Weekly Status: Difference between revisions

From Yocto Project
Jump to navigationJump to search
Line 1: Line 1:
== Yocto Project Weekly Status May 21st, 2024 ==
== Yocto Project Weekly Status May 28th, 2024 ==
Current Dev Position: YP 5.1 M1 <br/>
Current Dev Position: YP 5.1 M1 <br/>
Next Deadline: YP 5.1 M1 Build 20 May 2024<br/>
Next Deadline: YP 5.1 M1 Build 20 May 2024<br/>


'''Next Team Meetings:'''<br/>
'''Next Team Meetings:'''<br/>
*Bug Triage meeting Thursday May 23rd at 7:30 am PDT (https://zoom.us/j/454367603?pwd=ZGxoa2ZXL3FkM3Y0bFd5aVpHVVZ6dz09)
*Bug Triage meeting Thursday May 30th at 7:30 am PDT (https://zoom.us/j/454367603?pwd=ZGxoa2ZXL3FkM3Y0bFd5aVpHVVZ6dz09)
*Weekly Engineering Sync Tuesday May 21st at 8 am PDT (https://zoom.us/j/990892712?pwd=cHU1MjhoM2x6ck81bkcrYjRrcmJsUT09)
*Weekly Engineering Sync Tuesday May 28th at 8 am PDT (https://zoom.us/j/990892712?pwd=cHU1MjhoM2x6ck81bkcrYjRrcmJsUT09)
*Twitch -  See https://www.twitch.tv/theyoctojester
*Twitch -  See https://www.twitch.tv/theyoctojester


'''Key Status/Updates:'''<br/>
'''Key Status/Updates:'''<br/>
*YP 5.0.1 has been through QA and is pending review for release
*YP 5.0.1 has been released.
*We have suspected a performance problem for a while, this has now been identified/confirmed as an issue with hash equivalence. The public server using websockets/SSL has significantly slowed down things, to the point a 2 hour world build is taking 18+ hours. In trying to improve this, we ended up destablising the autobuilders entirely, breaking many builds and halting the patch merging.
The problem is visible both at build initialization time and also at any point it rehashes due to new hash information from the server. We are still trying to work out ways to improve things but there appears to be multiple levels of different problems, including the hashserve client parallelism seeming to have bugs. The issue affects scarthgap as well as master.
*A new gcc 14 compatible version of uninative was released allowing fedora 40 testing. Unfortunately there are issues with build failures on that host that are as yet uninvestigated.
*The WORKDIR to UNPACKDIR transition has merged and other layers are now working on the updates needed. There appear to be some challenges around go modules in particular which are needing more investigation and development time.
*We aim to build YP 5.1 M1 after the performance issue is addressed and we have stable builds.
*We were able to merge a chunk of pending patches but things are still processing slowly unfortunately.
*There are some CVE related announcements which are helpful to us:
*There are some CVE related announcements which are helpful to us:
**There was a CVE 5.1 Record format announcement with changes that encourage more useful data from our perspective
**There was a CVE 5.1 Record format announcement with changes that encourage more useful data from our perspective
**CISA announced on linkedin the creation of https://github.com/cisagov/vulnrichment
**CISA announced on linkedin the creation of https://github.com/cisagov/vulnrichment
**Unfortunately there are reports that the backlog of CVEs without version constraints is growing rapidly
*There is an open letter the project has created related to the CVE/NVD situation, more information is available here: https://lists.openembedded.org/g/openembedded-architecture/message/1990
*There is an open letter the project has created related to the CVE/NVD situation, more information is available here: https://lists.openembedded.org/g/openembedded-architecture/message/1990
**Please consider signing this to show support for those changes.
**Please consider signing this to show support for those changes.
**Signatures from small businesses/consultancies are as welcome as those from larger organizations as we want to demonstrate the breadth of the need for these changes.
**Signatures from small businesses/consultancies are as welcome as those from larger organizations as we want to demonstrate the breadth of the need for these changes.
*The project is sponsoring Syslinbit to separate out our CVE tooling from the build system into a standalone tool so that it can be used on software manifests of output at a later date. Marta sent out an update last week on this (thanks!)
*The project is sponsoring Syslinbit to separate out our CVE tooling from the build system into a standalone tool so that it can be used on software manifests of output at a later date.
*gcc 14 has merged to master. There are errors in meta-arm, meta-virtualization and meta-mingw. Now that has merged, we will build and release a new uninative version.
T*he project is in the process of testing a new way of hosting our autobuilder with plans to switch soon if successful. Our existing hardware is past end of life and showing increased hardware failures so is currently down on capacity meaning builds are taking longer than normal, also not helped with the performance issue identfiied above.
*More of the WORKDIR to UNPACKDIR transition has merged and devtool oe-local-files support has been dropped.
*Further performance graph improvements have merged and the graphs now have links to specific revisions allowing regressions to be much more easily tracked down.
*The next stage of the WORKDIR changes is pending testing in master-next.
*The aim is to build YP 5.1 M1 as soon as the various queued patches stabilize.
*The project is in the process of testing a new way of hosting our autobuilder with plans to switch soon if successful. Our existing hardware is past end of life and showing increased hardware failures so is currently down on capacity meaning builds are taking longer than normal.
*Automated testing continues to show large numbers of problems (e.g. a reproducibility issue in spirv-tools) but we continue to work through and fix issues as quickly as we can.
*Two issues causing failures on the new test infrastructure were tracked down to problems in bitbake. Both were effectively timing related issues.
PR server pass through support merged.
*A rust build issue of x86 hosted builds vs arm hosted builds has been identified as the cause of intermittent failures. Whilst this can be fixed, the fix breaks tests and/or reproducibility which gives challenges given the need to fix scarthgap. There are also challenges upgrading to newer rust versions.
*The performance graph improvements have merged.
*Improvements to the patch and CVE metrics page have also merged.


