OEDEM 2015: Difference between revisions

From Openembedded.org
Jump to navigation Jump to search
No edit summary
 
(28 intermediate revisions by 2 users not shown)
Line 23: Line 23:
== Agenda ==
== Agenda ==


* Metadata
* <s>Metadata</s>
** <s>OVERRIDE syntax: is it time to reconsider _?</s>
** <s>OVERRIDE syntax: is it time to reconsider _?</s>
** <s>meta-yocto -> meta-poky. People are still spending too much time explaining reality. (Crofton)</s>
** <s>meta-yocto -> meta-poky. People are still spending too much time explaining reality. (Crofton)</s>
** <s>variable expand = True by default</s>
** <s>variable expand = True by default</s>


* BSPs (tlwoerner)
* <s>BSPs (tlwoerner)</s>
** <s>Standards for BSP behavior. (Crofton)</s>
** <s>Standards for BSP behavior. (Crofton)</s>
** <s>feedback on layerindex to indicate BSP:
** <s>feedback on layerindex to indicate BSP:
Line 34: Line 34:
*** last successful boot
*** last successful boot
*** master branch parsability (perhaps ties into bugzilla 7792?)</s>
*** master branch parsability (perhaps ties into bugzilla 7792?)</s>
** several BSP layers to choose from for some boards, none for others (linux.com SBC survey)
** <s>several BSP layers to choose from for some boards, none for others (linux.com SBC survey)</s>
** <s>guidelines/requirements for setting up a new BSP on layerindex
** <s>guidelines/requirements for setting up a new BSP on layerindex
*** bugzilla
*** bugzilla
*** mailing list (use general mailing list(s) for patches?)
*** mailing list (use general mailing list(s) for patches?)
*** git.yoctoproject.org/<yourbsplayer></s>
*** git.yoctoproject.org/<yourbsplayer></s>
** making sure BSP layers only touch the variables/settings they should touch
** <s>making sure BSP layers only touch the variables/settings they should touch
*** should a BSP layer ever use = ?
*** should a BSP layer ever use = ?</s>
** does/should a BSP produce artifacts that are easy to install to SD cards? (Crofton)
** <s>does/should a BSP produce artifacts that are easy to install to SD cards? (Crofton)</s>


*Processes
*<s>Processes</s>
** <s>LayerIndex feature requests (tlwoerner)
** <s>LayerIndex feature requests (tlwoerner)
** build feedback (autobuilders and off-site)</s>
** build feedback (autobuilders and off-site)
** Build and test failures on the public Autobuilder (Bluelightening)
** Build and test failures on the public Autobuilder (Bluelightening)</s>


*Systemd
*<s>Systemd
** Systemd packaging. As new things appear, they are lumped into one package. Update split? (Crofton relaying info)
** Systemd packaging. As new things appear, they are lumped into one package. Update split? (Crofton relaying info)
** Can we get systemd's first-boot support to work? It is currently disabled in the systemd recipe by touching /etc/machine-id. (Saur)
** Can we get systemd's first-boot support to work? It is currently disabled in the systemd recipe by touching /etc/machine-id. (Saur)</s>
** How to make it easy to setup a system "the systemd way"? New DISTRO_FEATURE or IMAGE_FEATURE? (Saur)
** <s>How to make it easy to setup a system "the systemd way"? New DISTRO_FEATURE or IMAGE_FEATURE? (Saur)</s>
*** Link /bin -> /usr/bin, /lib -> /usr/lib, etc.
*** <s>Link /bin -> /usr/bin, /lib -> /usr/lib, etc.</s>
*** Change ${systemd_unitdir} to /usr/lib.
*** <s>Change ${systemd_unitdir} to /usr/lib.</s>
*** Make the systemctl wrapper install in ${systemd_unitdir} rather than ${systemconf}/systemd.
*** <s>Make the systemctl wrapper install in ${systemd_unitdir} rather than ${systemconf}/systemd.
**** This should also happen when installing a package in runtime.
**** This should also happen when installing a package in runtime.</s>


