Bug 34918

Summary: [GTK][Qt] Fails glyph reordering test fast/text/glyph-reordering.html
Product: WebKit Reporter: Gustavo Noronha (kov) <gustavo>
Component: TextAssignee: Nobody <webkit-unassigned>
Status: RESOLVED FIXED    
Severity: Normal CC: abecsi, mitz, ossy, vestbo
Priority: P2 Keywords: Gtk, Qt
Version: 528+ (Nightly build)   
Hardware: PC   
OS: OS X 10.5   
Attachments:
Description Flags
test in GTK+
none
Make the test less dependent on the complex text implementation none

Description Gustavo Noronha (kov) 2010-02-13 06:02:22 PST
Test output:

This tests text selection in complex scripts where glyph reordering occurs.

हिनà¥à¤¦à¥€
ि‌हनà¥à¤¦à¥€
FAIL: width was 116 instead of 174
Comment 1 Csaba Osztrogonác 2010-02-13 07:40:57 PST
This test fails on Qt port too, so I it was skipped by http://trac.webkit.org/changeset/54753 .

output:
This tests text selection in complex scripts where glyph reordering occurs.

हिन्दी
ि‌हन्दी
FAIL: width was 96 instead of 192
Comment 2 mitz 2010-02-13 10:16:27 PST
Interesting. If you open the test in a browser, do the two lines of Devanagari text appear identical?
Comment 3 Csaba Osztrogonác 2010-02-13 12:55:49 PST
(In reply to comment #2)
> Interesting. If you open the test in a browser, do the two lines of Devanagari
> text appear identical?

Unfortunately I saw only 6 and 8 black squares in QtLauncher. :(

Tor Arne, it seems that we haven't Devanagari fonts. Why not?
Comment 4 Gustavo Noronha (kov) 2010-02-15 17:56:31 PST
Created attachment 48783 [details]
test in GTK+

This is what I get in the GTK+ port
Comment 5 mitz 2010-02-16 07:27:49 PST
(In reply to comment #4)
> Created an attachment (id=48783) [details]
> test in GTK+
> 
> This is what I get in the GTK+ port

I am going to try to make the test less dependent on other glyphs and the behavior of the complex text implementation.
Comment 6 mitz 2010-02-16 08:09:44 PST
Created attachment 48815 [details]
Make the test less dependent on the complex text implementation

Is this better?
Comment 7 Csaba Osztrogonác 2010-02-16 16:54:36 PST
(In reply to comment #6)
> Created an attachment (id=48815) [details]
> Make the test less dependent on the complex text implementation
> 
> Is this better?

I tested your patch on Qt buildbot and the test passes. Thx.
Comment 8 Eric Seidel (no email) 2010-02-17 16:22:59 PST
Comment on attachment 48815 [details]
Make the test less dependent on the complex text implementation

Based on the above comments from Qt, this looks good.
Comment 9 mitz 2010-02-17 16:29:06 PST
Landed in r54912. Didn’t unskip the tests though. Leaving it up to the reporter to do, verify and close the bug.
Comment 10 Csaba Osztrogonác 2010-02-17 23:42:55 PST
(In reply to comment #9)
> Landed in r54912. Didn’t unskip the tests though. Leaving it up to the reporter
> to do, verify and close the bug.

Unskipped for Qt port by http://trac.webkit.org/changeset/54936 .
Comment 11 Gustavo Noronha (kov) 2010-02-18 10:59:17 PST
And for GTK+ in r54975. Thanks!