I'm not familiar with ctl(4), but it seems ctl.conf(5) says it support IPv6.

So does ctl(4) not yet support IPv6 ?

This bzero seems to be redundant. All members of struct sockaddr_l2cap are set via the following sequence.

Looks more readable.

Ditto.

or if (PTRIN(addr) != NULL) {?

And actually fixes a serious bug. Thanks!

softc->ha_lso is created in ctl_ha_listen() always as PF_INET socket.

No need for type casting.

From the compiler perspective, the type of the variable is known.
In this case typeof *sa == struct sockaddr_l2cap so the assignment can be shorted.

I compiled with clang and gcc, both work.

So unless the type casting is intended ( with a good reason ), I would prefer to remove them.

No need for type casting.

From the compiler perspective, the type of the variable is known.
In this case typeof *sa == struct sockaddr_l2cap so the assignment can be shorted.

I compiled with clang and gcc, both work.

How is that possible? For me it fails immediately on system clang:

/usr/src/FreeBSD/sys/netgraph/bluetooth/socket/ng_btsocket_l2cap.c:2522:8: error: expected expression
        *sa = {
              ^

I think the problem is not with compiler being able or not being able to determine type of variable. The problem is with compiler parser, which isn't able to tell a code block from structure value.

No need for type casting.

From the compiler perspective, the type of the variable is known.
In this case typeof *sa == struct sockaddr_l2cap so the assignment can be shorted.

I compiled with clang and gcc, both work.

How is that possible? For me it fails immediately on system clang:

/usr/src/FreeBSD/sys/netgraph/bluetooth/socket/ng_btsocket_l2cap.c:2522:8: error: expected expression
        *sa = {
              ^

I think the problem is not with compiler being able or not being able to determine type of variable. The problem is with compiler parser, which isn't able to tell a code block from structure value.

Sorry I mixed C and C++ features.

I compiled with clang and gcc, both work.

I tested that feature on godbolt.org and forgot to change lang of source code to C, it is default to C++ :-)

Please ignore my previous comment No need for type casting, and go ahead :)

Minor style changes, to keep align with current practices.

# egrep -r '=\s*\(struct \S+\s*\)\s*{' sys
sys/gdb/netgdb.c:	aux = (struct debugnet_proto_aux) {
...

I personally like this space. IMHO, it makes it more human readable. If you really insist I can delete it of course, but it is there intentionally. Not a typo.

I personally like this space. IMHO, it makes it more human readable. If you really insist I can delete it of course, but it is there intentionally. Not a typo.

Yeh I thought it is typo.

*sa = (struct sockaddr_l2cap ){

I spot this as I do not feel comfortable with it ( I can not tell why ) at the first glance. Probably because I had bad times arguing over code styles with my teammates. ;) We finally concluded a set of checkstyle rules shared between multiple teams and projects and checking styles is done by CI ( via pre commit hook ), then no one waste time arguing about styles any more ;) Anyway enthusiasm and intelligence are precious and as a team we should mainly focus on more important things such as documents / features and bugs.

Well I do not have strong option. I'm OK if you think it makes it more human readable.