*Advocacy
*<s>Advocacy
** There are still a lot of projects (too many?) not using OE (tlwoerner)
** There are still a lot of projects (too many?) not using OE (tlwoerner)
** why aren't more projects using OE?
** why aren't more projects using OE?
** how can we get more people using OE?
** how can we get more people using OE?
** who should we be targeting? (build people (obviously), kernel devs? user-space devs?)
** who should we be targeting? (build people (obviously), kernel devs? user-space devs?)</s>


== Attendees ==
== Attendees ==
Line 96: Line 96:


  RAW notes shown in boxes like this
  RAW notes shown in boxes like this
=> this is an action item
-> this is a desired thing


Introduce TSC
Introduce TSC
Line 153: Line 155:
  RP maintainers must be willing to come to triage process
  RP maintainers must be willing to come to triage process
  some issues don't have a clear maintainer
  some issues don't have a clear maintainer
------------------------------------------
BSP standards - standards exist,coming soon are methods for comparing a given layer with the standard
"oe has no rules, only half-documented best practices" - Koen
=> need wiki documentation for TSC to ratify
-> document distro vs. machine policies, other best practices
------------------------------------------
BSP layer availability
-> encourage board vendors to provide a BSP, make it easier to do so
people problem, not technical problem
add questionnaire about why people don't
ask LF to include a list of BSPs that are supported
=> ask yp ab to talk to linux.com [PB] - community issue
build for qemu, with limited resources
-> look at digi manual (i.e. "just add bsp changes in local.conf")
best practices : should be easy to use, instructions/tools to put it on a card, etc
wic incomplete but promising
------------------------------------------
touching variables - discuss on new arch list
------------------------------------------


=== Metadata ===
=== Metadata ===
Line 170: Line 191:


=== Processes ===
=== Processes ===
Layer Index feature requests discussed with layer index
build feedback>>
Build/test failures on public autobuilder
problems:
RP
different failures "random" - actually is a pattern, but initially looks random
build poky and poky-lsb, lsb turns on different distro features
boots a machine in qemu & runs scripts, most qemu machines, just says "kill" - unknown
release blocker causing qa failures
also, systemd having timeout issues
should OE care about these things? (yes)
PE
set up environment to detect regressions
did upgrading qemu and systemd create these? get to a correct point
not poky changes or bsp changes
eg poky enables ptest, doesn't affect systemd timeout
RP
have massively increased scope of tests
ptests, sdk..
resources always an issue
-> ask for cycles from employers to work on these issues
-> make people aware, help with triage, provide resources
KK -> plan B, C, etc. have had this problem for years
more issues recently
=== Security ===
security a long term issue, beginning to be noticed - now reluctant to address it, as it is being noticed/resourced
if patch had standard cv, could write automated tools to parse
-> need a method for tracking CVs
MH have to have someone whose job it is to pull CV list, look for differences, assess applicability
Michael Halstead : LF can run security audits of individual tools
could do layer index, toaster, ... busybox
MH independent academic audit found OE very good
=== Systemd ===
=== Systemd ===
currently scaling up
complaints: more package config, comments; look at packaging to have minimal systemd even with options available
-> need someone with time available to work on it
-> need to get systemd feedback, people using systemd talk to package maintainer
MH automotive is a heavy user, could ask genivi for help
peter, koen using systemd but not using official oe recipe - based on
Can we get systemd first boot support to work?
MH wind struggling on read-only filesystems
Easy setup "should work now"
yes can get it to work - talk to upstream
changing prefixes problematic
MH unitdir to /usr/lib - can't boot without base lib
should be able to boot minimal system w/shell and some commands
should work if reasonable default can be changed
patches are problematic - RP wants clear coherent vision
systemctl wrapper needs documentation for working with systemd
current docs targeted at app developers - not distro devs
want to use the install section
=> file systemctl wrapper bug
=== Advocacy ===
=== Advocacy ===
b
why aren't more projects using OE?
Belen:
difficult to understand & get started with
- toaster
don't understand the benefits
- not a distro
slower systems complain about performance (also don't understand sstate)
- could enable sstate mirrors by default
- now measuring size of builds
- massively examine macros in autotools
how can we get more people using OE?
outreach
documentation
intro on oe, what it is & benefits
no official body within OE to handle this - community mgmt, advocacy
-> board to discuss advocacy/community, invite jefro to meetings
need someone to run oe booth at fosdem, as paul is leaving [beth, belen]
value in tools:
devtool
adt
who should we be targeting? build people, kernel devs, userspace devs
* college kids, interns
* grad school programs
* makers
* white papers - active promotion (how is this different from debian etc) - at executive level; case studies, potato counter, comcast
* maintain community metrics
* koen: conference presentation comparing various build systems (oe, buildroot, etc)
* devs doing board bringup
* is oe for everyone?
== Potential OE Day at ELCE next year ==
PB: day of oe presentations
more like bofs
=> board: do a proper cfp
tlwoerner: lightning talks
[[Category:OEDEM]]

