My latest megapatch, was Re: [Rlib-devel] RLIB-1.3.2 questions, problems and an ODBC fix

[Zoltan B. <zb...@du...>]

Hi,

here's my last work against the latest CVS.

Two main changes since the last one:

- ODBC detection on POSIX systems considers $libdir before
   the hardcoded check for /usr/lib so on my x86-64 /usr/lib64
   is used for ODBC libs. rlib.pc no longer contains -L/usr/lib
   on my system so 64-bit compiled apps linked against rlib
   doesn't complain about incompatible libs in /usr/lib

- Warning fixes so compiling with "gcc4 -Wall -Werror" succeeds.
   GCC 4.0 produces many more warnings than previous versions.
   And guess what? It fixes the crashes on my 64-bit system finally.
   This was the last thing I could try as valgrind doesn't run on AMD64
   yet. So 1.3.3 will mark both Windows and at least Linux/AMD64 ports.

Cheers! :-)

One thing, though. Are you sure about your change in
rlib_set_datasource_encoding()?
As it stands in the CVS, line 407 of libsrc/api.c is

	tif->info.encoder =3D rlib_charencoder_new(encoding, "UTF-8");

rlib_charencoder_new() takes the parameters in (to, from) order.

I changed it in my patch to the order I originally proposed:

	tif->info.encoder =3D rlib_charencoder_new("UTF-8", encoding);

So when the datasource is read, the strings are converted from the set
encoding to UTF-8 and it's paired with rlib_set_output_encoding() so
when outputting the internally UTF-8 strings are converted from UTF-8
to the set output encoding.

Best regards,
Zolt=E1n B=F6sz=F6rm=E9nyi