Fixes related to export of skeleton rest-xforms #1130

pmolodo · 2021-01-28T23:15:31Z

Some bugfixes for edge cases, and made it so that we export restXforms whenever we export bindXforms (ie, the restXforms are calculated from the bindXforms if present).

We were getting some errors in other tools if rest xforms weren't present (as well as an annoying warning/error from usdview, IIRC).

…trix and members ...as the physicalIndex for the two attrs may not line up. Also, use members.numConnectedElements() instead of .numElements(), as using numElements() would sometimes cause a plug to get created when memberPlug.connectedTo() was subsequently called, which for some reason triggered a crash.

pmolodo · 2021-01-28T23:16:01Z

@dgovil recently touched some of this code, if he wants to take a look over...

dgovil · 2021-01-28T23:22:54Z

Also CC'ing @ysiewappl since he touched it too. (we treat my personal fork as a submission staging area)

ysiewappl

I think this makes sense to fallback on the bind transforms, and might solve a problem we're seeing here as well on our end.

ysiewappl · 2021-01-29T22:24:11Z

lib/usd/translators/jointWriter.cpp

@@ -222,16 +221,16 @@ static bool _FindDagPoseMembers(
    indices->resize(numDagPaths);

    std::vector<uint8_t> visitedIndices(numDagPaths, 0);
-    for (unsigned int i = 0; i < membersPlug.numElements(); ++i) {
+    for (unsigned int i = 0; i < membersPlug.numConnectedElements(); ++i) {


I don't mind so much because I think this makes sense (this connection needs to exist anyway to be meaningful so the change will work), but what catches my eye is the reasoning; your comment seems suspicious that somehow connectedTo created a new element. Might be worth trying using evaluateNumElements to see if this is somehow an evaluation caching issue and if so worth reporting as a separate Maya bug.

Yeah, the thing where it was somehow creating an element was odd. Unfortunately, I didn't have time to fully investigate at the time, and I don't know if I'd be able to replicate now. I guess I also didn't feel too motivated too, since I feel like using .numConnectedElements() is conceptually better anyway.
If you want, I can edit the commit message, and remove the bit about the crash / element being created, or throw a bunch more caveats around it.

Not that big of a deal; if it's not something that you cannot reliably reproduce already, this code is fine as is (IMHO).

~~To be clear - does the thumbs up mean you're ok with this as is? Or that you'd like me to edit the commit message?~~
Oops - guess you wrote a message that I didn't see while I was typing this. Will leave the commit as is.

ysiewappl · 2021-01-29T22:38:54Z

lib/usd/translators/jointWriter.cpp

    GfMatrix4d*              xform)
 {
    MStatus status;

    MPlug xformMatrixPlug = dagPoseDep.findPlug("xformMatrix");
-    if (index < xformMatrixPlug.numElements()) {
-        MPlug xformPlug = xformMatrixPlug[index];
+    MPlug xformPlug = xformMatrixPlug.elementByLogicalIndex(logicalIndex, &status);


Right now I know there's only one call site and the only indices stuffed in here should be the logical ones, but just in case, might be worth having debug assert to check if the logical index actually exists already via getExistingArrayAttributeIndices or similar.

My understanding is that if somehow that logical index doesn't exist yet, it will be created... which, as a fallback, seems ok to me.
Though I guess it might make sense to just put in a debug-only check / warning message, since it WOULD be an odd/unexpected occurence, which might warrent further investigation...

Ok, added a check...

addresses code review note: Autodesk#1130 (comment)

pmolodo · 2021-01-30T01:54:33Z

@kxl-adsk - are PF builds a little wonky again? I checked one of the failed builds, and the tests all passed - just some sort of permission error running a groovy script afterwards?

ppt-adsk · 2021-01-30T02:07:14Z

@kxl-adsk - are PF builds a little wonky again? I checked one of the failed builds, and the tests all passed - just some sort of permission error running a groovy script afterwards?

Yes, I got the same and reported it internally. Stay tuned.

kxl-adsk · 2021-01-30T11:41:46Z

lib/usd/translators/jointWriter.cpp

@@ -136,7 +137,6 @@ static GfMatrix4d _GetJointWorldBindTransform(const MDagPath& dagPath)
        CHECK_MSTATUS_AND_RETURN(status, GfMatrix4d(1));
        MPlug plgWorldMatrices = fnNode.findPlug("worldMatrix", false, &status);
        CHECK_MSTATUS_AND_RETURN(status, GfMatrix4d(1));
-        TF_VERIFY(membersIdx < plgWorldMatrices.numElements());


Can you comment on why this check wasn't good?