Latest revision as of 17:38, 12 August 2020

The 2015 OpenEmbedded Developer's European Meeting will take place Friday, Oct 9, 2015 in Dublin, Ireland. Part of this meeting will also constitute the annual OE General Assembly, and some corporation/eV matters will be discussed.

Location and Time

The Gibson Hotel
Point Village
Dublin 1, Ireland
+353 1 681 5000

9am - 5pm (note: times may change)

Gibson Hotel is at the end of the red LUAS line, adjacent to the stadium. Come up to the 3rd floor and follow the signs for the Cordoba room. (the sign says "Mtg", that's us) There is coffee here. See you all soon

OpenEmbedded eV General Assembly

We have a short administrative meeting before starting the Developer meeting. All are welcome for this.

  • Proposal to dissolve the current e.V. and re-constituting it in the U.S. to reduce the administrative burden for the project.
  • Define a method for identifying lapsed memberships.
  • Proxy form File:Proxy instructions-oe.pdf
  • Meeting Minutes

Agenda

  • Metadata
    • OVERRIDE syntax: is it time to reconsider _?
    • meta-yocto -> meta-poky. People are still spending too much time explaining reality. (Crofton)
    • variable expand = True by default
  • BSPs (tlwoerner)
    • Standards for BSP behavior. (Crofton)
    • feedback on layerindex to indicate BSP:
      • last successful compile (feedback from outside builders?)
      • last successful boot
      • master branch parsability (perhaps ties into bugzilla 7792?)
    • several BSP layers to choose from for some boards, none for others (linux.com SBC survey)
    • guidelines/requirements for setting up a new BSP on layerindex
      • bugzilla
      • mailing list (use general mailing list(s) for patches?)
      • git.yoctoproject.org/<yourbsplayer>
    • making sure BSP layers only touch the variables/settings they should touch
      • should a BSP layer ever use = ?
    • does/should a BSP produce artifacts that are easy to install to SD cards? (Crofton)
  • Processes
    • LayerIndex feature requests (tlwoerner)
    • build feedback (autobuilders and off-site)
    • Build and test failures on the public Autobuilder (Bluelightening)
  • Systemd
    • Systemd packaging. As new things appear, they are lumped into one package. Update split? (Crofton relaying info)
    • Can we get systemd's first-boot support to work? It is currently disabled in the systemd recipe by touching /etc/machine-id. (Saur)
    • How to make it easy to setup a system "the systemd way"? New DISTRO_FEATURE or IMAGE_FEATURE? (Saur)
      • Link /bin -> /usr/bin, /lib -> /usr/lib, etc.
      • Change ${systemd_unitdir} to /usr/lib.
      • Make the systemctl wrapper install in ${systemd_unitdir} rather than ${systemconf}/systemd.
        • This should also happen when installing a package in runtime.
  • Advocacy
    • There are still a lot of projects (too many?) not using OE (tlwoerner)
    • why aren't more projects using OE?
    • how can we get more people using OE?
    • who should we be targeting? (build people (obviously), kernel devs? user-space devs?)

