Fix font family based fallback font selection

Fix font family based fallback font selection

Introducing FontFaceCreationParams, used for creating a platform 
font face in the old way based on family, and in the fixed way, based on 
ttf filename and collection index.
Created a font based on Gardiner hieroglyphs font, same family name,
but two font files, one containing a bug glyph, one containing a cat
glyph. When the fonts are installed on the system, content_shell should
show both when opening the family-fallback-gardiner.html test case.
This works on Firefox, but on Chromium content_shell we used to only see
the bug symbol.

BUG=285303
R=eseidel,behdad,eae

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=176765

[behdad_google, 2014-06-02 20:02:09]

lgtm

Excellent test case!

[Dominik Röttsches, 2014-06-05 14:25:37]

This needs a bit more work, still some assertion failures in the old code,
however as http://roettsch.es/initialfix.png shows, it's basically working
nicely. (Not for the ttc case yet, but for the multiple ttfs).

I just wanted to ask for your (Eric, Emil?) feedback whether the
FontFaceCreationParams approach is the right direction?

Also, better suggestions for the new class name are welcome:
FontFaceCreateParams?

Thanks!

[eae, 2014-06-05 16:38:12]

Seems entirely reasonable to me. Not thrilled about the name
FontFaceCreationParams but can't come up with anything better either.

Thanks for doing this!

[behdad_google, 2014-06-05 18:54:24]

lgtm

Nice.  Thanks.

[Dominik Röttsches, 2014-06-06 14:55:42]

The patch is more or less ready, but there a couple of caveats:
a) void FontCache::getFontFamilyForCharacter, FontCache::SimpleFontFamily and
WebFontFamily need to be renamed since they do not only carry family info, we
might even get rid of the family info in these structures completely. I would
like to land such a rename separately.
b) I'll upload a separate IPC fix on the Chromium side so that this works not
only in --single-process after the above rename.
c) LayoutTesting currently does not work because: crbug.com/381684

[eae, 2014-06-06 15:11:22]

On 2014/06/06 14:55:42, Dominik Röttsches wrote:
> The patch is more or less ready, but there a couple of caveats:
> a) void FontCache::getFontFamilyForCharacter, FontCache::SimpleFontFamily and
> WebFontFamily need to be renamed since they do not only carry family info, we
> might even get rid of the family info in these structures completely. I would
> like to land such a rename separately.
> b) I'll upload a separate IPC fix on the Chromium side so that this works not
> only in --single-process after the above rename.
> c) LayoutTesting currently does not work because: crbug.com/381684

Ok, sounds like a good plan.

[Dominik Röttsches, 2014-06-17 12:58:48]

On 2014/06/06 15:11:22, eae wrote:
> On 2014/06/06 14:55:42, Dominik Röttsches wrote:
> > The patch is more or less ready, but there a couple of caveats:
> > a) void FontCache::getFontFamilyForCharacter, FontCache::SimpleFontFamily
and
> > WebFontFamily need to be renamed since they do not only carry family info,
we
> > might even get rid of the family info in these structures completely. I
would
> > like to land such a rename separately.
> > b) I'll upload a separate IPC fix on the Chromium side so that this works
not
> > only in --single-process after the above rename.

a) & b) are mostly complete now (still need to remove the old WebFontFamily.h in
a separate patch, but the rename is complete).
We can try to land this now, I'll run it through the bots.

[eae, 2014-06-17 14:56:25]

LGTM

[Dominik Röttsches, 2014-06-17 19:08:26]

.gypi fix.

[Dominik Röttsches, 2014-06-18 13:41:34]

1) Ugh, the bots and I were hitting the walls of the sandbox. :-(

SkTypeface::CreateFromFile is blocked in the renderer sandbox. I did not catch
that during my testing, since I was using --no-sandbox or --single-process.

Behind the scenes SkFontMgr_fontconfig is using a struct called FontIdentity:
When it's querying the sandbox support process for a font by family for example,
it receives a FontIdentity with an fId from the sandbox support process and the
sandbox support process keeps track of the filename corresponding to that id.
Later, when the font is instantiated, the FontIdentity is used and the sandbox
support process opens a filehandle which is then handed back to the renderer.

We just received a filename but no such id - and with the sandbox active that's
not enough to instantiate the font.

In https://codereview.chromium.org/340073002 I am extending WebFallbackFont by
fciId, and in https://codereview.chromium.org/336153003/ I am filling the fciId
and keeping track of the corresponding filename.

Once these two changes land, we can proceed with this one here. I am
instantiating my own SkFontConfigInterface to open a stream by fciId, and then
pass this one to SkTypeface::CreateFromStream. This way we're using the same
identifiers and the same IPC mechanism for opening the right file handles.

2) I simplied the FontCacheSkiaWin.cpp side - since that will never use
instantiation by fciId - only by family.

[Dominik Röttsches, 2014-06-23 13:01:11]

Windows link fix.

[Dominik Röttsches, 2014-06-23 14:42:17]

The CQ bit was checked by dominik.rottsches@intel.com

[Dominik Röttsches, 2014-06-23 14:42:46]

The CQ bit was unchecked by dominik.rottsches@intel.com

[Dominik Röttsches, 2014-06-23 14:45:56]

Emil, I am optimistic this can land now. Would you have a moment to take a look
at the latest (and hopefully last) change, using CreateFromStream and
SkFontconfigInterface to open the stream that is created from file outside the
sandbox? If it does look alright to you, feel free to tick the commit box.
Thanks!

[eae, 2014-06-23 15:07:01]

LGTM, Thanks for doing this Dominik!

[Dominik Röttsches, 2014-06-23 16:21:33]

The CQ bit was checked by dominik.rottsches@intel.com

[commit-bot: I haz the power, 2014-06-23 16:22:30]

CQ is trying da patch. Follow status at

https://chromium-status.appspot.com/cq/dominik.rottsches@intel.com/307243002/...

[commit-bot: I haz the power, 2014-06-23 17:23:59]

Change committed as 176765