LayoutTests/fast/dom/Range/range-expand.html always fails on Japanese SnowLeopard.
Created attachment 54689 [details] Test result diff
range expand uses ICU's text break iterator. Is the ICU's text break iterator different for OS X 10.5 and SnowLeopard? Or is it different for Japanese SnowLeopard? Since ICU does not support Chinese segmentation, the Chinese word are segmented character by character. Looks like in Japanese SnowLeopard, "在甲型" is segmented as one word. For Hebrew testing, the difference seems that in Japanese SnowLeopard, when segment into sentence and block, 0x0A (Line Feed) is added at the end of sentence/block. If the difference is due to platform difference. I will need to make the result as Mac OS X 10.5 only.
> Is the ICU's text break iterator different for OS X 10.5 and SnowLeopard? This test doesn't fail on Snow Leopard build bot, so this shouldn't be because of a different ICU version.
(In reply to comment #3) > > Is the ICU's text break iterator different for OS X 10.5 and SnowLeopard? > > This test doesn't fail on Snow Leopard build bot, so this shouldn't be because > of a different ICU version. It's curious. What's the difference between my machine and the bot? I changed the system locale to English. The test still failed. I tried both of 32bit build and 64bit build. The test failed on both settings. My OS version is Mac OS X 10.6.3 (10D573), my Xcode version is 3.2.2.