Name: krT82822 Date: 01/06/2000


(see also 4185525)

java version "1.2.2"
Classic VM (build JDK-1.2.2-W, native threads, nojit)

Java is supposed to finally eliminate the whole character set problem.
Everything is maintained in Unicode.
When put together with Windows NT, whose filesystem is in Unicode, and
whose filesystem APIs are available in Unicode, there should be no
more problems of writing servers whose filesystem manipulation
doesn't work depending on the client's codeset.
However, I have found that the JVM under NT converts all filesystem
names from Unicode in to the ANSI codeset, and then calls the narrow
versions of the filesystem APIs.
So java will convert Unicode to ANSI, and then the Win32 subsystem will
promptly convert back ANSI to Unicode and call the underlying NT wide
filesystem calls. What a waste!
Yes, I understand, you have one generic Win32 JVM which works on win95/98
as well as NT, and the Win95/98 doesn't support the wide calls.
This just makes it a little harder, but its standard Win32 programming
techniques at this point. There really aren't that many file system APIs.
(Review ID: 99666)
======================================================================

user later noted:
(I have just been pointed at bug #4185525 which is much more long winded
but is essentially the same bug.)

Thus, we'll dupe this bug to # 4185525.