Notes20101129

Agenda

Next Meeting
- Jan 3rd
Releases
- 5.6.1 (starting today)
  - short 1-week turn around times
- 5.5.1 (aim for Jan)
  - will start .pre1 a week or two after 5.6.1.pre1
- 5.3.4 (post 5.5.1)
Administrative
- Policy reminder on changes from last meeting
- Patch applying pain during a release?
  - cross release patches should be applied at the same time/revision, ideally
  - release on one branch breaks this (or delays applying stuff)
  - Results:
    1. Live with it
    2. use more branches (create a new branch just for the release, off the existing -patches branch)
  - Live with it
- IETF Report
  - KSM is probably going to resurface; hopefully with code contributions
  - power management is doing a lot of work too
Technical
- Unit tests
  - Wes got trounced in the last month and didn't write the promised documentation
  - bvassche has been creating lots of tests
  - tabled until next time when Wes can get docs written
- autoconf vs the world
  - still no takers on doing the work
- stream lined/stripped-down
  - work under way to make a "read only" code set using an ifdefs (along with a script to actually remove it as well)
  - thinking about multi-pass compile for really-reduced code. Something like sample at bottom.
  - discussion results captured below
Round Table
- dts12: upcoming releases
- Wes: most of the above discussion items
- rstory: dtls testing

Marking of features. Standard configure/make would always produce 1 for these ifdefs

// will only compile if needed by something else NETSNMP_BEGIN_FEATURE(counter64) ... code ... NETSNMP_END_FEATURE(counter64)

// something else says it's needed NETSNMP_REQUIRE_FEATURE(counter64)

Adding an extra "make reduce" or something would turn them into real ifdefs

.

ifdefs over macros
- but simple ifdefs. #ifdef NETSNMP_IMPLEMENTS_FEATURE
by default called from configure
- but actually runs a sub-script for non-configure-possible platforms
new parallel to config_require
- start with something new
- maybe convert the MIB config_require later if possible
defaults to including all features
will be useful for reducing backwards-compatability code