There must be something else in your case because the following html works in Chrome 66:
<iframe style="
    position: fixed;
    bottom: 10%;
    right: 10%;
    background: white;
    border: 2px solid #62afff;
    border-radius: 10px;
    box-shadow: 11px 17px 30px #005aca88;
"></iframe>

Apparently it's https://www.articbot.com and the iframe is shown after clicking the blue thingamajig in the bottom right corner, but the shadow is displayed here in Chrome 66. See if you can reproduce the bug in a new profile without any extensions. See if disabling hardware acceleration in browser settings helps. In case it does, attach the contents of chrome://gpu page.

.articbot001chat {
    all: unset;
    position: fixed;
    bottom: 100px;
    right: 30px;
    background-color: #33b8cc;
    background-image: none;
    border-radius: 10px 10px 10px 10px;
    padding: 0 0 0 0;
    margin: 0 0 0 0;
    box-sizing: border-box;
    border: none;
    float: none;
    box-shadow: 0 0 10px grey;
    overflow: hidden;
}

when i am removing the border-radius, box shadow working fine. this is the issue i am facing. but in the previous chrome version it works fine.

Deleted: chrome_66.PNG 124 KB

	chrome_66.PNG 124 KB View Download

The clip for the iframe border radius is clipping out the shadow. I see it in 66.0.3359.117

woxxom@, what version are you seeing the shadow on? We fixed a regression at some point related to iframe overflow clipping and I suspect that fix is causing this issue.

A reduced test case would be excellent. Is it just any non-composited iframe with border radius and a shadow?

Sorry for misleading, my Chrome 66 had chrome://flags/#enable-experimental-web-platform-features enabled, which prevented the bug.

A reduced test case is attached, it's basically #c1 with a yellow background.

Deleted: test.html 224 bytes

test.html 224 bytes View Download

Thanks woxxom@, that's super helpful.

It's odd to me that experimental-web-platform-features makes any difference. Maybe it's fixed by OOPIF.

As expected, bisects to https://chromium.googlesource.com/chromium/src/+/d03257dd6231949a244fea41890a35b7707f5f08 which was the patch re-adding overflow clip for non-composited iframes. However, it requires --disable-blink-features=SlimmingPaintV175 to reproduce.

My sense is that we should WontFix this because SPv175 is rolling out in M-67 so the bug will disappear shortly. Other thoughts on prioritizing this?

Agreed, WontFix if it is only if SPv175 is off. Please triple check that
that is the case though..

Every bisect run I tried without disabling SPv175 resulted in correct rendering, so I'm confident SPv175 fixes the issue. And I tried a lot of bisect runs because  I knew that this probably worked then didn't then did then didn't, and I was trying to isolate each state.

 Issue 843915  has been merged into this issue.

 Issue 847181  has been merged into this issue.