[kea-dev] Building KEA on OpenBSD

Patrik Lundin patrik at sigterm.se
Tue Dec 8 15:27:03 UTC 2015


On Mon, Dec 07, 2015 at 02:00:31PM +0100, Marcin Siodelski wrote:
> > 
> > After doing this the tests succeed for me. Are there any platforms you
> > are targeting that dont have pgrep available by default?
> > 
> 
> I have checked some boxes running Linux, BSD and OS-X, and they do seem
> to support 'pgrep'. So, I personally don't see issue with switching to
> 'pgrep'. I will file a ticket and we'll discuss it on our team meeting
> next Wednesday.
> 

Sounds good to me :).

> 
> Current use cases for random() seem to be:
> - Client Identifier generation for DHCPv6
> - Randomizing packet selection from the storage maintained by perfdhcp
> - Unit tests
> 
> I guess, none of those use cases require non-deterministic
> randomization. However, please note that we set a current-time based
> seed before randomization, so this provides sufficient mechanism for the
> current use cases.
> 
> I also don't think that switching to some other randomization mechanism
> would pose any issue for those use cases. But, it would add some
> complexity to the configure code.
> 
> We do have a ticket: http://kea.isc.org/ticket/3678, which also refers
> to http://kea.isc.org/ticket/3474 for an alternative solution. They both
> suggest to not use random(), so at some point we should take care of that.
> 
> However, we have treated this issue as a 'best practice' type of thing
> and thus it was not on our priority list. If this build warning causes
> some trouble for you integrating Kea as an OpenBSD package, please let
> me know and we'll discuss elevating priority for this work.
> 

Understood, thanks for taking a look and verifying that there are no
obvious problems here. I will leave these as-is unless I run into
trouble.

> 
> Current code (which will be released as Kea-1.0.0-beta this week)
> doesn't use strcat or sprintf. IIRC, in 0.9.2 we were using it
> indirectly with the ASIO implementation (in ext/) which we dragged with
> Kea. In 1.0.0 we rely on the boost::asio implementation, which should be
> installed on the system building Kea. The ASIO implementation has been
> removed from the Kea package. We don't have any plans to patch 'strcat'
> and 'sprintf' use in 0.9.2. The issue should go away with 1.0.
> 

Thanks again for a thorough response. I only mentioned these warning to
be complete. I'll keep an eye out when testing a newer release.

For interested parties the porting thread can be found here:
http://marc.info/?l=openbsd-ports&m=144943324603479&w=2

I do agree with a comment in a response I have recieved in that thread
that "message" is a pretty general name for a binary. Any thoughts on if
this should be renamed to "kea-message" or similar?

-- 
Patrik Lundin


More information about the kea-dev mailing list