[poppler] [PATCH] 2 workarounds for a Korean font on Microsoft Windows

[mpsuzuki at hiroshima-u.ac.jp]

Hi,

On Tue, 27 Mar 2012 00:32:30 +0900
mpsuzuki at hiroshima-u.ac.jp wrote:
>Also I fixed Korean entry by changing "kana" to "hang",
>but Batang (bundled to Microsoft Windows) have other
>problems, and I will submit another set of patches.

The problems in Batang (a Korean font bundled to Microsoft
Windows) are following.

1) entries in OpenType lookup table are sorted by source GID?

FoFiTrueType assumes the entries in lookup table are
sorted by source GID. If it finds a GID greater than
source GID, the lookup process is finished as "not
found in the table".

In Batang, the entries in the lookup subtable for
"vert" feature is not sorted by source GID, so some
substitution rule could not be found. So all entries
should be scanned to use Batang.

attached "fixOTLookupForUnsortedTable_20120326.diff"
is a workaround patch for this issue.



2) default language system or not?

FoFiTrueType assumes the vertical layout feature is
defined in default language system, for specified
script. FoFiTrueType::setupGSUB() takes only script
name. It is true for SimSun, MingLiU, MS-Mincho, but
false for Batang. In Batang, "hang" script supports
"KOR " language system (in "KOR " language system,
"vert" is defined), but default language system is
not "KOR " and it has no layout feature.

attached "addSetupGSUBWithLang_20120326.diff" is a
patch to add new variant of FoFiTrueType::setupGSUB()
taking both of "script" and "language". By giving both
of "hang" and "KOR ", the vertical layout feature of
Batang can be activated. Original method is kept as
a thin wrapper of new method.

attached "useSetupGSUBWithLang_20120326.diff" is a
patch to modify GfxCIDFont::getCodeToGIDMap() to use
new variant of FoFiTrueType::setupGSUB(script, lang).

--

If more detailed analysis and testing PDF are required,
I will post.

Regards,
mpsuzuki