Difference between revisions of "Notes20101129"

From Net-SNMP Wiki
Jump to: navigation, search
(Agenda)
(Fix link for next meeting)
 
(5 intermediate revisions by one other user not shown)
Line 1: Line 1:
 
<!-- Reminder for Spring 2011 - will need to use "time=19" to allow for DST -->
 
<!-- Reminder for Spring 2011 - will need to use "time=19" to allow for DST -->
{{meetingtime|time=20|year=2010|month=11|day=29|nextyear=2010|nextmonth=12|nextday=??|prevyear=2010|prevmonth=10|prevday=11}}
+
{{meetingtime|time=20|year=2010|month=11|day=29|nextyear=2011|nextmonth=01|nextday=03|prevyear=2010|prevmonth=10|prevday=11}}
  
  
Line 31: Line 31:
 
*** '''tabled until next time when Wes can get docs written'''
 
*** '''tabled until next time when Wes can get docs written'''
 
** autoconf vs the world
 
** autoconf vs the world
 +
*** ''still no takers on doing the work''
 
** stream lined/stripped-down
 
** 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)
 
*** 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.
 
*** thinking about multi-pass compile for really-reduced code.  Something like sample at bottom.
 +
*** '''discussion results captured below'''
 
* Round Table
 
* Round Table
 +
** dts12: upcoming releases
 +
** Wes: most of the above discussion items
 +
** rstory: dtls testing
  
 
=== feature possibility thinking ===
 
=== feature possibility thinking ===
Line 51: Line 56:
  
 
.
 
.
 +
 +
==== resulting conclusions ====
 +
 +
# 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

Latest revision as of 21:57, 22 December 2010

Previous: 2010-10-11
Next: 2011-01-03

This is the meeting agenda/notes for the Meeting held on 2010-11-29.

Date and Time: 2010-11-29 20:00 UTC


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

feature possibility thinking

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

.

resulting conclusions

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