resolves #7 - found appropriate reference in http://graphics.pixar.com/usd/docs/

…emoval later.

fix URL link - capitalization counts in this one

couple of clients. ArchStatCompare was only being used for it's mtime
equality operation. The SAME_FILE operation is dubious in our jungle of
networked file systems, and leaving just the mtime comparators seemed like
a lot of machinery for just an equals. We can revisit if/when we need to
abstract this away for some non-POSIX system.

Windows.
    JsValue::Get{Array,Object} -> ::GetJs{Array,Object}
    PlugPlugin::FindResource -> ::FindPluginResource
    PlugFindResource -> PlugFindPluginResource

This is an API-only change, no functionality should be modified.

This is a documentation/comment/whitespace-only change.

(Internal change: 1638538)

- Move Issues document into an Issues template. This allows github to
prefill issue requests for users, so they will include all of the information
we want, rather than having it in a separate document they may ignore.
- Create a pull request template in the same vein.
- Add a contributing document; this makes github display a banner
before users can submit a pull request. This is in an effort to
streamline the process of our CLAs. For further information on these see:

https://help.github.com/articles/creating-a-pull-request-template-for-your-repository/
https://help.github.com/articles/creating-an-issue-template-for-your-repository/
https://help.github.com/articles/setting-guidelines-for-repository-contributors/

(Internal change: 1640807)

issue #10. It turns out that github markdown changes all anchors to
lowercase during html generation. So a heading like # Foo Bar becomes
somePage.html#foo-bar.

(Internal change: 1640808)

This was caught by running the build against a newer compiler(4.8.5)

(Internal change: 1640836)

This change removes all dependencies in usdMaya on pxrUsdMayaGL. Instead,
pxrUsdMayaGL is made dependent on usdMaya for dealing with proxy shapes.

(Internal change: 1640854)

(Internal change: 1640881)

…e function

prototypes became out of sync with their definitions.

(Internal change: 1641070)

(Internal change: 1641072)

…ion to std::string.

For the pair of clients that were using TfHash on TfToken, just use the TfToken's
hashing directly.  Provide an overload in TfHash to prevent future clients from
unintentionally hashing by std::string.

(Internal change: 1641077)

that pointed to relocated prims caused composition errors.

This was due to how the mapping functions between variant
nodes and their parents were set up. Since variants are
conceptually just switches for local opinions, the namespace
for a variant node should be exactly the same as its parent
node. This is represented by an identity mapping function.
However, in this case Pcp was erroneously adding in
relocations to this mapping function. This caused target
paths that pointed to relocated prims to be unmappable,
because the combination of the identity and relocation
mappings would break the bijection between parent and
variant namespaces that PcpMapFunction maintains.

This change makes it so that Pcp no longer adds relocations
to the map functions used by variants. This behavior was
pushed into _CreateMapExpressionForArc, which is called
in exactly the places where relocates are necessary.

(Internal change: 1641091)

pxr/imaging and pxr/usdImaging are still in bad shape and
not referenced in the produced TOC.

(Internal change: 1641100)

…walking from prman terminals

(This is to silence benign but annoying shell messages for display shaders)

(Internal change: 1641164)

(Internal change: 1641549)
(Internal change: 1644828)

(Internal change: 1641194)
(Internal change: 1641823)

(Internal change: 1641266)

(Internal change: 1641267)

(Internal change: 1641269)

(Internal change: 1641271)

(Internal change: 1641272)

(Internal change: 1641273)

(Internal change: 1641274)

(Internal change: 1641275)

(Internal change: 1641276)

This is a documentation/whitespace -only change.

(Internal change: 1641282)

(Internal change: 1641318)

(Internal change: 1641332)

Hopefully, better errors will assist users with fixing invalid texture filepaths, as a short term solution before a long term fallback fix.

(Internal change: 1641348)

afoul of windows macro of the same name.

(Internal change: 1641404)

(Internal change: 1641405)

fixes issue #4

(Internal change: 1641406)

(Internal change: 1641407)

(Internal change: 1641408)

(Internal change: 1641409)

…nt64_t (we definitely have files bigger than 2.1 GB) and attempt to provide a Windows implementation.

(Internal change: 1641448)

(Internal change: 1641451)

(Internal change: 1641459)

rather than just knowing that there is one.

(Internal change: 1641507)

Changes:
- Added local class _SmoothNormalsWorker for containing the Compute wrapper.
- Removed adjacency from _ComputeSmoothNormals() as it was unused.

