311048 2026-03-29 18:23:07 -0700 Collapsed border color mismatch when table cell has different writing-mode 2026-03-31 06:27:14 -0700 1 1 1 Unclassified WebKit Tables WebKit Nightly Build Unspecified Unspecified RESOLVED FIXED https://github.com/web-platform-tests/wpt/pull/58874 InRadar P2 Normal --- 1 karlcow karlcow webkit-bug-importer zalan oldest_to_newest 2194940 0 478835 karlcow 2026-03-29 18:23:07 -0700 Created attachment 478835 testcase In computeCollapsedBeforeBorder and computeCollapsedAfterBorder, four calls to borderAfter() omit the tableWritingMode() parameter, causing the border value (width/style) to be read using the element's own writing mode while the border color is resolved using the table's writing mode. When a cell has a different writing mode from the table, these disagree about which physical edge is "after", producing a CollapsedBorderValue with width/style from one physical edge and color from another. The affected lines in RenderTableCell.cpp: - computeCollapsedBeforeBorder: - previousCell->style().borderAfter() - previousRow->style().borderAfter() - currSection->style().borderAfter() - computeCollapsedAfterBorder: - lastRowInSpan->style().borderAfter() All four should pass tableWritingMode() as the argument, matching the pattern already used by neighboring lookups in the same functions. The root cause is where afterColorProperty is resolved using tableWritingMode(): CSSPropertyID afterColorProperty = ... CSSProperty::resolveDirectionAwareProperty(CSSPropertyBorderBlockEndColor, tableWritingMode()) ...; This correctly maps to border-bottom-color for a horizontal-tb table. But previousCell->style().borderAfter() uses the cell's own writing mode, so for a vertical-rl cell it reads border-left properties. The resulting CollapsedBorderValue gets its width/style from border-left but its color from border-bottom-color. Steps to Reproduce: 1. With Safari Technology Preview 239 on macOS 2. Open testcase Expected Results: The border between rows should be 3px blue in both tables (cell-below's 3px blue top border wins over cell-above's 1px gray bottom border per the "wider border wins" rule). Actual Results: In the test table, the border between rows renders with the wrong color (gray instead of blue). The gray comes from border-bottom-color of the cell above, which leaks into the collapsed border value due to the writing-mode mismatch. The reference table correctly shows 3px blue. Firefox and Chrome both render 3px blue for both tables. 