Обсуждение: experimental Unicode support

I'm attaching a few tweaks to build psqlodbc with Unicode support on
Unix.  I changed some code that assumes two-byte SQLWCHAR and I changed
configure.ac and Makefile.am, as necessary.

I tested this with unixODBC 2.2.7 on SuSE Linux 8.2.  I built unixODBC
--with-iconv-ucode-enc=UCS-4LE and with SQL_WCHART_CONVERT defined, so that
SQLWCHAR is wchar_t.  I have not tested it with UCS-2 / unsigned short.

I also tested this with iODBC 3.51.1, which always uses wchar_t on Unix.

One caveat, if anyone actually plays with this: sqlext.h should include
sqlucode.h, so I haven't patched psqlodbc to do so.  The next release of
unixODBC will contain this fix, and I've mentioned it to the iODBC maintainer.

I realize that the demand for Unicode ODBC on Unix for PostgreSQL is pretty
small.  Nonetheless, I've learned a few things poking around the code, and
I figured someone might be interested.

Hi John,

I haven't tried this on a Unix system yet, but unfortunately it does
break the Win32 Unicode build:

Building the PostgreSQL Unicode 3.0 Driver for Win32...

No configuration specified. Defaulting to Unicode30.

        cl.exe /nologo /MT /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D
"_WINDOWS" /D "_MBCS" /D "_USRDLL" /D "PSQLODBC_EXPORTS" /D
"ODBCVER=0x0300" /D UNICODE_SUPPORT" /D "DRIVER_CURSOR_IMPLEMENT" /D
"WIN_MULTITHREAD_SUPPORT"  /Fp".\Unicode30\psqlodbc.pch" /YX
/Fo".\Unicode30\\" /Fd".\Unicode30\\" /FD /c win_unicode.c
win_unicode.c
win_unicode.c(28) : error C2059: syntax error : 'type'
NMAKE : fatal error U1077: 'cl.exe' : return code '0x2'
Stop.

Regards, Dave.

> -----Original Message-----
> From: Jon Willeke [mailto:willeke@users.sourceforge.net]
> Sent: 02 February 2004 23:23
> To: pgsql-odbc@postgresql.org
> Subject: [ODBC] experimental Unicode support
>
> I'm attaching a few tweaks to build psqlodbc with Unicode
> support on Unix.  I changed some code that assumes two-byte
> SQLWCHAR and I changed configure.ac and Makefile.am, as necessary.
>
> I tested this with unixODBC 2.2.7 on SuSE Linux 8.2.  I built
> unixODBC --with-iconv-ucode-enc=UCS-4LE and with
> SQL_WCHART_CONVERT defined, so that SQLWCHAR is wchar_t.  I
> have not tested it with UCS-2 / unsigned short.
>
> I also tested this with iODBC 3.51.1, which always uses
> wchar_t on Unix.
>
> One caveat, if anyone actually plays with this: sqlext.h
> should include sqlucode.h, so I haven't patched psqlodbc to
> do so.  The next release of unixODBC will contain this fix,
> and I've mentioned it to the iODBC maintainer.
>
> I realize that the demand for Unicode ODBC on Unix for
> PostgreSQL is pretty small.  Nonetheless, I've learned a few
> things poking around the code, and I figured someone might be
> interested.
>

I haven't looked at the code, yet, but I'm pretty sure I know the
problem.  I added the HAVE_ISWASCII condition to make win_unicode.c work
for Linux, which doesn't support this extension.  However, Windows uses a
different build system, so I need to guard it with HAVE_CONFIG, or
something like that.

I'll fix this and try to find a Windows build machine.

At 09:06 AM 2/3/2004 +0000, Dave Page wrote:
>I haven't tried this on a Unix system yet, but unfortunately it does
>break the Win32 Unicode build:
>
>Building the PostgreSQL Unicode 3.0 Driver for Win32...
>
>No configuration specified. Defaulting to Unicode30.
>
>         cl.exe /nologo /MT /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D
>"_WINDOWS" /D "_MBCS" /D "_USRDLL" /D "PSQLODBC_EXPORTS" /D
>"ODBCVER=0x0300" /D UNICODE_SUPPORT" /D "DRIVER_CURSOR_IMPLEMENT" /D
>"WIN_MULTITHREAD_SUPPORT"  /Fp".\Unicode30\psqlodbc.pch" /YX
>/Fo".\Unicode30\\" /Fd".\Unicode30\\" /FD /c win_unicode.c
>win_unicode.c
>win_unicode.c(28) : error C2059: syntax error : 'type'
>NMAKE : fatal error U1077: 'cl.exe' : return code '0x2'
>Stop.

I'm attaching an updated patch.  I haven't yet tested it on Windows, but
the fix is simple enough.

I have been unable to build psqlodbc satisfactorily with the PostgreSQL
config files, as the configure script fails to find support for shared
libraries.  I did a quick test with a statically linked client.

When I stripped out the PGAC* macros and inlined PGAC_VAR_INT_TIMEZONE, I
was able to build and test a shared library.  (I have not included those
changes.)

At 2/3/2004, Jon Willeke wrote:
>I haven't looked at the code, yet, but I'm pretty sure I know the
>problem.  I added the HAVE_ISWASCII condition to make win_unicode.c work
>for Linux, which doesn't support this extension.  However, Windows uses a
>different build system, so I need to guard it with HAVE_CONFIG, or
>something like that.
>
>I'll fix this and try to find a Windows build machine.

> -----Original Message-----
> From: Jon Willeke [mailto:willeke@users.sourceforge.net]
> Sent: 03 February 2004 20:02
> To: pgsql-odbc@postgresql.org
> Subject: Re: [ODBC] experimental Unicode support
>
> I'm attaching an updated patch.  I haven't yet tested it on
> Windows, but the fix is simple enough.

Thanks John - that got it.

Patch applied :-)

Regards, Dave.