On 07/21/2010 05:24 PM, Konstantin Osipov wrote:
> Perhaps a matter of taste, but assigning the error to TRUE
> at the beginning and resetting to FALSE only at the very end
> would save you a few lines.
A little bit like choosing between the plague and cholera ;-)
I've changed it the way you suggested.
> We should not release metadata locks on tables before calling
> close_thread_tables(). In my current patch I have an assert for
> it in close_thread_table().
> Thus here you also should call close_thread_tables() first.
--- Jon Olav