On Fri, 13 Dec 2019 13:40:38 +0100, Paul Chaignon wrote:
> > > @@ -176,7 +177,20 @@ prog_parse_fds(int *argc, char ***argv, int *fds)
> > > }
> > > NEXT_ARGP();
> > >
> > > - return prog_fd_by_tag(tag, fds);
> > > + return prog_fd_by_nametag(tag, fds, true);
> > > + } else if (is_prefix(**argv, "name")) {
> > > + char *name;
> > > +
> > > + NEXT_ARGP();
> > > +
> > > + name = **argv;
> > > + if (strlen(name) > BPF_OBJ_NAME_LEN - 1) {
> >
> > Is this needed? strncmp will simply never match, is it preferred to
> > hard error?
>
> I tried to follow the fail-early pattern of lookups by tag above.
Right although tag does a scanf and if we didn't scan all letters
we'd use uninit memory.
> I do like that there's a different error message for a longer than
> expected name. Since libbpf silently truncates names, typing a
> longer name is not uncommon.
Ugh, I didn't realize libbpf truncates names. Okay, let's keep the
error for now so we can switch to truncation if users complain.
