| Commit-Queue | +1 |
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Can you do this for blobs too? If we allow data url, we should allow blobs too. Can copy the tests from https://crrev.com/c/5876705.
Just FYI, I am landing metrics for the prevalence of this pattern in https://crrev.com/c/7687863.
Can you send a new blink-dev intent for this, or ressurect the old one? I think it's the right change.
if (image->IsSVGImage() && IsDataUrl()) {I think this works, but why do we have SVGImage::HasSingleSecurityOrigin at all? [This comment](https://source.chromium.org/chromium/chromium/src/+/main:third_party/blink/renderer/platform/graphics/image.h;drc=956f8d90f2e6a2a5e654adb872ec956bfbeadb51;l=126) just seems wrong. Can we just take all of this out?
I read the history of this area, and this was all added for visited links (https://bugs.webkit.org/show_bug.cgi?id=119492). But, we don't support links in svg images (https://source.chromium.org/chromium/chromium/src/+/main:third_party/blink/renderer/core/dom/node.cc;drc=956f8d90f2e6a2a5e654adb872ec956bfbeadb51;l=1577) any more, so this shouldn't matter. We could add a CHECK that we are not in an svg image in VisitedLinkState::DetermineLinkStateSlowCase to prove this.
var bitmap = await createImageBitmap(img);This test correctly fails without the patch. I don't understand why https://jsfiddle.net/progers/qhawnyeu allows readback of svg with foreign object. Is it because of this createImageBitmap call--do we detect that as tainting, but drawing img directly doesn't taint?
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |