Agenda

  1. dbGroup proposal of Dave’s
  2. Pull requests for API
  1. 1. Can the pull requests be merged?
  2. 2. Can getIntField etc. be removed now? (Independent of a)
  3. 3. Can getScalarArrayField etc. be removed now? (Dependent on a)
  1. Release Timetable, nominally

Feature freeze:        2015-07-30

pre1:        2015-08-13

rc1:        2015-08-27

Final release:        2015-09-10

Minutes

Present: AJ, MK, GW, RL, DH, MD

Chair: AJ

Minutes: GW

NEW TOPIC: dbGroup proposal of Dave’s

Problems are connected to “processing”. In particular what operations which are “atomic” in the V$ sense. Eg Can agent influence the order in which operations for processing are done? When you write to a group, what can be processing at that time? If you write to a PV of the group, what gets processed and in which order?

AJ: Also whether you use the PP processing flags in record type definition.

RL: Didn’t think of different ways to represent a group - per Dave.

DH: Thinks the choice of which alternative structure comes down to whether you want individual timestamps and alarms.

MD: [last week at BNL now, but after that, lots of time]

MD: Will probably start with “lossless”. [Probably] “Composite Full Structure” per DH’s document.

RL: Isn’t it that the client creates the structure that will be filled by the server?

RL: As a reminder, please do talk me about req and gotchas of dbGroup and pvaGateway.

[consensus]: dbGroup would be implemented by some variation of Composite full structure.

NEW TOPIC: Pull requests for API

AJ: Asks andrew what his time constraint for ITER codac core inclusion of pvAccess.

RL: Says there is not now significant time constraint w.r.t. codaccore. Next hard date is October.

MK: Thrashing getting multi-channel coded, suggest removing it from both C++ and Java for this release. There is still a way to do multi-channel in Java but not in C++.

AJ: EasyPvaJava still has multi-channel, so will still be able to  use it.

DH: How long might it take to implement?

MK: Quandary is presenting everything as union array, but clients probably want an array of doubles (say). Considering allow for NTMultiChannel, no extras, but this means more work for the client, and also API to access as an array of doubles. Have code for that in C++ but needs clean-up. Trying to get the best features of both in one, but it doesn’t work.

GW: What can be done in 2 weeks?

MK: get/put/process/monitor for a single channel can be done this week. Then in 2 more weeks, NTMultiChannel support.

MK: Q: What’s replacement for RELEASE_VERSIONS now we use git.

GW/MD: No change wrt RELEASE_VERSIONS, the building script has been ported to git.

RESOLUTION: The pull request for API changes [of Dave] can proceed.

DH: Will merge the outstanding pull requests.

DH: Can I do both the API change and the removal of deprecated methods with the same commit?

[no firm opinions]

Discussion on how long to delay release dates, will be at least 2 weeks possibly 3.

AJ: Dave, what else do you want to do still?

DH: Exception hierarchy, new normative types (NTEnum wrapper, NTScalarMultiChannel), NTMultiChannel has extra field isConnected – not documented, bug fixes.

AJ: Ok, please discuss details in email. Encourage work on the exception hierarchy, worth doing.

Meeting closed at xx:50.