[Swan-dev] Fwd: [Swan-commit] Changes to ref refs/heads/master

[Paul Wouters]

On Wed, 21 Nov 2018, Andrew Cagney wrote:

> - move the function to demux.[hc]
> The function, while currently IKEv2 specific, belongs in demux.[hc] as
> it operates on the msg_digest; and like the commit points out, several
> strange #includes were eliminated

Don't all functions that pull the list of payloads work on md? And those
are all in ikev[12]* ?


> - rename v2_msg_role() to msg_role()
> This one I umed and arhd over a bit.  It makes code like:
>    switch (msg_role(md)) {
>    case MESSAGE_REQUEST: ...
> slightly less verbose.

but now it is completely unclear this is an IKEv2 only function, and
someone accidentally might use it in ikev1 code. So for files in
programs/pluto that have no ikev* prefix, if the functions in those
files (demux.c, hostpair.c initiate.c etc) are specific to one version
of ike, the function name MUST reflect that.

> However, to your point, it seems that only
> IKEv2 can easily figure out if a message is a request or a response
> (which really scares me).  Given this, should
> MESSAGE_{REQUEST,RESPONSE} also get the V2_* treatment vis:
>    switch(v2_msg_role(md)) {
>    case V2_MESSAGE_REQUEST: ...
> but that is really looking cumbersome.

I think that is better, but I'm happy to also hear from Hugh,

Paul