Pending release of 2.1.9

[Alan DeKok]

John Dennis wrote:
> It passes basic sanity checking. It builds, installs, and runs. I have
> tested with radtest and with each of the eapol_test scripts. I do not
> have a stress testing environment, I think others do and it would be
> good to hear from them.

  OK.

> The Changelog notes several feature additions. I thought this was a bug
> fix update only. In fairness some of the feature additions were in the
> area of documentation, that's great and I don't have a problem with
> features which do not change code and make it easier for users to use.
> But shouldn't the other features have been reserved for the 2.2.x branch
> and limit 2.1.9 to only bug fixes?

  The features are:

- show stats for detail files
  Arguably a bug that it wasn't there originally.
  Added because people ran into problems where they couldn't see
  what was going on with a detail file
  The control socket isn't enabled in the default install, either.

- documentation

- better DHCP Option 82 support
  Arguably a bug: DHCP servers need Option 82 support.
  This affects only people who use DHCP. (i.e. not many)

- enabled "server" in NAS table
  arguably a bug that it wasn't there a year ago.
  Only affects *new* installations who use SQL.

  For me, all of these fall into the "arguably a bug fix" area.  There
are no major code changes, and will not affect existing systems.

> The one bug I was most concerned about I don't see specifically called
> out and I'm wondering what the disposition of that was. Sorry, but I'm
> going to be a little vague rather than citing a bug number. There was a
> problem reported by several people that resulted in a server crash and
> only seemed to appear under high load conditions after the server was up
> for a while. Alan said he was having a hard time reproducing it, that
> logically it seemed impossible from static code inspection, but
> acknowledged it was real because it had been reported often enough. Does
> that ring a bell? Does this update address that issue?

  Yes.  Bug #35.  There's a work-around which should help.

  I've run *billions* of packets through the server on the same machine
as people who claim to have problems.  I've been unable to reproduce the
issue.

  Alan DeKok.