(Internal change: 1641508)

(Internal change: 1641512)

implementations.

(Internal change: 1641910)

…nd tests.

(Internal change: 1641975)

(Internal change: 1641980)

…with tbb.

(Internal change: 1641992)
(Internal change: 1641994)

in two different tasks when unloading the last instance of that
master and loading a new instance of that master at the same
time. This could lead to crashes if those tasks were executed
concurrently.

This fix does some additional filtering when computing the
subtrees that need to be recomposed to ensure this doesn't
happen.

(Internal change: 1642055)

previously.  Also fix up signed/unsigned comparisons.

(Internal change: 1642056)

(Internal change: 1642057)

When we modify an instance we set the dirty bit on the instancer in the change tracker.  However, we go to sync, we start by looking at the rprims.  If the rprim is dirty, then we would look to see if the instancer is dirty and update the instance buffer.

So we need a way to dirty the rprims associated with an instancer when the instancer is dirty.

I've solved this by adding a reverse map to the change tracker, so that we can dirty the rprims.

Changes:
- The change tracker now keeps a set of rprims associated with each instancer.
- There are new entry points in the change tracker to add and remove these associations.
- When an instancer is dirtied, we also set the DirtyInstancer bit on all rprims associated with that instancer.
- The render index is responsible for updating the association with the change tracker when rprims are added and removed.

(Internal change: 1642257)

…y lived

in Ar. Another consumer came up that needed this, usdAbc, so this lofts it
to tf. Note that there isn't a wrapping provided, this is because python
already provides this via os.path.splitext().

(Internal change: 1642471)

(Internal change: 1642473)

when adding a feature to usddiff which required querying of various supported
file formats.

(Internal change: 1642474)

files. Update clients to use new --noeffect option to protect against
editing.

(Internal change: 1642475)

…ng, this made it look like textures had no mips and in the case of ambient environments it made it look like the cubemap was not moving.

(Internal change: 1642481)

are cleaned out in all cases.

(Internal change: 1642498)

(Internal change: 1642520)

(Internal change: 1642527)

…struct is

forward declared with the opposite keyword than the one with which it
was defined.

(Internal change: 1642538)

