Skip to contentSkip to conversation

minutes for Yocto Project Technical Team Meeting / Engineering Sync - June 9 2020

DISCUSSION Trevor Woerner 2 participants 2 messages Jun 9, 2020 – Jun 10, 2020

lore
Share & tools
lore
#0 lore raw

Yocto Technical Team Minutes, Engineering Sync, for June 9 2020
archive: https://docs.google.com/document/d/1ly8nyhO14kDNnFcW2QskANXW3ZT7QwKC5wWVDg9dDH4/edit

== disclaimer ==
Best efforts are made to ensure the below is accurate and valid. However, errors sometimes happen. If any errors or omissions are found, please feel free to reply to this email with any corrections.

== attendees ==
Trevor Woerner, Jan-Simon Möller, Stephen Jolly, Trevor Gamblin, Bruce Ashfield, Scott Murray, Paul Barker, Michael Halstead, Jon Mason, Richard Purdie, Joshua Watt, Armin Kuster, Tim Orling, Sathishkumar Duraisamy, henriknj, Ross Burton, Jeremy Puhlman

== notes ==
- 2.7.4 released last week
- 3.1.1 built, but has issues, not sent to QA yet
- M1 early next week (to build)
- thanks to AlexK for lots of fixes

== general ==
RP:
problems with 3.1.1. qemu(-mips?) failed with taskhash mismatches, not sure why

RP:
devtool, patchtest, patchwork, and wic looking for maintainers
PaulB:
patchtest/patchwork still on python2, needs 2→3 update. patchtest should be moved to OE-core so it can be noticed by more people

RP:
lots of recipe upgrades, happy with where master is (wrt to recipe versions), still need to update qemu to 5

JPEW:
did RP see my patch to mc-depends to make it backwards compatible?
RP:
dismayed it was more complex than hoped, but looks okay
JPEW:
i’m using a proxy object otherwise we need to duplicate a lot of code with a lot of if’s
RP:
saw the BBMASK thing for multiconfig that needed index checking, sad that we’ll be stuck with that going forward (in command.py) mostly for backwards compatibility (e.g. file checking). another option is to have a new API and append a ‘2’ to the end so callers will call the right things (old or new). the code’s probably going to get worse and worse with every new parameter we want to add (in the future)

JPEW:
re multiconfig, sometimes the old full “multiconfig” string doesn’t work but the new “mc” string does, can we drop “multiconfig”? changed in 2.7? or 3.0? there’s a bugzilla somewhere. some code handles both, some doesn’t
RP:
we can get rid of it, despite the expected screaming

RP:
re devtool, reports on mailing list that devtool isn’t mc-compliant
RP:
we have a horrible hack for PRserver
RP:
highlights the need for a dedicated maintainer

PaulB:
re patchwork, kernel.org are using a newer version of patchwork that is python3-compatible, is maintained, but missing some things we need
Timo:
i looked at freedesktop.org’s gitlab, they’ve made a lot of changes from where we forked from them, they’re also using python3 and django2. going with mainline will miss features that we’re used to
PaulB:
make a list of the gap items, then decide if we can live without these things
Timo:
i looked at the fd.o’s patchwork and saw that many of the things we added to our fork have been added (but differently) in fd.o’s. e.g. we added a superseded feature, doesn’t seem to be on any other patchworks
PaulB:
would like to spin up concurrent patchworks to see how the react to the mailing list
Timo:
good idea, maybe we can work together on that
PaulB:
for patchtest it seems obvious what steps are required, patchwork less clear
Timo:
https://gitlab.freedesktop.org/patchwork-fdo
Timo:
might be a lot of work to compare ours with fd.o’s patch by patch to see
PaulB:
we need to figure out if we can live with some upstream version, to cut down on the effort of having our own fork
RP:
but it’ll cause a lot of pain if, for example, the superceded feature is missing
Timo:
we’re better off putting resources elsewhere. upstream looks hard, but i think patchwork-fdo looks like it might work, maybe easier to submit to patchwork-fdo rather than upstream?
TrevorW:
maybe fdo would really like a superceded feature, they split from upstream for same reasons as us, they probably share the same frustrations
Timo:
agreed, we should reach out to them
RP:
we should get a list first before reaching out
Timo:
we probably have about 30 patches on top of upstream, although most of them of them are probably just to add the supercede feature
PaulB:
is there anything in bugzilla? 13684 is a django update
PaulB:
to take action to capture this discussion in bugzilla
Timo:
Amber working on layerindex, but maybe we can have her work on this too (not a promise), she has a lot of django experience

