[Bug 17500] New: regression: libs/wine: best-fit should be renamed to "not so" best-fit

http://bugs.winehq.org/show_bug.cgi?id=17500

Summary: regression: libs/wine: best-fit should be renamed to
                    "not so" best-fit
           Product: Wine
           Version: 1.1.15
          Platform: Other
        OS/Version: other
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: -unknown
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]
                CC: [e-mail]

...at least for cp932.
But I'm getting ahead of myself.

Just recently I've noticed that a few programs, that required Japanese
locale (not every single one, just a few) stopped running, or more exactly,
terminated with some conversion error.
First, I went through the tarballs. Last good was 1.1.13,
but I didn't have an idea at first, what's exactly wrong.
So, I've done a bisect. Checking out a whole tree take a long time -
really annoying, if you're not planning to keep it,
but without it, I wouldn't get so far.

After I've done a few bisects, build process failed suddenly
with following error:
../../tools/wrc/wrc --nostdinc -I. -I. -I../../include -I../../include 
-D__WINESRC__ -D_KERNEL32_  -fokernel.res kernel.rc
Source: ¢̤ a2 cc a4 eb
Unicode:  5341 6708
Back: ¤Q¤ë a4 51 a4 eb
nls/cht.nls:84:16: Error: String ¢Ì¤ë does not convert identically to
Unicode and back in codepage 950. Try using a Unicode string instead

It messed something with my locale, while it failed (my normal locale is
pl_PL.UTF-8). Luckily, I managed to get far enough, that I could notice the
probably buggy commit in webgit and reverted it manually against 1.1.15
tarball.
That fixed the regression.

As far as I could check, those programs were doing something like:
MultiByteToWideChar(CP_ACP, MB_PRECOMPOSED|MB_ERR_INVALID_CHARS, s, -1, NULL,
0);

The commit I'm talking about is : 8e16e7871063df35ca3a6de6fb165767c00bd087