# Conflicts:
#	pxr/base/lib/arch/attributes.h
#	pxr/base/lib/arch/demangle.h
#	pxr/base/lib/arch/errno.h
#	pxr/base/lib/arch/error.h
#	pxr/base/lib/arch/fileSystem.cpp
#	pxr/base/lib/arch/fileSystem.h
#	pxr/base/lib/arch/function.h
#	pxr/base/lib/arch/hash.h
#	pxr/base/lib/arch/mallocHook.h
#	pxr/base/lib/arch/math.h
#	pxr/base/lib/arch/nap.h
#	pxr/base/lib/arch/stackTrace.h
#	pxr/base/lib/arch/symbols.h
#	pxr/base/lib/arch/systemInfo.h
#	pxr/base/lib/arch/testenv/testFileSystem.cpp
#	pxr/base/lib/arch/threads.h
#	pxr/base/lib/arch/timing.h
#	pxr/base/lib/arch/vsnprintf.h
#	pxr/base/lib/gf/bbox3d.h
#	pxr/base/lib/gf/colorRamp.h
#	pxr/base/lib/gf/frustum.h
#	pxr/base/lib/gf/gamma.h
#	pxr/base/lib/gf/homogeneous.h
#	pxr/base/lib/gf/interval.h
#	pxr/base/lib/gf/line.h
#	pxr/base/lib/gf/line2d.h
#	pxr/base/lib/gf/lineSeg.h
#	pxr/base/lib/gf/lineSeg2d.h
#	pxr/base/lib/gf/math.h
#	pxr/base/lib/gf/matrix2d.h
#	pxr/base/lib/gf/matrix2f.h
#	pxr/base/lib/gf/matrix3d.h
#	pxr/base/lib/gf/matrix3f.h
#	pxr/base/lib/gf/matrix4d.h
#	pxr/base/lib/gf/matrix4f.h
#	pxr/base/lib/gf/matrixData.h
#	pxr/base/lib/gf/plane.h
#	pxr/base/lib/gf/quatd.h
#	pxr/base/lib/gf/quaternion.h
#	pxr/base/lib/gf/quatf.h
#	pxr/base/lib/gf/quath.h
#	pxr/base/lib/gf/range1d.h
#	pxr/base/lib/gf/range1f.h
#	pxr/base/lib/gf/range2d.h
#	pxr/base/lib/gf/range2f.h
#	pxr/base/lib/gf/range3d.h
#	pxr/base/lib/gf/range3f.h
#	pxr/base/lib/gf/ray.h
#	pxr/base/lib/gf/rect2i.h
#	pxr/base/lib/gf/rgb.h
#	pxr/base/lib/gf/rgba.h
#	pxr/base/lib/gf/rotation.h
#	pxr/base/lib/gf/size2.h
#	pxr/base/lib/gf/size3.h
#	pxr/base/lib/gf/transform.h
#	pxr/base/lib/gf/vec3d.h
#	pxr/base/lib/gf/vec3f.h
#	pxr/base/lib/gf/vec3h.h
#	pxr/base/lib/js/json.h
#	pxr/base/lib/js/value.h
#	pxr/base/lib/plug/plugin.h
#	pxr/base/lib/tf/anyWeakPtr.h
#	pxr/base/lib/tf/atomicOfstreamWrapper.h
#	pxr/base/lib/tf/callContext.h
#	pxr/base/lib/tf/debug.h
#	pxr/base/lib/tf/diagnosticLite.h
#	pxr/base/lib/tf/diagnosticMgr.h
#	pxr/base/lib/tf/diagnosticNotice.h
#	pxr/base/lib/tf/dl.h
#	pxr/base/lib/tf/enum.h
#	pxr/base/lib/tf/envSetting.h
#	pxr/base/lib/tf/errorMark.h
#	pxr/base/lib/tf/errorTransport.h
#	pxr/base/lib/tf/expiryNotifier.h
#	pxr/base/lib/tf/fileUtils.h
#	pxr/base/lib/tf/getenv.h
#	pxr/base/lib/tf/hash.h
#	pxr/base/lib/tf/instantiateSingleton.h
#	pxr/base/lib/tf/iterator.h
#	pxr/base/lib/tf/mallocTag.h
#	pxr/base/lib/tf/notice.h
#	pxr/base/lib/tf/noticeRegistry.h
#	pxr/base/lib/tf/nullPtr.h
#	pxr/base/lib/tf/pathUtils.h
#	pxr/base/lib/tf/patternMatcher.h
#	pxr/base/lib/tf/pointerAndBits.h
#	pxr/base/lib/tf/poolAllocator.h
#	pxr/base/lib/tf/preprocessorUtils.h
#	pxr/base/lib/tf/pyEnum.h
#	pxr/base/lib/tf/pyInterpreter.h
#	pxr/base/lib/tf/pyLock.h
#	pxr/base/lib/tf/pyObjWrapper.h
#	pxr/base/lib/tf/pyObjectFinder.h
#	pxr/base/lib/tf/pyTracing.h
#	pxr/base/lib/tf/pyUtils.h
#	pxr/base/lib/tf/pyWeakObject.h
#	pxr/base/lib/tf/pyWrapContext.h
#	pxr/base/lib/tf/refBase.h
#	pxr/base/lib/tf/refPtr.h
#	pxr/base/lib/tf/refPtrTracker.h
#	pxr/base/lib/tf/regTest.h
#	pxr/base/lib/tf/registryManager.h
#	pxr/base/lib/tf/scopeDescription.h
#	pxr/base/lib/tf/scriptModuleLoader.h
#	pxr/base/lib/tf/singleton.h
#	pxr/base/lib/tf/staticTokens.h
#	pxr/base/lib/tf/stl.h
#	pxr/base/lib/tf/stopwatch.h
#	pxr/base/lib/tf/stringUtils.h
#	pxr/base/lib/tf/token.h
#	pxr/base/lib/tf/type.h
#	pxr/base/lib/tf/weakBase.h
#	pxr/base/lib/tf/wrapTypeHelpers.h
#	pxr/base/lib/tracelite/trace.h
#	pxr/base/lib/vt/array.h
#	pxr/base/lib/vt/dictionary.h
#	pxr/base/lib/vt/value.h
#	pxr/base/lib/work/detachedTask.h
#	pxr/base/lib/work/threadLimits.h
#	pxr/usd/lib/usdShade/look.h
#	third_party/maya/lib/pxrUsdMayaGL/proxyDrawOverride.h

(Internal change: 1642555)

…conventions in cameraUtil.

