Changes

September 23October 13-2926, 2006

*Discussion There was discussion about profiling the testcase for bug 352367 on how the reflow branch. This bug is Firefox 1renders large tables with forms MUCH slower than IE (100x slower)The writer was asking about trying to reduce the calls to various functions like nsRuleNode::GetStyleData and nsStyleContext::GetStyleData .5.0.7 DOM generates XHTML inline elements using Wordpress. It Ultimately it was determined mentioned that the generated elements from Wordpress do not follow the WC3 guidelines in appendix C GetWidthInfo() took about 20% of XHTML 1.0. Although the W3 validation accepted the generated XHTML as valid, a bug identified within the HTML Working Group ignored the invalid elements. Details can be located at load time [http://groups.google.com/group/mozilla.dev.tech.layout/browse_thread/thread/42e27d7b95193c533efaffa2893333c4/81c1636a55a314744d088595f0df53a2#81c1636a55a31474 layout confusion4d088595f0df53a2 Perf question (reflow branch)].*Refactoring In MSIE table captions are displayed at the same size as content. But in Firefox the nsHTMLReflowState(ComputeBlockBoxDatatext displays at browser default, InitConstraints) It also exceeds the width of the table and nsImageFrame::GetDesiredSize which uses NS_INRINSICSIZE, into the following method: <pre>/** * Compute excess is hidden if the size that table is enclosed in a frame will occupybounding div. Called while * constructing the nsHTMLReflowState The writer wants to be used know if Is there any way of applying a font-size to Reflow firefox table caption text while keeping the frame, * correct display in order to fill its mComputedWidth and mComputedHeight member * variablesother browsers. * * The |height| member of In addition is it possible for preventing the return value may be * NS_UNCONSTRAINEDSIZE, but caption width from exceeding the |tables'width| member must not be[http://groups.google.com/group/mozilla.dev.tech.layout/browse_thread/thread/f7872557085da571/6c3550647c850aa0#6c3550647c850aa0 firefox table captions]. * * @param aAvailWidth The available Can the Gecko engine be used to get the width into which the element is * being placed information of some HTML elements (isuch as table,tr,td) [http://groups.google.com/group/mozilla.dev.etech., the layout/browse_thread/thread/12cce877ec767ac0/1bc65afbdf4df6b1#1bc65afbdf4df6b1 Help for getting HTML element width of its containing info]. * block)There was a discussion about interfaces allowing multiple presshells. * @param aMargin The sum of writer is for the left idea and right margins of the * frame, including actual values resulting from * percentages, but not including actual values * resulting from 'auto'believes this issue should be cleared up [http://groups.google.com/group/mozilla.dev.tech.layout/browse_thread/thread/962e3dd005e41bb7/198207d527ff7b4a#198207d527ff7b4a Multiple presshells]. * @param aBorder The sum of writer is seeking instructions on how to fix the table border color quirk that makes the left and right border widths of color dependent on the * framebackground. * @param aPadding The sum quirk is a pure NN4.6 quirk[http://groups.google.com/group/mozilla.dev.tech.layout/browse_thread/thread/a09888064bc1eaf1/0f6cb460af73c8bf#0f6cb460af73c8bf Removal of the left and right margins of the * frame, including actual values resulting from * percentagesNN4.6 table border color quirk]. * @param aShrinkWrap Whether the frame The writer is in trying to convert nsHTMLReflowState into a context where * non-replaced blocks should shrink-wrap (ebase class [http://groups.google.com/group/mozilla.gdev., * it's floating, absolutely positioned, or * inline-block)tech. *layout/browse_thread/thread/ virtual nsSize ComputeSize(nscoord aAvailWidth, nscoord aMargin, nscoord aBorder, nscoord aPadding, PRBool aShrinkWrap) = 0;<834413484e3a9c6c/pre>b79dcb6c19c4174c#b79dcb6c19c4174c pulling part of nsHTMLReflowState into a base class]. :Details can be located at refactoring some of * A few changes have been made to the reflow branch [http://groups.google.com/group/mozilla.dev.tech.layout/browse_thread/thread/9bbd4ceea2b72d1757efcb49c4ba32fc/aa3c40d434499e2b240f08f41872eff7#aa3c40d434499e2b nsHTMLReflowState240f08f41872eff7 reflow branch status]. *Discussion There was a discussion on how to determine whether there is pending CSS layout changes or reflows when using a popup. It behavior which was concluded that about faux columns with a frame model change is utilized within a popup, not a layout change. On the reflow branch, check the DIRTY mix of undefined and DIRTY_CHILDREN framestate flags. That won't help with pending style changes, but it'll work to detect cases when the popup or something in it needs to be reflown. Details can be located at fix width columns [http://groups.google.com/group/mozilla.dev.tech.layout/browse_thread/thread/dd6f6a3d6c8b1bad9f2644e5325f1fdd/f49d3694aa682f6933f85b837ea56bd5#f49d3694aa682f69 frame reflow33f85b837ea56bd5 faux columns with a mix of undefined and fix width columns].*Issues related There was a question on what modifications should be done to the ongoing implementation of MathML-in-HTML into deactivate a ComboBox from Mozilla, which include: :*Exposing the MathML implementation to tag soup.:*CSS Matching rules. :Details can be located at [http://groups.google.com/group/mozilla.dev.tech.layout/browse_thread/thread/f05a4077a9eac39fedf3f8d7f937c4fd/94d10105c7f627d777f727bf5bb6ff9b#94d10105c7f627d7 MathML-77f727bf5bb6ff9b ComboBox in-HTML5Mozilla].