RESOLVED FIXED 25698
add a test with a LTR complex script for the fix for bug 23786
https://bugs.webkit.org/show_bug.cgi?id=25698
Summary add a test with a LTR complex script for the fix for bug 23786
Jungshik Shin
Reported 2009-05-11 10:07:58 PDT
Bug 23786 was fixed a while ago and 3 layout tests were added along with the fix. Chromium used to have this problem whether it's RTL/Bidi or LTR complex scripts. With the fix, both cases work fine, but all 3 layout tests added were for the former. Although it's not likely that only the latter can regress without the former regressing as well, it may be a good idea to add another test with an LTR complex script. This patch adds a layout test with Hindi (I put 3 cases - white-space: normal, nowrap and pre - into a single test file). Sorry that I forgot to point this out when bug 23786 was still open. I meant to add a comment, but got distracted.
Attachments
add a new test (53.87 KB, patch)
2009-05-11 10:11 PDT, Jungshik Shin
abarth: review-
expected results for win added (104.89 KB, patch)
2009-06-22 15:45 PDT, Jungshik Shin
eric: review+
Jungshik Shin
Comment 1 2009-05-11 10:11:57 PDT
Created attachment 30193 [details] add a new test
Maciej Stachowiak
Comment 2 2009-05-21 21:55:45 PDT
Comment on attachment 30193 [details] add a new test Tests are good. r=me
Adam Barth
Comment 3 2009-06-01 02:02:25 PDT
Will land.
Adam Barth
Comment 4 2009-06-01 02:06:07 PDT
Sending LayoutTests/ChangeLog Adding LayoutTests/fast/text/international/hindi-whitespace.html Adding LayoutTests/platform/mac/fast/text/international/hindi-whitespace-expected.checksum Adding (bin) LayoutTests/platform/mac/fast/text/international/hindi-whitespace-expected.png Adding LayoutTests/platform/mac/fast/text/international/hindi-whitespace-expected.txt Transmitting file data ..... Committed revision 44320.
Adam Barth
Comment 5 2009-06-01 03:14:52 PDT
This test doesn't pass on windows. Reverted.
Adam Barth
Comment 6 2009-06-01 23:47:07 PDT
Comment on attachment 30193 [details] add a new test Apparently, I'm supposed to r- this to tell you that the patch needs to be changed so that the test passes on Windows.
Jungshik Shin
Comment 7 2009-06-17 16:10:52 PDT
(In reply to comment #5) > This test doesn't pass on windows. Reverted. I'm sorry I missed this completely. Do you happen to remember what's the cause of the failure? Because there's no test result for Windows (my patch only has test results for Mac), I'm not sure how it could fail on Windows. The buildbot does not have the information for the failure over 2 weeks ago. (I tried http://build.webkit.org/waterfall?last_time=1243877215 and it's gone other than the record of your checkin and rollback).
Adam Barth
Comment 8 2009-06-17 17:18:37 PDT
I'm sorry, I don't remember. The bot just went red with the test. Do you not have a window machine to test with?
Jungshik Shin
Comment 9 2009-06-22 11:29:17 PDT
I just built a Webkit on Windows, applied the patch and run webkit-test in fast/text/international. For all but one test, there's no expected results in platform/win. However, somehow run-webkit-tests script picked up the expected results for 17 of tests in that directory (including the one I'm adding) from platform/mac, compared them to the actual layout results on Windows and reported them as failures. They fail because of the font difference between Mac and Windows. Two issues/questions (if they're indeed issues, we have to deal with them in a separate bug): 1) Why does run-webkit-tests pick up the expected results from platform/mac in the absence of the expected result in platform/win? Is it supposed to do that? 2) Why does it do that for only a subset of tests in fast/text/international? Windows release (test) buildbot does not seem to do that at the moment ( http://build.webkit.org/builders/Windows%20Release%20(Tests)/builds/1737/steps/layout-test/logs/stdio ), but it may have done when my patch was landed by Adam. Anyway, I can generate and add an expected test result for Windows to my patch.
Jungshik Shin
Comment 10 2009-06-22 15:45:57 PDT
Created attachment 31684 [details] expected results for win added 1. Added the expected results for Windows. 2. Added the test to platform/gtk/Skipped so that the buildbot remains green with this test added. I guess this test will pass on gtk and it's just a matter of creating expected test results. 3. explicit set svn:mime-type to image/png for png files.
Eric Seidel (no email)
Comment 11 2009-06-23 17:44:35 PDT
Comment on attachment 31684 [details] expected results for win added This looks like just a test. And looks fine.
Eric Seidel (no email)
Comment 12 2009-06-26 02:20:08 PDT
Committing to http://svn.webkit.org/repository/webkit/trunk ... M LayoutTests/ChangeLog A LayoutTests/fast/text/international/hindi-whitespace.html M LayoutTests/platform/gtk/Skipped A LayoutTests/platform/mac/fast/text/international/hindi-whitespace-expected.checksum A LayoutTests/platform/mac/fast/text/international/hindi-whitespace-expected.png A LayoutTests/platform/mac/fast/text/international/hindi-whitespace-expected.txt A LayoutTests/platform/win/fast/text/international/hindi-whitespace-expected.checksum A LayoutTests/platform/win/fast/text/international/hindi-whitespace-expected.png A LayoutTests/platform/win/fast/text/international/hindi-whitespace-expected.txt Committed r45253 http://trac.webkit.org/changeset/45253
Note You need to log in before you can comment on or make changes to this bug.