'''Ways to contribute:'''<br/>
'''Ways to contribute:'''<br/>
Line 39: Line 37:


'''Tracking Metrics:'''<br/>
'''Tracking Metrics:'''<br/>
*WDD 2752 (last week 2740) (https://wiki.yoctoproject.org/charts/combo.html)
*WDD 2772 (last week 2752) (https://wiki.yoctoproject.org/charts/combo.html)
*OE-Core/Poky Patch Metrics
*OE-Core/Poky Patch Metrics
**Total patches found: 1127 (last week 1128)
**Total patches found: 1121 (last week 1127)
**Patches in the Pending State: 221 (20%) [last week 221 (20%)]
**Patches in the Pending State: 221 (20%) [last week 221 (20%)]
*https://autobuilder.yocto.io/pub/non-release/patchmetrics/
*https://autobuilder.yocto.io/pub/non-release/patchmetrics/
Line 56: Line 54:


'''Upcoming dot releases:'''<br/>
'''Upcoming dot releases:'''<br/>
*YP 5.0.1 is ready for release.
*YP 5.0.1 is released.
*YP 4.0.19 build date 2024-06-03
*YP 4.0.19 build date 2024-06-03
*YP 4.0.19 Release date 2024-06-14
*YP 4.0.19 Release date 2024-06-14

Revision as of 14:39, 28 May 2024

Yocto Project Weekly Status May 28th, 2024

Current Dev Position: YP 5.1 M1
Next Deadline: YP 5.1 M1 Build 20 May 2024

Next Team Meetings:

Key Status/Updates:

  • YP 5.0.1 has been released.
  • We have suspected a performance problem for a while, this has now been identified/confirmed as an issue with hash equivalence. The public server using websockets/SSL has significantly slowed down things, to the point a 2 hour world build is taking 18+ hours. In trying to improve this, we ended up destablising the autobuilders entirely, breaking many builds and halting the patch merging.

The problem is visible both at build initialization time and also at any point it rehashes due to new hash information from the server. We are still trying to work out ways to improve things but there appears to be multiple levels of different problems, including the hashserve client parallelism seeming to have bugs. The issue affects scarthgap as well as master.

  • A new gcc 14 compatible version of uninative was released allowing fedora 40 testing. Unfortunately there are issues with build failures on that host that are as yet uninvestigated.
  • The WORKDIR to UNPACKDIR transition has merged and other layers are now working on the updates needed. There appear to be some challenges around go modules in particular which are needing more investigation and development time.
  • We aim to build YP 5.1 M1 after the performance issue is addressed and we have stable builds.
  • We were able to merge a chunk of pending patches but things are still processing slowly unfortunately.
  • There are some CVE related announcements which are helpful to us:
    • There was a CVE 5.1 Record format announcement with changes that encourage more useful data from our perspective
    • CISA announced on linkedin the creation of https://github.com/cisagov/vulnrichment
    • Unfortunately there are reports that the backlog of CVEs without version constraints is growing rapidly
  • There is an open letter the project has created related to the CVE/NVD situation, more information is available here: https://lists.openembedded.org/g/openembedded-architecture/message/1990
    • Please consider signing this to show support for those changes.
    • Signatures from small businesses/consultancies are as welcome as those from larger organizations as we want to demonstrate the breadth of the need for these changes.
  • The project is sponsoring Syslinbit to separate out our CVE tooling from the build system into a standalone tool so that it can be used on software manifests of output at a later date.

T*he project is in the process of testing a new way of hosting our autobuilder with plans to switch soon if successful. Our existing hardware is past end of life and showing increased hardware failures so is currently down on capacity meaning builds are taking longer than normal, also not helped with the performance issue identfiied above.

  • Further performance graph improvements have merged and the graphs now have links to specific revisions allowing regressions to be much more easily tracked down.

Ways to contribute:

Tracking Metrics:

YP 5.1 Milestone Dates:

  • YP 5.1 M1 Build date 2024-05-20
  • YP 5.1 M1 Release date 2024-05-31
  • YP 5.1 M2 Build date 2024-07-08
  • YP 5.1 M2 Release date 2024-07-19
  • YP 5.1 M3 Build date 2024-08-26
  • YP 5.1 M3 Release date 2024-09-06
  • YP 5.1 M4 Build date 2024-09-30
  • YP 5.1 M4 Release date 2024-10-25

Upcoming dot releases:

  • YP 5.0.1 is released.
  • YP 4.0.19 build date 2024-06-03
  • YP 4.0.19 Release date 2024-06-14
  • YP 5.0.2 Build Date 2024-06-24
  • YP 5.0.2 Release Date 2024-07-05
  • YP 4.0.20 Build Date 2024-07-15
  • YP 4.0.20 Release Date 2024-07-26
  • YP 5.0.3 Build Date 2024-08-12
  • YP 5.0.3 Release Date 2024-08-23
  • YP 4.0.21 Build Date 2024-09-09
  • YP 4.0.21 Release Date 2024-09-20
  • YP 5.0.4 Build Date 2024-09-23
  • YP 5.0.4 Release Date 2024-10-04
  • YP 4.0.22 Build Date 2024-10-14
  • YP 4.0.22 Release Date 2024-10-25
  • YP 5.0.5 Build Date 2024-11-11
  • YP 5.0.5 Release Date 2024-11-22
  • YP 4.0.23 Build Date 2024-11-18
  • YP 4.0.23 Release Date 2024-11-29

The Yocto Project’s technical governance is through its Technical Steering Committee, more information is available at:
https://wiki.yoctoproject.org/wiki/TSC

Archives