For consistency, I tried to check first for GL, then for
GLES, which involved re-arranging a few if blocks.
PS 3 removes about 1.2KB (0.4 KB gzipped) from CanvasKit by
removing the GrGLInterface related checks from
GrGlInterface::validate()
PS 8 removes a total of 6.0 KB (2.6 KB gzipped) from
GrGlInterface::validate() and GrGlCaps::*
Bug: skia:8378
Change-Id: Ia91b732d888907f5d94b0eac8ca023084999fa7e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/201604
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Kevin Lubick <kjlubick@google.com>
one for GL, one for GLES, and one (soon) for WebGL
Bug: skia:8378
Change-Id: Ib13699b7432ed56cce99ac568840e5575bb4d2e5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/201654
Commit-Queue: Kevin Lubick <kjlubick@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>