Created attachment 467528 [details] Test case for the bug Safari 17 Beta displays output of Webgl Offscreen Canvases vertically mirrored when copying to a BitmapRenderer context when premultipliedAlpha is set to false. This seems to be the same bug as 259306 but I can't be certain as it doesn't appear to be fixed yet so I'm posting this bug anyways. The output of the canvas in the example can be compared with another browser to verify the bug.
Bug 259306 was supposed to be fixed in recent betas. However, I cannot reproduce this, the test case looks the same as in Chrome and in Firefox. Could you please add more information, such as your hardware model and exact OS build?
It's a Mac Mini (I'm not sure which one or which version of the OS as I'm using it remotely through a service for testing my web library) The version of safari it has is Version 17.0 (19616.1.20.11.3)
Created attachment 467532 [details] rendering in Safari, firefox, chrome I don't see any differences.
Created attachment 467535 [details] Screenshot of the bug, compare with other set of screenshots
What is the Mac mini model? In particular, is it an Apple Silicon or an Intel one? You can see the details in About this Mac under the Apple menu.
The service I'm using doesn't allow access to the apple menu as far as I can tell.
I cannot reproduce this on an Intel Mac mini either.
It also appears to happen on iPhone 13 running Safari 17. I'm using BrowserStack just in case you need to know the service.
According to a quick web search, BrowserStack advertises running on physical devices, not in virtualization, which makes this even more surprising. I'm not sure if we can do much more unless this reproduces in other environments, or BrowserStack works with Apple to provide details of their setup. Would you mind reaching out to BrowserStack about this problem, and having them get in touch? https://feedbackassistant.apple.com would be the best way, but additionally weighing in here would be appreciated too.
<rdar://problem/115224959>
I informed browserstack of the issue. They said they are looking into it.
I just received this email from browserstack: Hi Patrick, Thank you for your patience. I hope you are doing good. Our engineering team has checked this further, and we strongly believe that the issue is a Safari bug and not something related to BrowserStack Live implementation. This is based on a previous similar bug reported by others, as well as the original bug ticket with Apple. We are willing to provide any necessary information that you require from our end, but we don't have any follow-up steps to take in this case. Having said that, It may be worth asking the Apple Dev team to attempt to replicate the issue on the exact build of Safari. I have attached a screenshot for your reference. I hope this helps. Best Regards, Saud S BrowserStack Support
Thank you. Is there any chance BrowserStack could share the original bug ticket with Apple?
This doesn't reproduce on browserstack.com. I don't see what we can do about this, given that we cannot reproduce in our VMs, and that this does not affect BrowserStack at this time. Please feel free to reopen if actionable details become available.