During RenderBlock::layoutRunsAndFloatsInRange the lineOverflowsFromShapeInside is set to true if the lines don't fit the shape inside any more and must be placed at the bottom of the content box. This value is always initialized to false even though the block height is already after the shape bottom. This is wrong for partial layouts when only some of the lines are recomputed (see RenderBlock::determineStartPosition). If the start line is after the shape bottom, the flag will be initialized to false and the block height incorrectly updated to the bottom of the shape content box.
Tests are failing with the patch for 116296.
Linking to the master bug
(In reply to comment #1) > Tests are failing with the patch for 116296. The latest patch on bug #116296 no longer applies to the ToT, please update it and I'll take a look at it. As I see from the change only one shape-inside on regions test is failing due to this issue, please skip and link it to this bug if it's a blocker for you and I'll remove from the skipped list later. Thanks!
(In reply to comment #3) > (In reply to comment #1) > > Tests are failing with the patch for 116296. > > The latest patch on bug #116296 no longer applies to the ToT, please update it and I'll take a look at it. As I see from the change only one shape-inside on regions test is failing due to this issue, please skip and link it to this bug if it's a blocker for you and I'll remove from the skipped list later. > > Thanks! I've updated the WIP patch. The failing tests can be found in the TestExpectations file. For now, I think it's fine if I disable the tests in my patch.
I got rid of lineOverflowsFromShapeInside in bug #117757, thus this bug is not valid anymore.