Bug 78310 – MathML internals - use LayoutUnit

• Status: RESOLVED
• Resolution: WONTFIX
• Component: MathML

[Dave Barton]

MathML internals - use LayoutUnit

[Dave Barton]

Created attachment 126423 [details]
Patch

[Darin Adler]

Comment on attachment 126423 [details]
Patch

LayoutUnit will not be changing to float. It will be changing to a fixed point fractional type.

The people who need to review this are Levi and Emil, who know the rationale for when to use int and when to use LayoutUnit.

[Darin Adler]

It’s too bad this combines the LayoutUnit with the create change, because I could review+ the create change without a second thought!

[Dave Barton]

Wow! I'm glad Darin caught my LayoutUnit mistake. I don't know where I thought I read that it was changing to float. Is there somewhere I can read about the proposed future LayoutUnit so I'll get this right? Thanks!

[Dave Barton]

I had been relying on (misinterpreting?) WebCore/rendering/LayoutTypes.h which talked about replacing ints with floats. Obviously I won't do anything until I hear from Levi or Emil, and I don't want to conflict with bug 77916 which I've just discovered, but in the near future what should I do to program in RenderMathML* files? There's a lot of code that adds pixels constants to things, or sets them as padding widths or heights, etc. Can I use LayoutTypes.h roundedLayoutUnit(float) and ceiledLayoutUnit(float)? Are they intended to convert from pixel units to LayoutUnit? Will LayoutUnit turn into a class with an operator int() or operator float() or something that converts to pixels? If LayoutUnit becomes a fixed point type, don't we need to shift or multiply and divide when converting from and to pixel units?? Thanks for any enlightenment!

[Dave Barton]

All right, I have found http://svn.webkit.org/repository/webkit/branches/subpixellayout/Source/WebCore/platform/AppUnit.h and it answers a lot of my questions. Sorry if I asked too many, I think I was tired last night.

[Dave Barton]

The changes in this patch no longer really apply to the current MathML code. Most of the affected routines have been removed, and also I now understand better the preference for CSS-pixel-aligning most boxes, baselines, fraction bars, square root signs, etc.