On Wed, Nov 16, 2005 at 01:16:40PM +0100, Florian Weimer wrote: > * D. Richard Hipp: > > > If an error occurs in a step function, record that fact in > > the aggregate context. Then when the finalizer is called, > > check the error flag in the context and call sqlite3_result_error > > at that point if it is appropriate to do so. > > Does this really work? According to my experiments, the fact that > sqlite3_result_error has been called from the finalizer is ignored. > The sources seem to lack the appropriate check, but I'm not sure.
I haven't looked at the code that handles it, but in practice it seems that an error returned from the step function is noticed and aborts the query, but gives only the non-descript error message: "SQL logic error or missing database". An error returned from the final function is ignored, and the string of the error message is used as the result. --nate
