Bug 13440

Summary: minmax width "optimization" is completely bogus.
Product: WebKit Reporter: Dave Hyatt <hyatt>
Component: Layout and RenderingAssignee: Dave Hyatt <hyatt>
Status: RESOLVED FIXED    
Severity: Normal    
Priority: P2    
Version: 523.x (Safari 3)   
Hardware: Mac   
OS: OS X 10.4   
Bug Depends on: 13431    
Bug Blocks:    
Attachments:
Description Flags
Test case showing the problem none

Dave Hyatt
Reported 2007-04-21 17:21:57 PDT
recalcMinMaxWidth contains an optimization where it attempts to avoid doing a recalc if a block child's minwidth and maxwidth stay the same. This is a bogus optimization, since a child can have any number of properties change on it (such as float or clear or overflow) that will affect the parent's min max width computations
Attachments
Test case showing the problem (410 bytes, text/html)
2007-04-21 17:23 PDT, Dave Hyatt 		no flags
Details
View All Add attachment proposed patch, testcase, etc.
Dave Hyatt
Comment 1 2007-04-21 17:23:15 PDT
Created attachment 14128 [details] Test case showing the problem Here is a test case illustrating the issue. Maxwidth is computed for a container with two floats inside it. If you then dynamically unfloat one of the children, its minwidth/maxwidth don't change. Therefore the parent does not bother to recomputed its maxwidth, even though the new result would be different.
Dave Hyatt
Comment 2 2007-04-21 17:25:54 PDT
Relating this to 13431, since my rework of calcMinMaxWidth should fix this problem without adversely affecting performance.
Dave Hyatt
Comment 3 2007-04-25 15:24:56 PDT
Fixed by lazy minmaxwidth patch. I landed this test case in r21100.
Note You need to log in before you can comment on or make changes to this bug.