[unixODBC-support] Unicode and Wide Char strings

Neha Avasthy neha.avasthy at gmail.com
Tue Oct 19 10:40:35 BST 2010


Hi ,

I have a basic question w.r.t wide char literals. In the platforms I am
working on WIDE chars are of size 2 bytes however the L"string" literals are
of size wchar_t which is essentially 4 bytes.

My application however expects UTF16 and thats why to make the following
assignment work in the code

TCHAR    pSql[] = _T("SELECT NULL FROM DUAL");

typedef signed short        TCHAR;
typedef long int        wchar_t;

I am using the -fshort_wchar flag in the compile line .

The gcc compiler handles L as wchar_t string where wchar_t is of 2 bytes on
Windows and of 4
bytes on Unix. As application expects the unicode strings to be of type
unsigned short - 2 bytes so to convert 2 bytes unicode string the
-fshort-wchar flag was added.

However I figured out that

-fshort-wchar seems to be a compiler hack only available with gcc.
-fshort-wchar forces compiler (gcc) to treat wide charater literals as
unsigned short. Thats not a  part of language standards anyways.

This is has made my code non portable as equivalent flags are not available
on other native compilers for Solaris, HPI and AIX.

If I remove the -fshort_wchar flag from the compile line I get errors like

 error: a value of type "long [22]" cannot be used to initialize an entity
of type "TCHAR={signed short} []"
    TCHAR    pSql[] = _T("SELECT NULL FROM DUAL");

Can somebody tell me a way of dealing with this conversion. I have to ensure
that the L"string" treatment is that of a 2 byte quantity.

Thanks
-Neha
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.unixodbc.org/pipermail/unixodbc-support/attachments/20101019/5ce6079b/attachment.html>


More information about the unixODBC-support mailing list