Don't ignore RTL MARK and RTL OVERRIDE unicode characters

Don't ignore RTL MARK and RTL OVERRIDE unicode characters

In r/174133 unicode characters RIGHT-TO-LEFT MARK (0x200F) and
RIGHT-TO-LEFT OVERRIDE (0x202E) were added to complexCodePathRanges,
causing them to be ignored, making some tests fail.

See https://codereview.chromium.org/276883002

BUG=256333

[behdad_google, 2014-07-20 04:56:25]

I don't understand this patch.

The patch that originally went in to "ignore" these characters simply means that
they are not rendered.  Which is correct behavior.  However, I suppose you
observed a bidi behavior change?  Please elaborate and demonstrate the problem
you are fixing.

Thanks

[ingemara, 2014-07-21 06:31:34]

On 2014/07/20 04:56:25, behdad_google wrote:
> I don't understand this patch.
> 
> The patch that originally went in to "ignore" these characters simply means
that
> they are not rendered.  Which is correct behavior.  However, I suppose you
> observed a bidi behavior change?  Please elaborate and demonstrate the problem
> you are fixing.
> 
> Thanks

Well, seems to me that by "ignoring" to render them included ignoring to
evaluate them as well. The demonstration you're requesting is included in the
review as a LayoutTest (bidi-007). It's the same as
http://www.hixie.ch/tests/adhoc/css/box/inline/bidi/007.html, which as of the
original patch renders incorrectly, because the RTL OVERRIDE is ignored (not
only in rendering)

RTL MARK or OVERRIDE never did display as '&', which, if I understand the
original bug correctly, was the issue of 'Nominal Digit Shape'.

[behdad_google, 2014-07-21 15:02:58]

On 2014/07/21 06:31:34, ingemara wrote:
> On 2014/07/20 04:56:25, behdad_google wrote:
> > I don't understand this patch.
> > 
> > The patch that originally went in to "ignore" these characters simply means
> that
> > they are not rendered.  Which is correct behavior.  However, I suppose you
> > observed a bidi behavior change?  Please elaborate and demonstrate the
problem
> > you are fixing.
> > 
> > Thanks
> 
> Well, seems to me that by "ignoring" to render them included ignoring to
> evaluate them as well. The demonstration you're requesting is included in the
> review as a LayoutTest (bidi-007). It's the same as
> http://www.hixie.ch/tests/adhoc/css/box/inline/bidi/007.html, which as of the
> original patch renders incorrectly, because the RTL OVERRIDE is ignored (not
> only in rendering)

Ok, then someone needs to investigate why that is.

[ingemara, 2014-09-26 13:54:23]

On 2014/07/21 15:02:58, behdad_google wrote:
> On 2014/07/21 06:31:34, ingemara wrote:
> > On 2014/07/20 04:56:25, behdad_google wrote:
> > > I don't understand this patch.
> > > 
> > > The patch that originally went in to "ignore" these characters simply
means
> > that
> > > they are not rendered.  Which is correct behavior.  However, I suppose you
> > > observed a bidi behavior change?  Please elaborate and demonstrate the
> problem
> > > you are fixing.
> > > 
> > > Thanks
> > 
> > Well, seems to me that by "ignoring" to render them included ignoring to
> > evaluate them as well. The demonstration you're requesting is included in
the
> > review as a LayoutTest (bidi-007). It's the same as
> > http://www.hixie.ch/tests/adhoc/css/box/inline/bidi/007.html, which as of
the
> > original patch renders incorrectly, because the RTL OVERRIDE is ignored (not
> > only in rendering)
> 
> Ok, then someone needs to investigate why that is.

So, what happened here? I see that a revert was commited
(https://codereview.chromium.org/463713002) not too long ago, doing practically
the same thing suggested in this review (sans layouttests).

[behdad_google, 2014-09-30 19:39:37]

On 2014/09/26 13:54:23, ingemara wrote:
> On 2014/07/21 15:02:58, behdad_google wrote:
> > On 2014/07/21 06:31:34, ingemara wrote:
> > > On 2014/07/20 04:56:25, behdad_google wrote:
> > > > I don't understand this patch.
> > > > 
> > > > The patch that originally went in to "ignore" these characters simply
> means
> > > that
> > > > they are not rendered.  Which is correct behavior.  However, I suppose
you
> > > > observed a bidi behavior change?  Please elaborate and demonstrate the
> > problem
> > > > you are fixing.
> > > > 
> > > > Thanks
> > > 
> > > Well, seems to me that by "ignoring" to render them included ignoring to
> > > evaluate them as well. The demonstration you're requesting is included in
> the
> > > review as a LayoutTest (bidi-007). It's the same as
> > > http://www.hixie.ch/tests/adhoc/css/box/inline/bidi/007.html, which as of
> the
> > > original patch renders incorrectly, because the RTL OVERRIDE is ignored
(not
> > > only in rendering)
> > 
> > Ok, then someone needs to investigate why that is.
> 
> So, what happened here? I see that a revert was commited
> (https://codereview.chromium.org/463713002) not too long ago, doing
practically
> the same thing suggested in this review (sans layouttests).

We reverted the whole series.  eae@ is working on an actual fix.  For now,
consider these abandoned and start from an actual reproducible bug.