Katana plugin v0.7.6: crashes when expanding Usd PointInstances

[bfloch]

Description of Issue

When a point instancing usd file is loaded in Katana and expanded to the instancer
katana crashes with:

0x00007fffbaa626e5 in pxrInternal_v0_7__pxrReserved__::PxrUsdInShipped_PointInstancerUtils::ComputeInstanceTransformsAtTime(
    std::vector<std::vector<pxr
::GfMatrix4d> >,
    std::allocator<std::vector<pxrInternal_v0_7__pxrReserved__::GfMatrix4d, 
    std::allocator<pxrInternal_v0_7__pxrReserved__::GfMatrix4d> > ::vector<pxrInternal_v0_7__pxrReserved__::UsdTimeCode, 
    std::allocator<pxrInternal_v0_7__pxrReserved__::UsdTimeCode> > const&, pxrInternal_v0_7__pxrReserved__::UsdTimeCode)
() from /.../third_party/katana/plugin/Libs/pxrUsdInShipped.so

Looks related to the changes at: 5740c6b

Same file works with 0.7.5.

Steps to Reproduce

1. In Katana drop a pxrUsdIn node.
2. Set the fileName to an usd file with pointInstances
3. In the Scene Graph expand the tree until the point instancer should be loaded in the Viewer
4. Crash.

System Information (OS, Hardware)

Linux, CentOS 7.3

Package Versions

Katana 2.6v1
Usd 0.7.6

Build Flags

Pretty much default.

[bfloch]

I found the issue.
We do not have Scale Attributes in our point instances.
This will fix it:

--- a/third_party/katana/plugin/pxrUsdInShipped/pointInstancerUtils.cpp
+++ b/third_party/katana/plugin/pxrUsdInShipped/pointInstancerUtils.cpp
@@ -83,8 +83,8 @@ PxrUsdInShipped_PointInstancerUtils::ComputeInstanceTransformsAtTime(
         if (positions.size() != numInstances) {
             break;
         }
-        scalesAttr.Get(&scales, sampleTimes[a]);
-        orientationsAttr.Get(&orientations, sampleTimes[a]);
+        bool hasScales = scalesAttr.Get(&scales, sampleTimes[a]);
+        bool hasOrientations = orientationsAttr.Get(&orientations, sampleTimes[a]);
         if (scales.size() > 0 and scales.size() != numInstances) {
             break;
         }
@@ -95,8 +95,14 @@ PxrUsdInShipped_PointInstancerUtils::ComputeInstanceTransformsAtTime(
         for (auto i = decltype(numInstances){0}; i < numInstances; ++i) {
             GfTransform transform;
             transform.SetTranslation(positions[i]);
-            transform.SetScale(scales[i]);
-            transform.SetRotation(GfRotation(orientations[i]));
+            if(hasScales)
+            {
+                transform.SetScale(scales[i]);
+            }
+            if(hasOrientations)
+            {
+                transform.SetRotation(GfRotation(orientations[i]));
+            }
             curr.push_back(transform.GetMatrix());
         }

You care for a PR or is it ok for you guys handling it?

[spiffmon]

Thanks, Blazej! Agreed that's the fix, and I think we can handle it here. Sorry for the regression.

…

On 07/14/2017 10:37 AM, Blazej Floch wrote: I found the issue. We do not have Scale Attributes in our point instances. This will fix it: |--- a/third_party/katana/plugin/pxrUsdInShipped/pointInstancerUtils.cpp +++ b/third_party/katana/plugin/pxrUsdInShipped/pointInstancerUtils.cpp @@ -83,8 +83,8 @@ PxrUsdInShipped_PointInstancerUtils::ComputeInstanceTransformsAtTime( if (positions.size() != numInstances) { break; } - scalesAttr.Get(&scales, sampleTimes[a]); - orientationsAttr.Get(&orientations, sampleTimes[a]); + bool hasScales = scalesAttr.Get(&scales, sampleTimes[a]); + bool hasOrientations = orientationsAttr.Get(&orientations, sampleTimes[a]); if (scales.size() > 0 and scales.size() != numInstances) { break; } @@ -95,8 +95,14 @@ PxrUsdInShipped_PointInstancerUtils::ComputeInstanceTransformsAtTime( for (auto i = decltype(numInstances){0}; i < numInstances; ++i) { GfTransform transform; transform.SetTranslation(positions[i]); - transform.SetScale(scales[i]); - transform.SetRotation(GfRotation(orientations[i])); + if(hasScales) + { + transform.SetScale(scales[i]); + } + if(hasOrientations) + { + transform.SetRotation(GfRotation(orientations[i])); + } curr.push_back(transform.GetMatrix()); } | You care for a PR or is it ok for you guys handling it? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#239 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AF7qaCo98FVwhcwZJIM8JTpF3bQjtdQaks5sN6dSgaJpZM4OXoSq>.

[jtran56]

Filed as internal issue #148741.