[yocto] Minutes: Yocto Project Weekly Meeting 7/30/2019

[Reyna, David]

Yocto Project Weekly Meeting
MINUTES: 7/30/2019

Attendees: Richard, Michael, JoshuaW, Ross, Tim, David, Trevor, Jan-Simon

Note: consolidated bridge for Tuesday meetings: https://zoom.us/j/990892712

Richard:
  * Weekly status update will go out later today
  * 2.8 (3.0) M2 sent to QA
  * 2.6.3. ready to build, then QA
  * The Technical Steering Committee (TSC) updated their Wiki mission page (https://wiki.yoctoproject.org/wiki/TSC), stressing that they are the committee of last resort.
  * Mentioned again the YP-2.8 will be renamed as YP-3.0. Michael noted that we need to change the hold release “2.99” to “3.99”.
  * With working with the optimization patches, may want to break/cleanup API, specifically around task naming (e.g. array, filename-XXX, filename:XXX)
  * Hash equivalency work is breaking the eSDK. Joshua offered to help, noted that they do not use the eSDK and therefore had not tested against it.
  * Linux Foundation has launched a New Community Bridge (https://www.linuxfoundation.org/press-release/2019/03/the-linux-foundation-launches-new-communitybridge-platform-to-help-sustain-open-source-communities/). Nico has forwarded this to the YP community to encourage mentors/mentees.
  * We are slightly behind schedule for M2, but wanted to get the runqueue patches in and part of the QA

Joshua: Reproducible builds, wants to know why simple class is different. Ross will ping Juro. Richard recalls the more complex class was to get sensible dates, and it has been working very well to date (with a few exceptions), and there may be a bug in the reproducibility build class. Richard is aiming for reproducibility build tests, and these builds are the builds to prove it [to us and to the general community].

Joshua: RSS (?) versus host tools. There are issues, for example perl-native man pages can get different dates in the text depending if it is pulled from host or cache. We should review host tools versus native tools. Richard: specifically host tools to build native and then native to build target, and there should never be docs built for native.

Richard: Goal for YP-3.0 reproducibility. Joshua: is there something on autobuilder yet?

Richard: Running more Autobuilder full builds as opposed to quick builds to help flush out issues with recent big changes (e.g. runqueue). Full builds 1-3 times a weeks, cover pTest/LTP/build stats for x64 and arm64.

Tim: working with data analytics team in Intel, analyzing QA logs. Poses question to group: “what questions do we want to answer from the data?”. This will help guide the effort to extract meaningful results. Richard: noted that build performance is in a different database than what they are looking at right now.

Tim: Ported Autobuilder to Intel. Will update the wiki with learned lessons (e.g. local layers). Discovered that putting the sstats-cache on an NFS server brought the system to a crawl, was considering SSHFS. Local builds fast, but cache not shared. Richard noted the issue is that git does a lock on repos during check, and is slow/blocking especially for repos with long history. There was a fix/alternate process where the build is local and then the results copied to shared cache. Will provide patch to Tim, this is simpler that SSHFS. Noted that the Autobuilder was designed against YP’s build cluster. Tim noted that all the work on shared-caches over the years really helps.

Trevor: happy but curious why YP finally adopted “gsock”? Richard: still has concerns but will give it a go,  community and general experience has moved forward significantly over time, plus the fact it is LF project. We do need people to pick this up (not RP).

Tim: for community bridge topic, notes that care should be taken in choosing mentees: some become long term contributors, some never execute. Asked if there was any funding for mentees? Richard: there is some matching funds from LF especially for the underserved, also could get some support from YP sponsors. Trevor gave kudos to Xlinx.

- David


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.yoctoproject.org/pipermail/yocto/attachments/20190730/4b485d6d/attachment.html>