Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Bringing dev_win_ip up to date with dev #100

Merged
merged 392 commits into from
Nov 3, 2016

Conversation

c64kernal
Copy link
Contributor

Changes required to bring dev_win_ip up to date with current dev.

takahito-tejima and others added 30 commits September 19, 2016 09:44
(Internal change: 1654553)
Modify clip time interpolation code to avoid doing
any computations when given times that exactly match
what's authored in the clipTimes metadata. This
helps avoid floating point precision issues that
may appear if consumers use UsdTimeCode::SafeStep
to author clipTimes that mimic jump discontinuities
(e.g., for looping time-sampled animation from a clip).

(Internal change: 1654654)
…ionBatch

This change allows clients of TestIntersectionBatch() to differentiate
between instances. Otherwise, clients will only see the path to the instancer
prim and only a single hit will be returned.

(Internal change: 1654681)
There are currently two places in pxrUsdMayaGL that use this logic, so this
change consolidates them into a single function.

(Internal change: 1654874)
in case if instance indirection buffer becomes more than
HD_MAX_VBO_SIZE (=1GB by default), stop populating
the buffer, set the number of instances to 1, and raise a
warning.

ideally hydra should split that kind of instance draw into
separate drawItems so that it still draws (even if it's slow)

(Internal change: 1654947)
This is so users aren't bombarded with warnings wrt specific versions
of flex and bison. Most(99%) users will not need to edit the ascii parsers
for any reason so they shouldn't have to care about the warnings as new
code will not be getting generated.

(Internal change: 1655084)
- Show diffs using diff libs.
- Discern arguments wherever possible for the user.
- Clearly display what is changed vs unchanged.

(Internal change: 1655104)
…stancer path available.

(Internal change: 1655114)
…gy of a set of clips

without generating root layer metadata. This is in place to facilitate the upcoming
template metadata which will allow users to have the clip metadata generated for them
by UsdStage.

(Internal change: 1655228)
… range are found.

(Internal change: 1655460)
- HdxDrawTarget inherits from HdSprim
- move related drawTarget structs to Hdx
- localize all drawTarget tokens and dirtybits
- Add general state tracking in HdChangeTracker, to replace
  drawTargetSetVersion (key: HdxDrawTargetTokens->drawTargetSet)
- fix default drawingRange of drawTargetRenderPass to (0.9, -1.0)
- add a unit test for HdxDrawTarget and HdxDrawTargetTask.

- HD_API   26 to 27
- HDX_API 4 to 5.

(Internal change: 1655531)
PrimId display was not working, since HdRenderPassState overrides
ALPHA_TO_COVERAGE to be enabled even in ID Rendering.

adding unit test.

(Internal change: 1655532)
(Internal change: 1655663)
ideally, or at least, we like to put this lock
in UsdImagingDelegate, but instancer rprim syncing is a special
case which we haven't split delegate query into two-step sync
(delegate sync, then rprim sync), so it goes all the way down to
UsdStage directly from HdRenderIndex::Sync in parallel
(i.e. _UpdateSingleValue).

(Internal change: 1655762)
(Internal change: 1655955)
Using -j seems to have hurt build times, which
caused CI builds to time out.

(Internal change: 1655977)
(Internal change: 1656079)
Add the USD plugin install directory to the plugin search path.
This ensures that plugins like usdAbc are usable without having
to manually specify the plugin directory via an environment
variable at runtime.

USD plugins are now installed under the "plugin/usd" directory
instead of just "plugin". This avoids collisions with other
contents of that directory, and also provides a clear location
to install other USD plugins.

(Internal change: 1656966)
(Internal change: 1657174)
(Internal change: 1657247)
… subdiv mesh export

The Maya export code already exports using the OSD3-style face varying linear
interpolation attribute in USD rather than the now deprecated face varying
interpolate boundary attribute from OSD2. This change adds a Maya attribute
for specifying the value directly. We still consider the old attribute if it
is present for backwards compatibility.

(Internal change: 1656200)
…et a rounding error when the knot vectors are scaled (for example, an order-2 knot vector can look like (-0.99999999999 0 0.99999999999 1.999999999 despite the double precision math).

However, the knot ranges are always forced to [0,1] when normalized, so the result is that you can get a uRange or vRange that exceeds the actual knots by some extremely small amount.

This results in validation errors downstream in Presto, and currently causes some sort of presto memory corruption (bugs 136704 and 136705) although it seems OK in other parts of the pipeline.

This fix is to actually change the boundary knot values so that they do not fall inside the [0,1] normalized range.

(Internal change: 1656282)
created by ArchMakeTmpFile, which creates files with permissions 0640, and
is then renamed into place. If there is an existing file at the rename
destination, the permissions of the temporary file are adjusted to match
the destination. However, if no file exists at the destination, the
permissions are left as returned by ArchMakeTmpFile, which means that the
files are only readable by the owner and group.

This change modifies the wrapper to set the permissions for new files to
the default file permissions with the current umask applied. This ensures
that if the system umask specifies that new files should be readable by
others, files created by the wrapper adhere to this.

(Internal change: 1656307)
…] operator

where we didn't intend to modify the hash table.  Even though our code
guarantees that using [] won't logically mutate the container, the []
implementation speculatively grows the table before attempting to find the
element, so it's never safe to invoke concurrently without locking.  Use find()
instead.

(Internal change: 1656375)
a schema-generated header, which does not involve nesting class
declarations inside other classes.  Caught the current issue
because it made doxygen complain.

(Internal change: 1656512)
(Internal change: 1656527)
@sunyab sunyab merged commit 2bea7ca into PixarAnimationStudios:dev_win_ip Nov 3, 2016
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.