REGRESSION (r130570): fast/writing-mode/vertical-subst-font-vert-no-dflt.html fails on non-Chromium platforms

• Component: Tools / Tests
• Status: RESOLVED WONTFIX

[Zan Dobersek]

fast/writing-mode/vertical-subst-font-vert-no-dflt.html is failing on non-Chromium platforms since being introduced in r130570.
http://trac.webkit.org/changeset/130570
http://test-results.appspot.com/dashboards/flakiness_dashboard.html#group=%40ToT%20-%20webkit.org&tests=fast%2Fwriting-mode%2Fvertical-subst-font-vert-no-dflt.html

[Zan Dobersek]

Created attachment 167442 [details]
GTK actual output

[Zan Dobersek]

Created attachment 167444 [details]
GTK expected output

[Zan Dobersek]

(In reply to comment #1)
> Created an attachment (id=167442) [details]
> GTK actual output

(In reply to comment #2)
> Created an attachment (id=167444) [details]
> GTK expected output

To analyze the GTK results, it seems as the actual output is correct while the expected output is not rendered correctly.

[Xianzhu Wang]

(In reply to comment #3)
> (In reply to comment #1)
> > Created an attachment (id=167442) [details] [details]
> > GTK actual output
> 
> (In reply to comment #2)
> > Created an attachment (id=167444) [details] [details]
> > GTK expected output
> 
> To analyze the GTK results, it seems as the actual output is correct while the expected output is not rendered correctly.

No. According to Unicode standard, the puctucations in CJK vertical text, the punctuations should be rotated, that is, the glyphs should be substituted with the corresponding vertical forms of them according to the GSUB table in the font.

[Xianzhu Wang]

(In reply to comment #3)
> (In reply to comment #1)
> > Created an attachment (id=167442) [details] [details]
> > GTK actual output
> 
> (In reply to comment #2)
> > Created an attachment (id=167444) [details] [details]
> > GTK expected output
> 
> To analyze the GTK results, it seems as the actual output is correct while the expected output is not rendered correctly.

(Sorry for early enter of #4. Please ignore it)

No. According to Unicode standard, for the puctucations in CJK vertical text, the glyphs should be substituted with the corresponding vertical forms of them according to the GSUB table in the font. The new layout test is testing this behavior. It's a bug that the glyphs are not properly substituted.

[Xianzhu Wang]

(In reply to comment #5)
> (In reply to comment #3)
> > (In reply to comment #1)
> > > Created an attachment (id=167442) [details] [details] [details]
> > > GTK actual output
> > 
> > (In reply to comment #2)
> > > Created an attachment (id=167444) [details] [details] [details]
> > > GTK expected output
> > 
> > To analyze the GTK results, it seems as the actual output is correct while the expected output is not rendered correctly.
> 
> (Sorry for early enter of #4. Please ignore it)
> 
(In reply to comment #5)
> (In reply to comment #3)
> > (In reply to comment #1)
> > > Created an attachment (id=167442) [details] [details] [details]
> > > GTK actual output
> > 
> > (In reply to comment #2)
> > > Created an attachment (id=167444) [details] [details] [details]
> > > GTK expected output
> > 
> > To analyze the GTK results, it seems as the actual output is correct while the expected output is not rendered correctly.
> 
> (Sorry for early enter of #4. Please ignore it)
> 
> No. According to Unicode standard, for the puctucations in CJK vertical text, the glyphs should be substituted with the corresponding vertical forms of them according to the GSUB table in the font. The new layout test is testing this behavior. It's a bug that the glyphs are not properly substituted.

Sorry I misunderstood the comment about expected and actual results. #3 is correct. In the expected result the vertical forms are unnecessarily rotated, perhaps because the characters are not treated as CJK characters (which should not be rotated in vertical text).

[Csaba Osztrogonác]

Skipped on Qt by https://trac.webkit.org/changeset/130628.
Please unskip it with the proper fix.

[Xianzhu Wang]

Checked the GTK output again, and found that this is a more fundamental issue: the GTK port just doesn't do anything for vertical CJK text. It just rotates all CJK glyphs same as other glyphs, so the original forms of the punctuations are rotated "as expected", and the ideographs and vertical forms are also rotated unexpectedly. (bug 50619)

[Xianzhu Wang]

(In reply to comment #7)
> Skipped on Qt by https://trac.webkit.org/changeset/130628.
> Please unskip it with the proper fix.

Could you attach the actual and expected images on Qt? Thanks.

[Csaba Osztrogonác]

Qt results: http://build.webkit.sed.hu/results/x86-64%20Linux%20Qt%20Release/r130627%20%2843536%29/fast/exclusions/shape-inside/shape-inside-rounded-rectangle-003-diffs.html

[Csaba Osztrogonác]

Ooops, it was wrong URL.

Here is the correct one: http://build.webkit.sed.hu/results/x86-64%20Linux%20Qt%20Release/r130627%20%2843536%29/fast/writing-mode/vertical-subst-font-vert-no-dflt-diffs.html

[Xianzhu Wang]

(In reply to comment #11)
> Ooops, it was wrong URL.
> 
> Here is the correct one: http://build.webkit.sed.hu/results/x86-64%20Linux%20Qt%20Release/r130627%20%2843536%29/fast/writing-mode/vertical-subst-font-vert-no-dflt-diffs.html

Thanks for the result. So this is the bug: https://bugs.webkit.org/show_bug.cgi?id=51584.

[Stephen Chenney]

Marked LayoutTest bugs, bugs with Chromium IDs, and some others as WontFix. Test failure bugs still are trackable via TestExpectations or disabled unit tests.