Notes20100419

From Net-SNMP Wiki
Revision as of 20:16, 19 April 2010 by Wes (Talk | contribs)

Jump to: navigation, search
Previous: 2010-03-15
Next: 2010-05-19

This is the meeting agenda/notes for the Meeting held on 2010-04-19.

Date and Time: 2010-04-19 20:00 UTC

Agenda

  • Next Meeting Date
    • May 19th
  • Releases
    • 5.2.x -- dts12
    • 5.4.x -- dts12
    • Text sent from dts12:
 With one exception, I believe the 5.4.3 and 5.2.6
 releases are ready to go out.   That exception is,
 of course, the status of the Windows builds.

 There are two issues here:
   a) Notification Log MIB
   b) IPv6 support

 As it stands at the moment, the code (for both lines)
 builds OK using both Visual Studio (run interactively)
 and using 'build.bat'.  However previous versions of the
 5.4.x agent & trapd included support for the Notification
 Log MIB when built with Visual Studio, and this is no
 longer active.

 I believe there are three options here:
   i)   Accept the current code
          (build.bat works, MSVC build loses functionality)
   ii)  Reactivate the NLog-MIB support
          (MSVC works as before, build.bat fails)
   iii) Add support to build.bat framework to include NLog
      • iii is the right option; bartvanassche will look into a fix to build.bat which should be easy as it works in trunk and 5.5
 As regards IPv6 support, the current status is (I believe)
 that MSVC builds OK, but (definitely) build.bat fails when
 IPv6 support is enabled.  The patch I proposed is apparently
 not suitable - Bart is looking to come up with a better version.

 There are two options here:
    I)   Accept the current code
           (and document that build.bat does not support IPv6)
    II)  Use whatever fix Bart comes up with.
      • bartvanassche posted a new patch to -coders that should fix everything; coders are encouraged to voice support (or not) for it
      • Release Status
    • 5.6 -- hardaker
      • hardaker will likely start the process after the others get out and/or in early May
    • Release Testing
      • Notes from Dave:
This recent flurry of activity has brought to light a number
of problems with the project - in particular knowing what has
or hasn't been tested.
  A few things for you to think about at the meeting:
 
 - Do we need an equivalent of "make test" for Windows?
   Currently, I'm just building the code, and checking that
   this succeeds.  A test suite would help detect run-time
   problems of the type Bart has been reporting.

 - Currently, we rely on people telling us about problems with
   the pre- and rc- builds.  If we hear nothing, then we assume
   things are OK.
      Should we be asking for explicit success reports as well?

 - I seem to remember discussion of "make test" offering to
   send a build report to the project.  I'm increasingly of
   the opinion that this is probably a good idea.
       But what about Windows?  (see above)
      • Wes will see if he can get more non-sh tests in place for the future
      • Supposedly make test does actually work on windows already
      • make test is currently broken on linux for agentx? Possibly due to new transport code?
    • Binary Releases
      • Notes from Dave:
 We recently had a query about whether the 5.4.2 Windows binaries
 included SSL support or not.   Comparing this release with previous
 versions showed one set of binaries (?without) rather than two
 (both with/without).   But I realised that I didn't know how to
 generate the missing set (or even whether I could)

 Again, this raises a number of issues: 

   - Should we continue to offer two sets of Windows binaries?

   - How do users know what functionality is/isn't supported
     in any given binary distribution?  Should these packages
     include a summary of the configuration options used?
       (Not just Windows - this applies to other O/S as well)

   - Who among the core team is able/responsible for creating
     binaries?  (Again, various systems, not just Windows).
       I suspect it might be useful to have proper Build
     Instructions for each of the supported architecture
        • Agreement that everything should be documented
        • Wes sent a request to Alex for windows instructions
  • Administrative
    • Test server updates
      • rstory: v3 gets are working again
      • will work on writable support once dtls is online as well
  • Technical Discussions
    • Perl NetSNMP::agent regression test 5 fails [1].
      • Wes will look into it
    • Net-SNMP V5 get/set request handler documentation [2].
      • General agreement we should take the older ones offline entirely
      • Volunteers needed to do restructuring of the handler (especially) documentation so it's all in one place again
    • Whether or not the triple build infrastructure should be unified to a single build infrastructure.
      • Issues surrounding multiple build systems, especially on windows.
      • Moving files, etc, breaks windows builds
      • Possible solutions would include moving away from configure and system to something like CMake which supports most platforms.
        • But CMake is C++ internally and may be a problem within certain embedded device build environments?
      • Maybe it's possible to build the windows files from the unix configure system?
      • People with ideas are encouraged to write up high-level descriptions of how a replacement might work
        • noting that an important aspect of Net-SNMP is it's third-party easy-to-use plugin mechansims
  • Round Table -- What are ya up to?