(Internal change: 1642792)

A previous change switched the hydra capability check from GL version
to GLEW variable in RenderContextCaps (needed for portability).

There's a fallout that if client doesn't initialize glew before calling
IsEnabledHydra, it never gets enabled. This fix adds a call of glew
initialization in UsdImagingGL::IsEnabledHydra.

(Internal change: 1642808)

…etain the file identifier suffix

which allows users to tell which is which, but we now rely on context managers for the temp
files, so cleanup is handled.

(Internal change: 1642839)

(Internal change: 1642854)

(Internal change: 1642904)

… behavior

with large multi-line string tokens and also avoids us having to do clever mmap
business to ensure that flex gets its two bytes of null padding.  It does mean
that we pay the cost of reading all the bytes even in the (ultra rare) case
that we're reading the metadata section only, but we think this is an okay
tradeoff.

(Internal change: 1642939)

…wn problematic shapes

(Internal change: 1642980)

(Internal change: 1643439)
(Internal change: 1644007)
(Internal change: 1645072)

(Internal change: 1643655)

…s time varying and if so, we will push updates to the gpu when changing time!

(Internal change: 1643690)

AttributeQuery::Get(time).  The main problem is that with clips, the "value
source" for an attribute might itself be time varying.  During times that clips
are active, the source would be clips, at other times, the source might be time
samples.  To handle this we add a new value source:
Usd_ResolveInfoSourceIsTimeDependent, which is returned for timeless
ResolveInfo queries where clips are involved.  In this case,
AttributeQuery::Get() will fall back to the ordinary Attribute::Get() code
path.

(Internal change: 1643699)

…rim in its subtree

if usd native instance doesn't have any renderable prim in its subtree,
book keeping of instance to master prim isn't needed for that path.

(Internal change: 1643828)

"opensubdiv3/" include path had been used to support both 2.x and 3.x
at the same time so that we can compare the results between those versions.
now that opensubdiv 2.x support in hydra has been removed,
we no longer need to distinguish the version by include path.

closes #3
#3

(Internal change: 1643830)

The build now uses a simpler FindPySide CMake module for
detecting the presence of PySide and pysideuic. Also,
pxr_shared_library and pxr_plugin now accept .ui files
under a new 'PYSIDEUI_FILES' argument, which is more
consistent with how we specify other files to build.

(Internal change: 1643902)

This allows us to remove all direct includes of boost::python
headers from tf/type.h, which helps with the Python include
ordering issue.

(Internal change: 1644046)

(Internal change: 1644089)

…usd stage

teardown by ~2x.

(Internal change: 1644215)

inline assembly atomic ops to C++11 std::atomic.

The speed hit came from changing TfRefPtr's internal pointer
from TfRefBase* to std::atomic<TfRefBase*>.  That change was
to accomodate TfWeakPtr's requirement to atomically initialize its
ref-counted Tf_Remnant pointer.  Early testing with godbolt.org
and simple test cases left us expecting similar x86-64 code as
before.  However, perf tests and codegen analysis revealed that
we were getting worse code out of gcc 4.8.2: extra branches and
repeated identical MOV ops.

Back that out in favor of a targeted solution for TfWeakPtr:

* TfRefPtr goes back to holding a TfRefBase*
* TfWeakBase holds its remnant as std::atomic<Tf_Remnant*>
* Tf_Remnant remains TfRefBase
* TfWeakBase manually manages its +1 ref to the remnant instead
  of holding a TfRefPtr; this lets it initialize atomically
  without pushing that constraint onto TfRefPtr
* Corresponding updates to downstream use of Tf_Remnant

(Internal change: 1644269)

callbacks.

Two advantages:

- WorkSerialForN wouldn't allow functors with const operator() before,
  which broken with its intention to closely follow the API of
  WorkParallelForN (which did allow it).

- WorkParallelForN no longer uses boost::function.

(Internal change: 1644395)

usdReferenceAssembly nodes would paths with Ar but the exporter's stage
would not have this context.  This means paths would resolve in maya but
not in the stage we export.

(Internal change: 1644491)

(Internal change: 1644626)

commas in pretty-printed numbers, and initializaing
locale in a non-robust way.

Switch to "simpler" string-formatting idiom that allows us
to avoid locale altogether.

Fixes #17 github issue

(Internal change: 1644691)