A strange error with no clear description

[newDINO]

Do anyone know what this error may be?

wgpu error: Validation Error

Cause by:
  In a CommandEncoder
    In a pass parameter
      Encoder is invalid

There's no further information except stack backtrace.
I got this error after modifying a render pass and pipeline. If I don't run this pass, there is no error.
Does it mean there is an error with the configuration of a render pass?

update:
I can do many things with the render pass, but when I call draw() method this error shows up.

[newDINO]

It turns out to be a buffer size mismatch error. The uniform buffer size stated on the cpu side is not equal to the buffer size in the shader.

[ErichDonGubler]

If you could make a minimum reproducible example, that would help with talking about whether the diagnostic is good enough or not. I'm surprised that there's no further information than what you've posted in the OP; if that's truly the case, then I would consider it a bug.

[newDINO]

I'll try reproducing it in a minimum example. In my previous experience with wgpu, buffer size mismatch usually didn't produce any error, maybe the latest change begins to report this error.

[newDINO]

Trying to reproduce in a simple example actually gives the right error report.

wgpu error: Validation Error

Caused by:
  In a CommandEncoder
    In a draw command, kind: Draw
      Buffer is bound with size 4 where the shader expects 64 in group[0] compact index 0

[kpreid]

I found how to cause this problem (by coincidence, in my own program today): #8009

[andyleiserson]

It would also be useful to know the wgpu version, there were changes to error reporting in v26 that could affect how this is reported.

The "encoder is invalid" error is raised when attempting to use the encoder for a subsequent operation, after a previous error invalidated the encoder. The first error is obviously more interesting. In working on the error reporting changes, I tried to prioritize reporting of the first error, but improving diagnostics wasn't the primary intent of the change and there may still be cases where the first error is hidden. It's also possible that the two errors are both reported, but in different ways, and potentially even out of order (if the first error is a case where we correctly defer reporting of the error until finish, and the second is not).

#7881 is one issue in this area, but it doesn't sound like your case fits that one.

[newDINO]

wgpu version is 26.0.1.