Open javagl opened 2 months ago
Also reported in https://community.cesium.com/t/maximum-call-stack-size-exceeded-rangeerror/33315/13 , but I have not yet checked that model and where exactly the error comes from (but I'd place my bet on another instance of push.apply
here...)
Can I take on this issue, please?
@Tim-Quattrochi Yes, go ahead!
Just providing a WIP update, I am still working on this.
@Tim-Quattrochi You can (if you want to, and only if you want to) create a 'Draft' pull request for your work in progress. This way, people can track the progress and have a look at the intermediate state. (For larger, "structural" changes, it can be a good way to gather early feedback, but it might not be sooo important here: The changes are probably 'uncontroversial', insofar that it's pretty clear what the changes should look like...)
The issue actually is about a
RangeError: Maximum call stack size exceeded
. But I'll go out on a limb and write this directly as the proposed solution here.There are several places where the
push.apply
is used for potentially large arrays. Butpush.apply
puts the arguments on the stack! (The same as with the...spread
operator). That's not good.The following is an example to reproduce the main problem:
Just running it with
node _CallStackTest.js
will cause theRangeError: Maximum call stack size exceeded
The
push.apply
pattern should be replaced by afor
-loop. Preferably, with a target array with a pre-allocated size (offering an additional potential performance benefit - to be benchmarked if necessary...).(An aside: As seen in the code snippet, there's actually an eslint rule for that - but the
...spread
operator would not solve the issue here!)One specific example of this pattern is in
GltfLoader.parse
:This actually did trigger the
RangeError
for a real-world glTF model that contained >150000 accessors. While such a model may certainly fall into the category of Models that challenge engine's performance or runtime limits , it should not cause a crash.Simply replacing the above snippet with
solved the error. But of course, there are other places that may have to be fixed.
If someone wants to try it out: Here is an artificial model that causes the error:
callStackTest_250x250.zip
Note: This was sparked by a forum thread at https://community.cesium.com/t/maximum-call-stack-size-exceeded-rangeerror/33315 , but should be considered more holistically.