Re-enable vertical flow in SimplePath

Re-enable vertical flow in SimplePath

This patch changes the default code path of vertical flow text back to the
simple code path. This was changed to complex path in crbug.com/408992, but
it turned out that it degrades the performance by 100-150%.

The performance work for the complex path is under progress in
crbug.com/446011, but in order to ship better code, this patch makes a
temporary switch before the stable release. The complex path will be
re-enabled once we're back to work-mode.

BUG=445589
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=190462

[kojii, 2015-02-18 16:29:08]

Patchset #2 (id:20001) has been deleted

[kojii, 2015-02-18 19:40:33]

kojii@chromium.org changed reviewers:
+ eae@chromium.org

[kojii, 2015-02-18 19:40:34]

eae@, PTAL.

linux_blink_dbg is failing for worker, I'll re-run the test tomorrow.

Note that I had to change two tests as Failure since they were ImageFailure in
crbug.com/453002 but now text fail too. I'll change them back on revert, but FYI
in case.

[eae, 2015-02-18 20:30:33]

On 2015/02/18 19:40:34, koji wrote:
> eae@, PTAL.
> 
> linux_blink_dbg is failing for worker, I'll re-run the test tomorrow.
> 
> Note that I had to change two tests as Failure since they were ImageFailure in
> crbug.com/453002 but now text fail too. I'll change them back on revert, but
FYI
> in case.

LGTM :(

[kojii, 2015-02-19 01:40:33]

The CQ bit was checked by kojii@chromium.org

[commit-bot: I haz the power, 2015-02-19 03:25:53]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/932033002/60001

[kojii, 2015-02-19 12:28:31]

The CQ bit was unchecked by kojii@chromium.org

[kojii, 2015-02-19 12:28:40]

The CQ bit was checked by kojii@chromium.org

[commit-bot: I haz the power, 2015-02-19 12:31:35]

Committed patchset #3 (id:60001) as
https://src.chromium.org/viewvc/blink?view=rev&revision=190462

[bungeman-skia, 2015-02-19 18:49:51]

Not sure where to point this out, but fewer and fewer of the installed fonts on
Mac use the 'GSUB' table 'vert' feature for vertical substitutions. Instead,
they're using AAT tables ('mort'/'morx') or other shaping mechanisms. As a
result, most 'simple' vertical shaping is broken on Mac by this (and has been
since M40). It doesn't matter how fast it draws if the wrong thing is drawn. (In
other words, I'm sad to see 'simple' vertical text on Mac is going to continue
to be broken in M42.)

[kojii, 2015-02-21 12:35:39]

On 2015/02/19 18:49:51, bungeman1 wrote:
> Not sure where to point this out, but fewer and fewer of the installed fonts
on
> Mac use the 'GSUB' table 'vert' feature for vertical substitutions. Instead,
> they're using AAT tables ('mort'/'morx') or other shaping mechanisms. As a
> result, most 'simple' vertical shaping is broken on Mac by this (and has been
> since M40). It doesn't matter how fast it draws if the wrong thing is drawn.
(In
> other words, I'm sad to see 'simple' vertical text on Mac is going to continue
> to be broken in M42.)

Thank you for the feedback! Not many people talks about non-OpenType
technologies in CSS WG, so feedback like this is greatly appreciated.

Mostly from my curiosity, but could you tell me which scripts they are?

If they are one of horizontal-only scripts as defined in CSS:
http://dev.w3.org/csswg/css-writing-modes-3/#script-orientations
I strongly recommend to use "-webkit-text-orientation: sideways-right".

Currently, Blink is NOT comformant to the spec in terms of inline alignment in
vertical flow, and I'm hearing requests to fix this from vertical scripts
people. If we fix this, however, it will have unpleasant side effects for
horizontal-only scripts, and this property will protect you from such fix. See
the wiki below.
https://wiki.csswg.org/spec/css3-writing-modes/author-css#horizontal-only-scr...

And if I understand correctly, when it's specified, Blink can render AAT
correctly.

I'd like to make Blink good for as much people as possible in all combinations,
but it'll take time. There are reasons why vertical scripts people needs
performance, and I don't want to pick one by sacrificing the other. I appreciate
your patience.

Note that, if things don't work even with "-webkit-text-orientation:
sideways-right", I understand it's higher priority than when not specified.
Please do let me know if that's the case.