Gabriel Krisman Bertazi <krisman@xxxxxxxxxxxxxxx> writes: > In the current patchset, I'm rejecting invalid sequences (or at least I > thought so) as soon as the normalization fails. The open syscall will > trigger an -EINVAL once it identifies a bad sequence, for instance. Hmm, this is true for utf8n but not ASCII. My bad. It is an error in the patchset. -- Gabriel Krisman Bertazi