How to get wchar_t support under Cygwin?

[Bruce Reid]

My reliance on GCC includes the following:

1) App must be able to be built/run on: Linux, Solaris, and Cygwin
2) App needs Xerces-C XML parser, which includes unicode support, and I 
must be able to build/link the library on all three platforms mentioned 
above.

I did initial development on Solaris, then later ported to Linux. Both 
of these were relatively painless (both configurations are well-
supported by the Xerces development team; since Cygwin is not so well 
supported, I'm treading a path not many others have used, according to 
info available from the Xerces-C mailing list.)

Prior to the recent introduction of the Xerces-C element into my 
application, I was able to build it successfully under Cygwin using the 
gcc 2.95 binary that came bundled with Cygwin. Now, when trying to 
build the Xerces library on Cygwin, I'm running into trouble with a 
file that tries to include wctype.h

Thinking that improvements since 2.95 may have solved the problem, I 
obtained the latest GCC release (3.0.4, full package), and built it 
successfully, though the build process produced a message indicating 
that wctype.h is missing, and wchar_t support is disabled. I have 
searched from the Cygwin root directory (which encompasses GCC) for any 
file containing the text 'wctype', and have found one instance of 
wctype.h, and several other related files in various subdirectories.

Still, I can't successfully compile a source file containing #include 
<wctype.h> (or #include <cwctype>)

I've attached a text file contining pertinent info about my platform, 
the GCC configure/build process, the locations of the '*wctype*' files, 
errors produced by attempting to compile my test file, code snippets 
extracted from the troublesome source file necessitating the need for 
wctype.h, etc.

I'm not sure why this was no problem at all on Solaris or Linux, but is 
causing so much trouble on Cygwin. I've looked through the archives for 
discussions of this topic, and unless I've missed something, all of the 
related patches, etc. seem to be in place in my installation.

If anyone can tell me whether it's possible to get GCC to properly 
handle this wide char business on my Cygwin platform, and if so, what 
steps I need to take to do so, I will be extremely grateful.

Sincerely,
Bruce Reid
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: gccinfo.txt
URL: <http://gcc.gnu.org/pipermail/libstdc++/attachments/20020309/f0cb5c09/attachment.txt>