2194959 1 webkit-bug-importer 2026-03-29 19:10:32 -0700 <rdar://problem/173655092> 2194960 2 karlcow 2026-03-29 19:16:26 -0700 Pull request: https://github.com/WebKit/WebKit/pull/61647 2195513 3 ews-feeder 2026-03-31 06:23:08 -0700 Committed 310303@main (81700e22624d): <https://commits.webkit.org/310303@main> Reviewed commits have been landed. Closing PR #61647 and removing active labels. 2195514 4 karlcow 2026-03-31 06:27:14 -0700 Submitted web-platform-tests pull request: https://github.com/web-platform-tests/wpt/pull/58874 478835 2026-03-29 18:23:07 -0700 2026-03-29 18:23:07 -0700 testcase bug01-collapsed-border-writing-mode.html text/html 3501 karlcow PCFET0NUWVBFIGh0bWw+CjwhLS0KQnVnIDE6IE1pc3NpbmcgdGFibGVXcml0aW5nTW9kZSgpIGlu IGNvbGxhcHNlZCBib3JkZXIgcmVzb2x1dGlvbiAoUmVuZGVyVGFibGVDZWxsLmNwcDoxMDMzKQoK SW4gY29tcHV0ZUNvbGxhcHNlZEJlZm9yZUJvcmRlciwgbGluZSAxMDMzIGNhbGxzOgogICAgcHJl dmlvdXNDZWxsLT5zdHlsZSgpLmJvcmRlckFmdGVyKCkgICAgICAgICAgIOKGkCB1c2VzIGNlbGwn cyBPV04gd3JpdGluZyBtb2RlCkJ1dCB0aGUgY29sb3IgYXQgbGluZSAxMDI0IGlzIHJlc29sdmVk IHZpYToKICAgIENTU1Byb3BlcnR5OjpyZXNvbHZlRGlyZWN0aW9uQXdhcmVQcm9wZXJ0eShDU1NQ cm9wZXJ0eUJvcmRlckJsb2NrRW5kQ29sb3IsIHRhYmxlV3JpdGluZ01vZGUoKSkKICAgICAgICAg ICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAg4oaQIHVzZXMgVEFCTEUn cyB3cml0aW5nIG1vZGUKCkZvciBhIGNlbGwgd2l0aCB3cml0aW5nLW1vZGU6dmVydGljYWwtcmwg aW4gYSBob3Jpem9udGFsLXRiIHRhYmxlOgogIC0gYm9yZGVyQWZ0ZXIoKSBbY2VsbCdzIHZlcnRp Y2FsLXJsXSDihpIgcmVhZHMgYm9yZGVyLUxFRlQgKHZhbHVlOiB3aWR0aCArIHN0eWxlKQogIC0g YWZ0ZXJDb2xvclByb3BlcnR5IFt0YWJsZSdzIGhvcml6b250YWwtdGJdIOKGkiByZXNvbHZlcyB0 byBib3JkZXItQk9UVE9NLWNvbG9yCgpUaGlzIGNyZWF0ZXMgYSBGcmFua2Vuc3RlaW4gQ29sbGFw c2VkQm9yZGVyVmFsdWU6IHdpZHRoL3N0eWxlIGZyb20gYm9yZGVyLWxlZnQsCmNvbG9yIGZyb20g Ym9yZGVyLWJvdHRvbS4gVGhlIG1pc21hdGNoZWQgY29sb3IgY2FuIGNvbnRhbWluYXRlIHRoZSBi b3JkZXIKcmVzb2x1dGlvbiBhbmQgcHJvcGFnYXRlIHRoZSB3cm9uZyBjb2xvciB0byB0aGUgZmlu YWwgcGFpbnRlZCBib3JkZXIuCgpWRVJJRklFRDoKICAtIFNhZmFyaTogIGJvcmRlciBiZXR3ZWVu IHJvd3MgaXMgZ3JheSAoV1JPTkcgLSBjb2xvciBmcm9tIGJvcmRlci1ib3R0b20gbGVha2VkKQog IC0gRmlyZWZveDogYm9yZGVyIGJldHdlZW4gcm93cyBpcyBibHVlIChDT1JSRUNUKQogIC0gQ2hy b21lOiAgYm9yZGVyIGJldHdlZW4gcm93cyBpcyBibHVlIChDT1JSRUNUKQoKQm90aCB0YWJsZXMg dXNlIGlkZW50aWNhbCBDU1MgYm9yZGVycy4gVGhlIE9OTFkgZGlmZmVyZW5jZSBpcyB3cml0aW5n LW1vZGU6dmVydGljYWwtcmwKb24gdGhlIHRvcCBjZWxsLiBUaGUgcmVmZXJlbmNlIHRhYmxlIChu byB3cml0aW5nLW1vZGUgY2hhbmdlKSBzaG93cyBibHVlIGluIGFsbCBicm93c2Vycy4KLS0+Cjxo dG1sPgo8aGVhZD4KPG1ldGEgY2hhcnNldD0idXRmLTgiPgo8dGl0bGU+QnVnIDE6IENvbGxhcHNl ZCBib3JkZXIgY29sb3IgbWlzbWF0Y2ggd2l0aCBtaXhlZCB3cml0aW5nLW1vZGU8L3RpdGxlPgo8 c3R5bGU+CiAgdGFibGUgewogICAgYm9yZGVyLWNvbGxhcHNlOiBjb2xsYXBzZTsKICAgIHdyaXRp bmctbW9kZTogaG9yaXpvbnRhbC10YjsKICAgIHdpZHRoOiAzMDBweDsKICB9CiAgdGQgeyBwYWRk