It's because membersIdx is a logical index, while numElements() gives the number of physical plugs... so the two aren't related.

kxl-adsk · 2021-01-30T11:44:33Z

lib/usd/translators/jointWriter.cpp

-            "The Skeleton's 'restTransforms' property will not be "
-            "authored.",
-            skelPath.GetText());
+        TF_DEBUG(PXRUSDMAYA_TRANSLATORS)


Interested to know why we want to switch this warning to a debug trace? (and same for the one below)

Because we now have an additional fallback - we set the restTransforms from the bindPose - whereas before this would mean that we would "give up", and not author any restTransforms.

I don't mind changing either this or the next one back to TF_WARN if you want though.

Actually - on looking again, I think I'll definitely change the next one back to TF_WARN, because the inability to read the local transform shouldn't happen. (Whereas with this one, it merely indicates that there was no dagPose... which, while not standard, is still kosher.)

Changed the second to TF_WARN, left this one as TF_DEBUG for now - let me know if you want this one back to TF_WARN as well:
afaea55

kxl-adsk · 2021-01-30T11:46:37Z

test/lib/usd/translators/testUsdExportSkeleton.py

@@ -205,6 +207,49 @@ def testSkelWithoutBindPose(self):
            animSource.GetTranslationsAttr().Get(5.0),
            Vt.Vec3fArray([Gf.Vec3f(5.0, 5.0, 0.0)]))

+    def testSkelRestXformsWithNoDagPose(self):
+        """
+        Tests export of of rest xforms when there is no dagPose node at all.


typo: of of

Fixed in afaea55

- Switch TF_DEBUG back to TF_WARN if unable to read local transform from dagPose Autodesk#1130 (comment) - "of of" typo Autodesk#1130 (comment)

Paul Molodowitch added 3 commits January 28, 2021 14:21

joint translator: fix for warning about joints missing from bind pose

a83340a

author restXforms from bindXforms if dagPose can't be used

fb76e95

pmolodo assigned ecp-maya-devops-adsk Jan 29, 2021

ysiewappl approved these changes Jan 29, 2021

View reviewed changes

ysiewappl reviewed Jan 29, 2021

View reviewed changes

add a debug check to verify logicalIndex exists

ad03c38

addresses code review note: Autodesk#1130 (comment)

kxl-adsk reviewed Jan 30, 2021

View reviewed changes

kxl-adsk added the import-export Related to Import and/or Export label Jan 30, 2021

address code review notes from kxl-adsk

afaea55

- Switch TF_DEBUG back to TF_WARN if unable to read local transform from dagPose Autodesk#1130 (comment) - "of of" typo Autodesk#1130 (comment)

kxl-adsk approved these changes Feb 1, 2021

View reviewed changes

kxl-adsk added the ready-for-merge Development process is finished, PR is ready for merge label Feb 1, 2021

pmolodo assigned ecp-maya-devops-adsk and unassigned ecp-maya-devops-adsk Feb 1, 2021

kxl-adsk merged commit 4b98c88 into Autodesk:dev Feb 2, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fixes related to export of skeleton rest-xforms #1130

Fixes related to export of skeleton rest-xforms #1130

pmolodo commented Jan 28, 2021

pmolodo commented Jan 28, 2021

dgovil commented Jan 28, 2021

ysiewappl left a comment •

edited

Loading

ysiewappl Jan 29, 2021

pmolodo Jan 30, 2021

ysiewappl Jan 30, 2021

pmolodo Jan 30, 2021 •

edited

Loading

ysiewappl Jan 29, 2021 •

edited

Loading

pmolodo Jan 30, 2021

pmolodo Jan 30, 2021

pmolodo commented Jan 30, 2021

ppt-adsk commented Jan 30, 2021

kxl-adsk Jan 30, 2021

pmolodo Feb 1, 2021

kxl-adsk Jan 30, 2021

pmolodo Feb 1, 2021

pmolodo Feb 1, 2021

kxl-adsk Jan 30, 2021

pmolodo Feb 1, 2021

Fixes related to export of skeleton rest-xforms #1130

Fixes related to export of skeleton rest-xforms #1130

Conversation

pmolodo commented Jan 28, 2021

pmolodo commented Jan 28, 2021

dgovil commented Jan 28, 2021

ysiewappl left a comment • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

pmolodo Jan 30, 2021 • edited Loading

Choose a reason for hiding this comment

ysiewappl Jan 29, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

pmolodo commented Jan 30, 2021

ppt-adsk commented Jan 30, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ysiewappl left a comment •

edited

Loading

pmolodo Jan 30, 2021 •

edited

Loading

ysiewappl Jan 29, 2021 •

edited

Loading