214743 - remove unnecessary/ununsed jamo fallback encoder/decoder helper functions in usca/ugen.c

Status: RESOLVED FIXED

(Core :: Internationalization, defect)

Description

[Jungshik Shin]

Back in late 1990's, most Korean fonts for X11 came in ksc5601.1987-[01]
encoding and could cover only 2,350 syllables out of 11,172 complete syllables
(for modern orthography). So, I asked ftang to put in the fallback code to 
render  the rest 8,822 syllables with sequences of 2 or 3 Jamos (vowels and
consonants).  

Even after X11 fonts with broader coverage (like Sun's ksc5601.1992-3 and
iso10646-1 fonts in XFree86), I just left them alone because it wouldn't hurt. 
Users can just choose fonts with broad coverage to avoid fallback rendering.

Today, I found (talking to a Korean HP engineer) that HP/UX has rather unique
combinations of Korean fonts:

1.   ksc5601.1987-1 : covers the repertoire of 
                    KS X 1001:1998(KS C 5601-1987)
2.   ucs2-ko (XLFD to be confirmed): iso10646-1-like fonts with
                     glyphs for 8,822 precomposed syllables ONLY

That is, the repertoire of #1 and #2 are mutually exclusive.
  
When font #1 is selected, 8,822 syllables are rendered with the fallback code
instead of rendering them with font #2. If font #2 is selected, it would take 
longer to find a font for a lot more frequently used 2,350 syllables because
iso10646-1-like fonts(#2) have to be searched for first (which always fails)
before Mozilla tries #1. 

Because these days most Linux/FreeBSD users use Xft build or they have
ksc5601.1992-3 fonts (actually truetype fonts presented via xfs as
ksc5601.1992-3 fonts) and/or iso10646-1 fonts, we don't need the fallback
rendering any more. As for commercial Unix, Solaris has no trouble at all (they
have ksc5601.1992-3) and I guess AIX 5.x should be fine as well. On the other
hand, removing helps HP/UX (and possibly other commercial Unix) in addition to
cutting down the code size a bit.

Comment 1

[Jungshik Shin]

I realized that uscan.c (used for 'ToUnicode' conversion) also has a 'scannmer'
function for 8byte fallback encoding of 8,822 syllables _in GL_. Because EUC-KR
uses KS X 1001 in GR and ISO-2022-KR decoder relies on EUC-KR decoder. We don't
need that scanner, either.
 
I'm changing the OS to all from HP/UX.  

This will cut down the size further.

Comment 2

[Jungshik Shin]

Attachment: a patch

Comment 3

[Jungshik Shin]

Comment on attachment 129016 [details] [diff] [review]
a patch

asking for r/sr. 

The original file seems to have tabs in a couple of places. I'll fix the
indentation when landing.

Comment 4

[Simon Montagu :smontagu]

Comment on attachment 129016 [details] [diff] [review]
a patch

r=smontagu

Comment 5

[Jungshik Shin]

Fix was checked into the trunk. Thank you all.