aW5nOiAxMHB4OyB9CiAgI2NlbGwtYWJvdmUgewogICAgd3JpdGluZy1tb2RlOiB2ZXJ0aWNhbC1y bDsKICAgIGJvcmRlci1ib3R0b206IDFweCBzb2xpZCBncmF5OyAgIC8qIHRhYmxlJ3MgImFmdGVy IiBlZGdlIC0gY29sb3IgbGVha3MgaW50byByZXN1bHQgKi8KICAgIGJvcmRlci1sZWZ0OiA1cHgg c29saWQgZ3JlZW47ICAgIC8qIGNlbGwncyBvd24gImFmdGVyIiBlZGdlIGluIHZlcnRpY2FsLXJs ICovCiAgICBib3JkZXItdG9wOiAxcHggc29saWQgZ3JheTsKICAgIGJvcmRlci1yaWdodDogMXB4 IHNvbGlkIGdyYXk7CiAgICBoZWlnaHQ6IDYwcHg7CiAgfQogICNjZWxsLWJlbG93IHsKICAgIGJv cmRlci10b3A6IDNweCBzb2xpZCBibHVlOyAgICAgIC8qIHNob3VsZCB3aW46IDNweCA+IDFweCAq LwogICAgYm9yZGVyLWJvdHRvbTogMXB4IHNvbGlkIGdyYXk7CiAgICBib3JkZXItbGVmdDogMXB4 IHNvbGlkIGdyYXk7CiAgICBib3JkZXItcmlnaHQ6IDFweCBzb2xpZCBncmF5OwogICAgaGVpZ2h0 OiA2MHB4OwogIH0KCiAgLyogUmVmZXJlbmNlOiBpZGVudGljYWwgYm9yZGVycywgbm8gd3JpdGlu Zy1tb2RlIG92ZXJyaWRlICovCiAgI3JlZi1hYm92ZSB7CiAgICBib3JkZXItYm90dG9tOiAxcHgg c29saWQgZ3JheTsKICAgIGJvcmRlci1sZWZ0OiA1cHggc29saWQgZ3JlZW47CiAgICBib3JkZXIt dG9wOiAxcHggc29saWQgZ3JheTsKICAgIGJvcmRlci1yaWdodDogMXB4IHNvbGlkIGdyYXk7CiAg ICBoZWlnaHQ6IDYwcHg7CiAgICBwYWRkaW5nOiAxMHB4OwogIH0KICAjcmVmLWJlbG93IHsKICAg IGJvcmRlci10b3A6IDNweCBzb2xpZCBibHVlOwogICAgYm9yZGVyLWJvdHRvbTogMXB4IHNvbGlk IGdyYXk7CiAgICBib3JkZXItbGVmdDogMXB4IHNvbGlkIGdyYXk7CiAgICBib3JkZXItcmlnaHQ6 IDFweCBzb2xpZCBncmF5OwogICAgaGVpZ2h0OiA2MHB4OwogICAgcGFkZGluZzogMTBweDsKICB9 Cjwvc3R5bGU+CjwvaGVhZD4KPGJvZHk+CjxoMz5CdWcgMTogQ29sbGFwc2VkIGJvcmRlciBjb2xv ciBtaXNtYXRjaCB3aXRoIG1peGVkIHdyaXRpbmctbW9kZTwvaDM+Cgo8cD48Yj5UZXN0IHRhYmxl PC9iPiAodG9wIGNlbGwgaGFzIHdyaXRpbmctbW9kZTogdmVydGljYWwtcmwpOjwvcD4KPHRhYmxl PgogIDx0cj48dGQgaWQ9ImNlbGwtYWJvdmUiPlRvcCBjZWxsICh3cml0aW5nLW1vZGU6IHZlcnRp Y2FsLXJsKTwvdGQ+PC90cj4KICA8dHI+PHRkIGlkPSJjZWxsLWJlbG93Ij5Cb3R0b20gY2VsbCAo bm9ybWFsKTwvdGQ+PC90cj4KPC90YWJsZT4KCjxwPjxiPlJlZmVyZW5jZSB0YWJsZTwvYj4gKHNh bWUgYm9yZGVycywgbm8gd3JpdGluZy1tb2RlIGNoYW5nZSk6PC9wPgo8dGFibGU+CiAgPHRyPjx0 ZCBpZD0icmVmLWFib3ZlIj5Ub3AgY2VsbCAobm8gd3JpdGluZy1tb2RlIGNoYW5nZSk8L3RkPjwv dHI+CiAgPHRyPjx0ZCBpZD0icmVmLWJlbG93Ij5Cb3R0b20gY2VsbCAobm9ybWFsKTwvdGQ+PC90 cj4KPC90YWJsZT4KCjxwPjxiPkV4cGVjdGVkOjwvYj4gQm90aCB0YWJsZXMgc2hvdyBhIDxzcGFu IHN0eWxlPSJjb2xvcjpibHVlO2ZvbnQtd2VpZ2h0OmJvbGQiPjNweCBibHVlPC9zcGFuPgpib3Jk ZXIgYmV0d2VlbiByb3dzIChjZWxsLWJlbG93J3MgM3B4IGJsdWUgdG9wIHdpbnMgb3ZlciBjZWxs LWFib3ZlJ3MgMXB4IGdyYXkgYm90dG9tKS48L3A+CjxwPjxiPlNhZmFyaSBidWc6PC9iPiBUaGUg dGVzdCB0YWJsZSBzaG93cyBhIDxzcGFuIHN0eWxlPSJjb2xvcjpncmF5O2ZvbnQtd2VpZ2h0OmJv bGQiPmdyYXk8L3NwYW4+CmJvcmRlciBiZXR3ZWVuIHJvd3MuIFRoZSBncmF5IGNvbG9yIGxlYWtz IGZyb20gY2VsbC1hYm92ZSdzIGJvcmRlci1ib3R0b20tY29sb3IKYmVjYXVzZSB0aGUgYm9yZGVy IHZhbHVlIGFuZCBjb2xvciBhcmUgcmVzb2x2ZWQgdXNpbmcgZGlmZmVyZW50IHdyaXRpbmcgbW9k ZXMuPC9wPgo8L2JvZHk+CjwvaHRtbD4K