Attendees

If you plan to attend, please list your name below.

  1. Jeff Osier-Mixon (jefro)
  2. Philip Balister (Crofton)
  3. Trevor Woerner (tlwoerner)
  4. Anders Darander (andersd)
  5. Marco Cavallini (mckoan)
  6. Koen Kooi (koen)
  7. Alexandre Belloni (abelloni)
  8. Armin Kuster (armpit)
  9. Belén Barros Pena (belen)
  10. Peter Kjellerstedt (Saur)
  11. Alejandro del Castillo (adelcast)
  12. Denys Dmytriyenko (denix)
  13. Sean Hudson (darknighte)
  14. Nathan Rossi (nrossi)
  15. Mark Hatle (fray)
  16. Nicolas Dechesne (ndec)
  17. Paul Eggleton (bluelightning)
  18. Andrei Gherzan (agherzan)
  19. Felipe F. Tonello (ftonello)
  20. Changhyeok Bae (chbae)
  21. Saul Wold (sgw)
  22. Ashish Shrivastava
  23. Pascal Bach (bachp)
  24. Richard Purdie (RP)
  25. Michael Halstead (halstead)

Proceedings

RAW notes shown in boxes like this
=> this is an action item
-> this is a desired thing

Introduce TSC

  • Martin Jansa (jama)
  • Mark Hatle (fray)
  • Paul Eggleton (bluelightning)
  • Richard Purdie (RP)
  • Khem Raj (khem)

go over agenda BSP topic first to accommodate member travel

BSPs & Layer Index

Paul described layer index process & how it works
BBP layers also come into toaster, expose to people, dependency list very important - snapshot when setting up toaster
module branches - will look in repo for branches
some layers don't have a master branch, don't show content - need way to present in UI
REST API
DD automatically parse layer to check for required content? PE do want, necessitates background queuing process
parse w/dependencies
also affects YP Compatible, RP's proposed layer validation tool
error reporting system - tie together with info in layer index, intention there
JH ability to register layer & specify quality
------------------------------------------
not getting resources to work on core BSPs
people working on own BSPs - not really a complaint, but an observation
SH would love to see REST APIs
PE would love some help with the layer index
RP love ideas, but people need to resource, step up & drive
------------------------------------------
=> oe driving toward machine readable files, can also drive from layer test (yp compat) side - KK
------------------------------------------
oe-core or oe-devel?  oe-core
------------------------------------------
=> publicize defs of each list? discuss on members list - where to overall architecture discussions belong
discussion: decided on oe-core
oe needs its own goals, interact with the yocto project
------------------------------------------
status email to oe-core list, useful? to most people.  cross-postd to yocto list
KK should be about oe-core, not about poky
=> TSC to add new mailing list specifically about project architecture, no patches, and determine list name
------------------------------------------
KK realize that OE has no project vision
RP is there any vision that needs to be represented to YP, and are there resources
KK as a project take an active role to work with YP
marketing/visibility standpoint, wear OE hat
if list of oe bugs that should be raised, bring to philip
-> KK bring discussion to new mailing list
------------------------------------------
TW can new layers use bugzilla
PE struggle with volume of bugs currently in triage process
RP triage process - yp has taken on oe-core eg. don't want many bugs from idle layers, YP judged on metrics from that system
SW some layers already in that state
DD attended call regularly
RP maintainers must be willing to come to triage process
some issues don't have a clear maintainer
------------------------------------------
BSP standards - standards exist,coming soon are methods for comparing a given layer with the standard
"oe has no rules, only half-documented best practices" - Koen
=> need wiki documentation for TSC to ratify
-> document distro vs. machine policies, other best practices
------------------------------------------
BSP layer availability
-> encourage board vendors to provide a BSP, make it easier to do so
people problem, not technical problem
add questionnaire about why people don't
ask LF to include a list of BSPs that are supported
=> ask yp ab to talk to linux.com [PB] - community issue
build for qemu, with limited resources
-> look at digi manual (i.e. "just add bsp changes in local.conf")
best practices : should be easy to use, instructions/tools to put it on a card, etc
wic incomplete but promising
------------------------------------------
touching variables - discuss on new arch list
------------------------------------------

