[PATCH] [Bug 91955] Bidi algorithm: incorrect implementation of rule N2

[Mitz Pettel]

Hi,

Thanks for making the test results available to me. I see there are 
still many unnecessarily split runs (mostly when a line begins with a 
neutral type, it seems). I'll try to eliminate the splits.

Regarding the failures:
1. text-transform.html is really puzzling. Why does it single out those 
few characters? Aren't they all DirL? I'll look into it, but it's very 
strange. From your remark about the selectors bug I gather that my 
patch isn't the only difference you're testing. Could there have been a 
change to Unicode data between versions?
2. sec09-10a.htm - the "OUT" looks exactly like Firefox and MSIE render 
it, so I'd say it's a success, actually!
3. sec09-10b.htm - the "OUT" looks more like Firefox renders it: [right 
cl1] and [right cl3] are like Firefox. [right cl2] is broken both in 
"BASE" and in "OUT" (it should look something like
	g f e d c b a 31 21 11 01 9 8 7 6 5 4 3 2 1 [2lc thgir]
and indeed that's what it looks like in Firefox).
4. css3-modsel-d4.html - the selectors bug.

Regarding the fixes: cool! What I don't get is why they're flagged as 
RESOLVED/FIXED in the bug database if they weren't really fixed.

Regards,
-- Mitz Pettel

On Nov 4, 2004, at 4:20 PM, Stephan Kulow wrote:

> Am Thursday 04 November 2004 00:46 schrieb Mitz Pettel:
>> On Nov 3, 2004, at 10:20 PM, Stephan Kulow wrote:
>>
>>> Am Mittwoch 03 November 2004 17:21 schrieb Mitz Pettel:
>>>> I see. I think changing line 859 from
>>>>                       if( bidi.status.eor != QChar::DirL ) {
>>>> to
>>>>                       if( bidi.status.eor != QChar::DirL &&
>>>> bidi.status.eor!=QChar::DirEN ) {
>>>> should take care of most of these.
>>>
>>> I will try tomorrow. Thanks for the fix.
>>
>> Forget it. I tried it myself and it failed one of the test cases.
>> Here's the latest revision of the patch. This one looks good on
>> bidi.html, and should not generate excessive runs (however I didn't
>> test that part).
>
> Hi!
>
> I get two failures in test cases and two fixes (I listed this morning).
>
> See yourself at http://ktown.kde.org/~coolo/regression/output.bidi/
> (the RP are the ones with visible changes)
>
> There is a selector test case that fails now where I can hardly believe
> it's the fault of this patch.
>
> Greetings, Stephan