On Tue, Jul 28, 2009 at 07:06:50PM +0100, Martin J. Evans wrote: > Just a quick question about DBIh_SET_ERR_SV. I've produced a patch for > DBD::Oracle and it failing to handle non-unicode strings but running our > test code against DBI 1.609 produces a segfault. It appears: > > DBIh_SET_ERR_SV(h, imp_xxh, errcode_sv, errstr_sv, &sv_undef , &sv_undef); > > works fine and > > DBIh_SET_ERR_SV(h, imp_xxh, errcode_sv, errstr_sv, Nullsv, Nullsv); > > does not, no doubt due to the following code in set_err_sv: > > if (SvREADONLY(state)) state = sv_mortalcopy(state); > > Is that right as they seem inconsistent?
Some bits of perl internals check for null pointers and others don't. Looks like set_err_sv() assumes it won't be given null pointers. So don't do that :) I'll tweak the docs in DBI::DBD. Tim.
