[unixODBC-dev] ODBC on 64 bit linux - test64

Nick Gorham nick at lurcher.org
Thu Mar 13 09:45:09 GMT 2008


Kent Boortz wrote:

>Nick Gorham <nick at lurcher.org> writes:
>  
>
>>I have been putting it off for as long as possible because I know there 
>>will be a world of pain to follow.
>>    
>>
>
>Nick,
>
>Maybe suggested before, or a stupid idea, but couldn't you reduce the
>problem with driver and manager mismatch by requiring that "real 64
>mode" drivers define a symbol (maybe even controlled by unixODBC
>headers), lets say
>
>  unixodbc_real_64_bit_mode
>  
>
Yes its not a bad idea, but then you stil have the problem with drivers 
that don't follow the standard. Maybe its worth having two symbols, one 
for each, that way we can detect a mismatch, but know ehen we can't tell.

Maybe a

int unixodbc_sqlulen_size = sizeof( SQLULEN );

>and when loading the driver at runtime, check for this symbol. If not
>defined, refuse to load the driver with a nice error message?
>
>Well, maybe too much of a hack....
>
>I think a table like the one attached (but corrected where it is
>incorrect) would be useful on the unixODBC site, as there is much
>confusion about types in ODBC. I think, I'm not really a user of ODBC,
>I just build the ODBC stuff at MySQL AB ;-)
>  
>
Hmm, that would be a big help, but I suspect the real problem is going 
to be with applications. I know of at least one app that changed between 
minor versions :-(

>The table shows unixODBC SQLROWSETSIZE to differ from Microsoft 64
>bits, but that might be incorrect or corrected after 2.2.11,
>  
>
Maybe, I think (hope) its correct in 2.2.13

-- 
Nick



More information about the unixODBC-dev mailing list