Metadata

meta-yocto to meta-poky:

=> RP to take action on the YP side, possibly not for 2.0

override syntax

underscore character to designate overrides, underscore is overloaded
need better method
KK maybe _OVERRIDE_
dot character?
require all lowercase overrides? benefits for bitbake speed
=> formalize lowercase as mandatory requirement for overrides, look for other options in the future

Processes

Layer Index feature requests discussed with layer index build feedback>>

Build/test failures on public autobuilder

problems:
RP
different failures "random" - actually is a pattern, but initially looks random
build poky and poky-lsb, lsb turns on different distro features
boots a machine in qemu & runs scripts, most qemu machines, just says "kill" - unknown
release blocker causing qa failures
also, systemd having timeout issues
should OE care about these things? (yes)
PE
set up environment to detect regressions
did upgrading qemu and systemd create these? get to a correct point
not poky changes or bsp changes
eg poky enables ptest, doesn't affect systemd timeout
RP
have massively increased scope of tests
ptests, sdk..
resources always an issue
-> ask for cycles from employers to work on these issues
-> make people aware, help with triage, provide resources
KK -> plan B, C, etc. have had this problem for years
more issues recently

Security

security a long term issue, beginning to be noticed - now reluctant to address it, as it is being noticed/resourced
if patch had standard cv, could write automated tools to parse
-> need a method for tracking CVs
MH have to have someone whose job it is to pull CV list, look for differences, assess applicability
Michael Halstead : LF can run security audits of individual tools
could do layer index, toaster, ... busybox
MH independent academic audit found OE very good

Systemd

currently scaling up
complaints: more package config, comments; look at packaging to have minimal systemd even with options available
-> need someone with time available to work on it
-> need to get systemd feedback, people using systemd talk to package maintainer
MH automotive is a heavy user, could ask genivi for help
peter, koen using systemd but not using official oe recipe - based on
Can we get systemd first boot support to work?
MH wind struggling on read-only filesystems
Easy setup "should work now"
yes can get it to work - talk to upstream
changing prefixes problematic
MH unitdir to /usr/lib - can't boot without base lib
should be able to boot minimal system w/shell and some commands
should work if reasonable default can be changed
patches are problematic - RP wants clear coherent vision
systemctl wrapper needs documentation for working with systemd
current docs targeted at app developers - not distro devs
want to use the install section
=> file systemctl wrapper bug

Advocacy

b why aren't more projects using OE?

Belen:
difficult to understand & get started with
- toaster
don't understand the benefits
- not a distro
slower systems complain about performance (also don't understand sstate)
- could enable sstate mirrors by default
- now measuring size of builds
- massively examine macros in autotools

how can we get more people using OE?

outreach
documentation
intro on oe, what it is & benefits
no official body within OE to handle this - community mgmt, advocacy
-> board to discuss advocacy/community, invite jefro to meetings
need someone to run oe booth at fosdem, as paul is leaving [beth, belen]
value in tools:
devtool
adt

who should we be targeting? build people, kernel devs, userspace devs

  • college kids, interns
  • grad school programs
  • makers
  • white papers - active promotion (how is this different from debian etc) - at executive level; case studies, potato counter, comcast
  • maintain community metrics
  • koen: conference presentation comparing various build systems (oe, buildroot, etc)
  • devs doing board bringup
  • is oe for everyone?

Potential OE Day at ELCE next year

PB: day of oe presentations
more like bofs
=> board: do a proper cfp
tlwoerner: lightning talks