[Swan-dev] I think 229e2d24a4 needs to be reverted

[Paul Wouters]

On Wed, 8 May 2019, Antony Antony wrote:

> I am not commenting on 229e2d24a4, instead the line of argument you made.
>
> your statement sounds a bit scary to me.  If true it would mean we went back
> to stone ages!
>
> Why do you say - "stupid thing of IKEv2 using STATE_PARENT_I3 and
> STATE_PARENT_R2 for both IKE and CHILD SA" ?
>
> The states STATE_PARENT_I3 STATE_PARENT_R2 are not shared between IKE and
> Child SA state since the following commits.
>
> svm entry change e58685b2b8
> Added constant fa3ecaa60
>
> The terminal states of Child SA are STATE_V2_IPSEC_I STATE_V2_IPSEC_R.

Ahhh right. That is true. It is not as bad currently as I remembered it
was in the past. Still, changing such fundamental macros just before a
release is scary. Code might still depend on the parts removed from the
macro.

> NOTE as far as I recollect - there one place where parent and child share
> same state name - only on the initiator for STATE_PARENT_I2, it is only for
> short while. Then it become STATE_V2_IPSEC_I. I haven trying to find a quiet
> period, test results are stable for a week or so, to change this. See the
> wiki page for the propsed change. May be you are confused because of this.

But we agreed a few weeks ago to do that change post 3.28 as well :)

Paul