0. Which version the attached patch applies to
2.2.19
1. What this patch supposed to fix
1-1. Wrong case manipulation in namei.c
From 2.2.18 namei.c uses toupper(), islower() to make uppercase
shortname, and since it assumes the character range of 0xe0-0xff as
lowercase, it broke the DBCS(double byte character set) character.
This patch tries to acknowledge the codepage charset it currently uses
and if the charset is one of DBCS charset, it uses (if (c>='a' ||
c<='z') c = c - 32) instead. Non-DBCS charsets are not affected.
1-2. Spliced doublebyte character shortname
If filename longer than 8 byte base has doublebyte character in 6th and
7th place, like
'a' 'b' 'c' 'XY' 'XY' 'XY' '.' 'b' 'm' 'p'
(X and Y means first and latter half of double byte character)
Current implementation makes shortname like
'A' 'B' 'C' 'XY' 'X' '~' '1' '.' 'B' 'M' 'P'
which is different from DBCS windows implementation
'A' 'B' 'C' 'XY' '~' '1' '.' 'B' 'M' 'P'
2. Tests
Copy entire korean win95,98,Me fresh-installed partition to samba share
to make original file set and make other computer that has both ext2
linux partition and vfat partition. Boot this computer in linux, mount
vfat partition and nfs partition (for original filesets) and copy these
to vfat partition. Finally reboot to windows and run dir, scandisk,
etc..
without patch, scandisk complains at several spots and dir shows like
this.
dir c:\windows\*.bmp
물뮷완 BMP 190 96-10-30 1:00 물방울.bmp
(notice the difference in shortname and longname)
WIN95_1 BMP 240 96-10-30 1:00 WIN95_1.bmp
(correct name of this file is WIN95 설치 화면.bmp)
with patch applied scandisk doesn't complain anymore.
3. Feedback
I welcome feedback from anyone interested, especially those from other
DBCS country. If you have comments, feedback please mail me.
This archive was generated by hypermail 2b29 : Wed Oct 31 2001 - 21:00:25 EST