"Ross J. Reedstrom" <reedstrm@wallace.ece.rice.edu> forwards:
> This absolutely beyond my mind. This means that user, being
> inserted explicitly, is correctly truncated. If it is inserted
> by default, it is wider than column! Seems something very strange
> is going on here.
Yes, I thought that some cases of default insertion were still broken,
and I was right. The patch Bruce put in before only fixes the case
of a wrong-length string constant being given as the default for a
char(n) field. When the default is not a constant, there needs to be
a run-time length coercion to char(n), and there isn't.
I have some work to do in that part of the parser anyway; will take care
of it (but not in time for 6.5.1, I fear).
regards, tom lane