Re: LuaSQL for Lua 5.1

New clues! It worked (sort of): it has something to do with the use of lua stack.. when I use a non-empty table as a reference to pass into cur_fetch it works well.. if the table is empty all hell break lose.
..
row = cur:fetch ({ test="test" }, "n") -- returns values correctly
row = cur:fetch ({ }, "n") -- access violation
..

On the source code, all seems right.. the table passed as a parameter is in stack index 2 when the following line runs:
ls_sqlite3.c: 153 lua_rawset(L, 2);

If this table is empty, this like violates memory - I can´t figure out why.. same for numeric indexes (ls_sqlite3.c: 141).

Luís Santos escreveu:

[
Here goes my modified makefile:

#---------------------------------------------------
#config
#---------------------------------------------------
# Driver (leave uncommented ONLY the line with the name of the driver)
#T= mysql
#T= oci8
#T= odbc
#T= postgres
#T= sqlite
T=sqlite3

# Installation directories
# System's libraries directory (where binary libraries are installed)
LUA_LIBDIR= dependencies/org.lua.engine/lib
# System's lua directory (where Lua libraries are installed)
LUA_DIR= dependencies/org.lua.engine/lib
# Lua includes directory
LUA_INC= dependencies/org.lua.engine/include

# Lua version number (first and second digits of target version)
LUA_VERSION_NUM= 501

# OS dependent
#LIB_OPTION= -shared #for Linux
LIB_OPTION= #-bundle -undefined dynamic_lookup #for MacOS X

LIBNAME= $T.so
COMPAT_DIR= ../compat-5.1r5

# Compilation parameters
# Driver specific
######## MySQL
#DRIVER_LIBS= -L/usr/local/mysql/lib -lmysqlclient -lz
#DRIVER_INCS= -I/usr/local/mysql/include
######## Oracle OCI8
#DRIVER_LIBS= -L/home/oracle/OraHome1/lib -lz -lclntsh
#DRIVER_INCS= -I/home/oracle/OraHome1/rdbms/demo -I/home/oracle/OraHome1/rdbms/public
######## PostgreSQL
#DRIVER_LIBS= -L/usr/local/pgsql/lib -lpq
#DRIVER_INCS= -I/usr/local/pgsql/include
######## SQLite
#DRIVER_LIBS= -lsqlite
#DRIVER_INCS=
######## SQLite3
DRIVER_LIBS= -Ldependencies/org.sqlite.engine/lib -lsqlite3
DRIVER_INCS= -Idependencies/org.sqlite.engine/include
######## ODBC
#DRIVER_LIBS= -L/usr/local/lib -lodbc
#DRIVER_INCS= -DUNIXODBC -I/usr/local/include

WARN= -Wall #-Wmissing-prototypes -Wmissing-declarations -ansi
INCS= -I$(LUA_INC)
CFLAGS= -O2 $(WARN) -I$(COMPAT_DIR) $(DRIVER_INCS) $(INCS) $(DEFS)
CC= g++

# $Id: config,v 1.6 2007年04月06日 23:46:04 mascarenhas Exp $


#---------------------------------------------------
#makefile
#---------------------------------------------------
V= 2.0.3
CONFIG= ./config

include $(CONFIG)

ifeq "$(LUA_VERSION_NUM)" "500"
COMPAT_O= $(COMPAT_DIR)/compat-5.1.o
endif

OBJS= src/luasql.o src/ls_$T.o $(COMPAT_O)


SRCS= src/luasql.h src/luasql.c \
src/ls_postgres.c \
src/ls_odbc.c \
src/ls_oci8.c \
src/ls_mysql.c \
src/ls_sqlite.c

AR= ar rcu
RANLIB= ranlib


lib: src/$(LIBNAME)

src/$(LIBNAME): $(OBJS)
export MACOSX_DEPLOYMENT_TARGET="10.3"; $(CC) $(CFLAGS) -o $@ $(LIB_OPTION) $(OBJS) $(DRIVER_LIBS)

$(COMPAT_DIR)/compat-5.1.o: $(COMPAT_DIR)/compat-5.1.c
$(CC) -c $(CFLAGS) -o $@ $(COMPAT_DIR)/compat-5.1.c


dll: $T.dll

$T.exp: src/$T.def
echo ------------------------------
echo Generating Definitions - $@
echo ------------------------------
dlltool --def src/$T.def --dllname $T.dll --output-exp $T.exp --output-lib $T.a

$T.dll: $(OBJS) $T.exp
echo ------------------------------
echo Assembling DLL - $@
echo ------------------------------
$(CC) -shared -o $T.dll $(LIB_OPTION) $T.exp $(OBJS) -L$(LUA_LIBDIR) -llua $(DRIVER_LIBS) -Wall -Wl,--out-implib,$T.a
strip $T.dll


install:
mkdir -p $(LUA_LIBDIR)/luasql
cp src/$(LIBNAME) $(LUA_LIBDIR)/luasql

jdbc_driver:
cd src/jdbc; make $@

clean:
rm -f src/$(LIBNAME) src/*.o $(COMPAT_O)

# $Id: Makefile,v 1.52 2007年04月06日 23:46:04 mascarenhas Exp $

Aditionally, I have folowed the problem into the C file, and cornered the Access Violation exactly in the loop to get the values in the cur_fetch() function. It expodes on either numeric and string keys, in the very first field. Value is obtained ok from sqlite (I log it to disk) but something went wrong in the rawset() function.

I am beginning to run out of clues.. but I will poke it a little more.. :-)


Fabio Mascarenhas escreveu: [
--
Fabio Mascarenhas

On 6/11/07, Luís Santos <lsantos@itquality.com.br > wrote:

I have compiled every library and binary with my G++ 3.4.2 (MinGW Edition)..
The only thing I believe to be strange here is that it seems that luasql wasn´t prepared to compile under G++ (I needed to add some const char casts and an 'export "C" for luaopen_*')

Fabio Mascarenhas escreveu:

Did you compile Lua on your own, or are you using someone else's binaries? What about other modules? Your problem looks like mixup of different C runtimes.

--
Fabio Mascarenhas

--

Luís Eduardo Jason Santos
Coordenador Técnico
IT Quality Systems
lsantos@itquality.com.br
[21]2242-7959 ramal 49