Able to reproduce this Windows 10, debian Rodete and Mac 10.13.6 with chrome Stable/Beta: 69.0.3497.92, Dev: 70.0.3538.16, Canary: 71.0.3551.0
Issue broken in M69

Regression Range:
GB: 69.0.3489.0, #574445
BB: 69.0.3491.0, #575129

After executing per-revision bisect script, got the following CL:
https://chromium.googlesource.com/chromium/src/+log/a875a2ea21d8a321295a11440c80af01bcf7fa86..f8b8872f2080130c2b5bdf026f32f46aad7598da

Suspect CL: https://chromium.googlesource.com/chromium/src/+/f8b8872f2080130c2b5bdf026f32f46aad7598da

junov@ Could you please look into it.

RB-Stable for M-69 as this is regressed in M-69, feel free to remove if this should not be blocking.

Hmm, this looks tricky.  Adding in Canvas folks, fserb@ is out - is someone else comfortable looking at this regression ASAP?

I'm 99% sure this is a duplicate of  Issue 878545 . A fix is almost complete which is small and which could be merged back.

The application requires a login and I can't reproduce the behavior from the login screen. Please send me an invitation link (at kbr at chromium.org) if we need to log in to the app to reproduce and confirm the fix.

The issue does not require login to reproduce. Just resize the window up and down until all the text disappears.

Strangely, I seem to be having issues reproducing the bug on desktop, but the issues persist on mobile where resize is not required to trigger the bug. On mobile, the issue seems to happen only on textures that have had a LOT of draw calls, which seems consistent with kbr@'s conclusion that this is a dupe of  Issue 878545 .

I will work with kbr@ to verify that his fix addresses this issue as well.

Reporter: please watch  Issue 878545  and help us test the new Canary (71.0.3551.3 or later) tomorrow or whenever it comes out on Android, and ensure that your app is fixed with it. Thanks.

Do you have a sense for when the patch for 69.x would get released?

Once the fix is verified on Canary, it'll be backported to 69. Please help us test Canary tomorrow. Thanks.

Latest release works fine. Feel free to close this issue since it is a duplicate of 878545.

Thanks submitter for re-testing.

The merge back to M69 was non-trivial, so to avoid the possibility of destabilizing Chrome we are not going to do it. The fix will be in Chrome 70. Submitter, in the future, please test with Chrome Beta at least (if not also Chrome Canary) and file bugs that you see earlier so we can have a chance to fix them sooner. Thanks.

In that case, can you rollback the original offending change on chrome 69 instead?

While we are happy to see prompt responses by the chrome team, it is extremely disappointing that we may have to wait an entire month for a resolution. If the chrome team views canvas APIs as second class citizens, we may regret our decision to even attempt building a webgl-enabled product.

It seems a bit unreasonable to expect startups to accept the burden of regression testing browser releases, but this may be the cost we have to pay as early adopters.

It is not possible to roll back the M69 change which caused this regression; many other changes were built on top of it.