For information on the steps to reproduce the issue, read the following bug: 
https://bugs.chromium.org/p/chromium/issues/detail?id=833809

We're opening this bug to follow-up because the change we have reverted to address the regression should not have been reverted. We only made our implementation conform to the GL specification, stating that a texture of size 0 should return 0 bits for all the channels. this was the original change we want to reintroduce in our codebase : 
https://chromium.googlesource.com/angle/angle/+/c0db9addeaebc76c7cc99b26aa27df5e432097ac%5E%21/#F0

We cannot reintroduce the change for now because it causes the regression described in the bug referenced above. We highly suspect that a chrome component is actually relying on our old behavior that was actually wrong, and this is causing the regression.

This bug goal is to find what component is expecting this wrong behavior, fix it, and then reintroduce our original change.