1134947 - Ahem font shows different height for zh-TW lang on Windows

Status: NEW

(Core :: Layout: Text and Fonts, defect)

Description

[Xidorn Quan [:xidorn] UTC+11]

Attachment: zh-tw.html

When lang is zh-TW, Ahem font behaves slightly differently than in other lang. See the attachment. You need to install Ahem font [1] before. Do we do anything special for zh-TW?

[1] http://www.w3.org/Style/CSS/Test/Fonts/Ahem/

Comment 1

[Xidorn Quan [:xidorn] UTC+11]

Attachment: screenshot

Comment 2

[Xidorn Quan [:xidorn] UTC+11]

Seems to be a windows-specific problem. It works fine on Mac.

Comment 3

[Xidorn Quan [:xidorn] UTC+11]

Attachment: testcase

Comment 4

[Xidorn Quan [:xidorn] UTC+11]

Attachment: testcase

Use https for the font file so that it can be checked directly on bugzilla.

Comment 5

[Xidorn Quan [:xidorn] UTC+11]

The reason that this happens only on zh-TW on Windows is that, the default font list of zh-TW contains some bad underline font family. Although it is also true for ja and zh-CN fonts, the fonts in the list seem to have an larger underline offset.

The underline offset:
Ahem: ~ -2.66
First bad in ja list: -2.00
First bad in zh-TW list: ~ -3.81

in gfxFontGroup::GetUnderlineOffset(), the smaller one is peeked as the underline offset, which in case of zh-TW is -3.81.

After some further calculation, it finally adds one pixel to the descent and the height.

But simply adding descent probably shouldn't cause this problem?

Comment 6

[John Daggett (:jtd)]

Need to look at this a little closer but guessing we shouldn't be using "first font in list" logic like this.

Comment 7

[Xidorn Quan [:xidorn] UTC+11]

So this problem happens in nsLineLayout::VerticalAlignFrames(), which calls nsLayoutUtils::GetCenteredFontBaseline() to calculate the minBCoord, which is later used in calculation of the baseline of the block.

nsLayoutUtils::GetCenteredFontBaseline() uses the result of nsFontMetrics::MaxHeight() which finally calls gfxFontGroup::GetUnderlineOffset().

Finally, the block outside the text of zh-TW will have a wrong baseline.

Not sure how should we fix this problem.