On Tue, 2023-09-19 at 14:18 +0200, Pablo Neira Ayuso wrote:
> Hi Thomas,
>
> On Tue, Sep 19, 2023 at 01:28:03PM +0200, Thomas Haller wrote:
> > It's not obvious that two enum values are missing (or why).
> > Explicitly
> > set the values to NULL, so we can see this more easily.
>
> I think this is uncovering a bug with these selectors.
>
> When concatenations are used, IIRC the delinerize path needs this.
>
> TYPE_CT_EVENTBIT does not need this, because this is a statement to
> globally filter ctnetlink events events.
>
> But TYPE_CT_LABEL is likely not working fine with concatenations.
>
> Let me take a closer look.
Hi Pablo,
Thank you.
FYI, I have a patch with a unit test that performs some consistency
checks of the "datatypes" array. Only TYPE_CT_LABEL + TYPE_CT_EVENTBIT
are missing.
You don't need to write a test about that. The test is however on top
of "no recursive make" patches, which I will resent at a later time.
Thomas
>
> > Signed-off-by: Thomas Haller <thaller@xxxxxxxxxx>
> > ---
> > src/datatype.c | 4 +++-
> > 1 file changed, 3 insertions(+), 1 deletion(-)
> >
> > diff --git a/src/datatype.c b/src/datatype.c
> > index 70c84846f70e..bb0c3cf79150 100644
> > --- a/src/datatype.c
> > +++ b/src/datatype.c
> > @@ -65,6 +65,7 @@ static const struct datatype *datatypes[TYPE_MAX
> > + 1] = {
> > [TYPE_CT_DIR] = &ct_dir_type,
> > [TYPE_CT_STATUS] = &ct_status_type,
> > [TYPE_ICMP6_TYPE] = &icmp6_type_type,
> > + [TYPE_CT_LABEL] = NULL,
> > [TYPE_PKTTYPE] = &pkttype_type,
> > [TYPE_ICMP_CODE] = &icmp_code_type,
> > [TYPE_ICMPV6_CODE] = &icmpv6_code_type,
> > @@ -72,8 +73,9 @@ static const struct datatype *datatypes[TYPE_MAX
> > + 1] = {
> > [TYPE_DEVGROUP] = &devgroup_type,
> > [TYPE_DSCP] = &dscp_type,
> > [TYPE_ECN] = &ecn_type,
> > - [TYPE_FIB_ADDR] = &fib_addr_type,
> > + [TYPE_FIB_ADDR] = &fib_addr_type,
> > [TYPE_BOOLEAN] = &boolean_type,
> > + [TYPE_CT_EVENTBIT] = NULL,
> > [TYPE_IFNAME] = &ifname_type,
> > [TYPE_IGMP_TYPE] = &igmp_type_type,
> > [TYPE_TIME_DATE] = &date_type,
> > --
> > 2.41.0
> >
>
