EPICS V4 Telecon, 18-Nov-2014

AGENDA: Status of 4.4 release


- pvaSrv put behavior [1]
- what’s up with exampleJava?
- rc1 readiness and schedule (new tags will be needed, and therefore
dependencies re tagged also).

MINUTES

Present: AJ, DH, GC, MK, MS, RL, GW

Chair: AJ

Scribe: GW

NEW TOPIC: pvaSrv put behavior

MK: if you have process true, and block false, then it just calls …. and it’s done.

Before, he was going through notify code. now he directly call dbProcess.

dbProcess does not hang. If it had already started process, then it continued.

[not sure I got all of this]

MK: so we need DH to check proper functionality now

DH: <doesn’t understand hg tags [for pvaSrv]>

<meeting can’t see where ralph made relevant changes>

MK <brings rl up to date>

MK: I found it on 0.1.0

RL: The Thursday change is the code MK was proposing,but does not solve your problem

DH: the thing that doesn’t work is if the put doesn’t work, nothing cleans up

RL: the notify doesn’t get called, so the [channel] doesn’t get cleaned up

MS: pvaccess server code creates a “channel” that takes a ref to a channel. The pvserver (pvasrv) should create the channel. There may be a circular dependency - we have to check

RL: I see the channel being destroyed.

When putDone is called, the channel is destroyed

RL: regular operation:

[RL describes regular operation, and the circumstances in which putDestry is not called]

MS: pvaSrv does call destroy on a channel only. It is the channel’s responsibility to clean up the channel.

RL: ok, I know what I have to do.

MK: But I’m not sure if that’s true [couldn’t catch whey]

MS: for that case [of MK] use shared pointers

MK: you have a list of channels already at [?] level, now you are suggesting a channel must also keep a list. Is that sensible

RL; Matej and I can sort this out

MK: and include me

AJ: ok we will fix this in email

aj: can we fix it prior to release?

aj: i think it’s reasonable for this to go in

RL: but it’s a very special case - it’s a use case which uses the busyRecord beyond its documentation

Resolution: We await fixing of the pvaSrv fix, it will be included when it’s ready but won’t compromise the final release (include in -rc1 if by end of this week, in -rc2 if next week). Will come out with a 4.4.1 release soon afterwards if it takes longer.

NEW TOPIC: What’s up with exampleJava?

GW: Nothing now [it’s fixed]. 2 things held up: exampleJava with the flow model, GW gets error message that Ralph doesn’t. 2 issues: directoryService version number; caj & jca not in our maven area so not included in package from remotes.

GC: They are in maven central. The mvn repo on SF should be phased out, use maven central.

GW: Nobody made that decision.

GC: JCA & CAJ have not been updated on SF in 18 months.

MS: Talking about moving all our modules to maven central.

GC: Can you change the script to get them from maven central?

GW: Just change the URL? Compatible?

GC: Final releases are at

Search: http://search.maven.org/#search|ga|1|caj

Released version: https://repo1.maven.org/maven2/org/epics/caj/1.1.14/

Snapshot: https://oss.sonatype.org/content/repositories/snapshots/org/epics/caj/

AI on GW: Modify the Makefile script to take jca & caj from maven central

AI on MS & GC: Move other core modules to Maven Central

GC: Maven can automake creating tar files with build info, I can help with that.

NEW TOPIC: readiness and schedule

AJ: Nominally then we decided on an rc1 next week. And rc2 a week later.

AJ: Is that ok with DH?

DH confirms

AJ: so we will be preparing rc1 on november the 24th, rc2 on Dec 1st.

AJ: If all the changes we have been waiting for [to pvaSrv in particular] are in rc1, then we don’t necessarily need an rc2.

--- Meeting ends 8:48 PST ---