patch for postgresql driver

[Wols Lists]

On 16/04/13 19:25, Noel Grandin wrote:
> On Tue, Apr 16, 2013 at 6:36 PM, Wols Lists <antlists at youngman.org.uk> wrote:
>>
>> I need to force some type of type conversion - basically from "const
>> char *" to "const rtl::OUString&", I presume.
> 
> You can either call
>    matchIgnoreAsciiCase( OUString ( p ) )
> or
>    matchIgnoreAsciiCase( p, strlen(p) )
> 
This still refuses to work. I'm guessing the problem lies in my
declaration of the array...

   static const char* keyword_list[] = {
        "password",
        "user",
        "port",
        "dbname",
        "connect_timeout",
        "options",
        "requiressl"
    };

    for( int i = 0; i < args.getLength() ; ++i )
    {
        bool append = false;
//        for( int j = 0; j < (int) ( sizeof( keyword_list ) / sizeof(
char * )); j++)
        for( int j = 0; j < (int) SAL_N_ELEMENTS( keyword_list ); j++)
        {
            if( args[i].Name.matchIgnoreAsciiCase( OUString(
keyword_list[j]), strlen( keyword_list[j] )))
            {

/home/anthony/gitstuff/loffice/connectivity/source/drivers/postgresql/pq_connection.cxx:
In function ‘void pq_sdbc_driver::properties2arrays(const
com::sun::star::uno::Sequence<com::sun::star::beans::PropertyValue>&,
const
com::sun::star::uno::Reference<com::sun::star::script::XTypeConverter>&,
rtl_TextEncoding, pq_sdbc_driver::cstr_vector&,
pq_sdbc_driver::cstr_vector&)’:
/home/anthony/gitstuff/loffice/connectivity/source/drivers/postgresql/pq_connection.cxx:502:77:
error: invalid conversion from ‘const char*’ to ‘sal_Unicode {aka short
unsigned int}’ [-fpermissive]
/home/anthony/gitstuff/loffice/instdir/unxlngx6.pro/sdk/include/rtl/ustring.hxx:150:14:
error:   initializing argument 1 of
‘rtl::OUString::OUString(sal_Unicode)’ [-fpermissive]
make[1]: ***
[/home/anthony/gitstuff/loffice/workdir/unxlngx6.pro/CxxObject/connectivity/source/drivers/postgresql/pq_connection.o]
Error 1
make[1]: *** Waiting for unfinished jobs....
make: *** [build] Error 2

Bear in mind arrays of strings is funny, I'm guessing the test to make
sure the conversion is allowed is exploding.

If we look at the definition of RTL_CONSTASCII_STRINGPARAM

#define RTL_CONSTASCII_STRINGPARAM( constAsciiStr ) (&(constAsciiStr)[0]), \
    ((sal_Int32)SAL_N_ELEMENTS(constAsciiStr)-1)

it's doing something weird with the first argument as a validity check.
I'm feeding it the second dimension of a string array. I wonder.

Is the following valid syntax, and would it fix the problem?

   static const char* keyword_list[] = {
        &("password"),
        &("user")
    }

Cheers,
Wol