Timo:
re perl dependency. there’s an existing oe-package-data util script, i’m using almost exact same code in a bitbake task, should this be refactored to lib/oe?
RP:
sounds good, but depends on the implementation, in general i’m in favour of strong APIs in lib/oe. e.g. “packagedata” is too generic. the history was that packagedata.py was created as a way of pulling code out of a bbclass

#1 lore raw
Quoted 114 lines · Trevor Woerner

Yocto Technical Team Minutes, Engineering Sync, for June 9 2020 archive:
https://docs.google.com/document/d/1ly8nyhO14kDNnFcW2QskANXW3ZT7QwKC5wWVDg9dDH4/edit

== disclaimer ==
Best efforts are made to ensure the below is accurate and valid. However, errors sometimes happen. If any errors or omissions are found, please feel free to reply to this email with any corrections.

== attendees ==
Trevor Woerner, Jan-Simon Möller, Stephen Jolly, Trevor Gamblin, Bruce Ashfield, Scott Murray, Paul Barker, Michael Halstead, Jon Mason, Richard Purdie, Joshua Watt, Armin Kuster, Tim Orling, Sathishkumar Duraisamy, henriknj, Ross Burton, Jeremy Puhlman

== notes ==
- 2.7.4 released last week
- 3.1.1 built, but has issues, not sent to QA yet
- M1 early next week (to build)
- thanks to AlexK for lots of fixes

== general ==
RP:
problems with 3.1.1. qemu(-mips?) failed with taskhash mismatches, not sure why

RP:
devtool, patchtest, patchwork, and wic looking for maintainers
PaulB:
patchtest/patchwork still on python2, needs 2→3 update. patchtest should be moved to OE-core so it can be noticed by more people

RP:
lots of recipe upgrades, happy with where master is (wrt to recipe versions), still need to update qemu to 5

JPEW:
did RP see my patch to mc-depends to make it backwards compatible?
RP:
dismayed it was more complex than hoped, but looks okay
JPEW:
i’m using a proxy object otherwise we need to duplicate a lot of code with a lot of if’s
RP:
saw the BBMASK thing for multiconfig that needed index checking, sad that we’ll be stuck with that going forward (in command.py) mostly for backwards compatibility (e.g. file checking). another option is to have a new API and append a ‘2’ to the end so callers will call the right things (old or new). the code’s probably going to get worse and worse with every new parameter we want to add (in the future)

JPEW:
re multiconfig, sometimes the old full “multiconfig” string doesn’t work but the new “mc” string does, can we drop “multiconfig”? changed in 2.7? or 3.0? there’s a bugzilla somewhere. some code handles both, some doesn’t
RP:
we can get rid of it, despite the expected screaming

RP:
re devtool, reports on mailing list that devtool isn’t mc-compliant
RP:
we have a horrible hack for PRserver
RP:
highlights the need for a dedicated maintainer

PaulB:
re patchwork, kernel.org are using a newer version of patchwork that
is
python3-compatible, is maintained, but missing some things we need
Timo:
i looked at freedesktop.org’s gitlab, they’ve made a lot of changes from where we forked from them, they’re also using python3 and django2. going with mainline will miss features that we’re used to
PaulB:
make a list of the gap items, then decide if we can live without these things
Timo:
i looked at the fd.o’s patchwork and saw that many of the things we added to our fork have been added (but differently) in fd.o’s. e.g. we added a superseded feature, doesn’t seem to be on any other patchworks
PaulB:
would like to spin up concurrent patchworks to see how the react to the mailing list
Timo:
good idea, maybe we can work together on that
PaulB:
for patchtest it seems obvious what steps are required, patchwork less clear
Timo:
https://gitlab.freedesktop.org/patchwork-fdo
Timo:
might be a lot of work to compare ours with fd.o’s patch by patch to see
PaulB:
we need to figure out if we can live with some upstream version, to cut down on the effort of having our own fork
RP:
but it’ll cause a lot of pain if, for example, the superceded feature
is
missing
Timo:
we’re better off putting resources elsewhere. upstream looks hard, but i think patchwork-fdo looks like it might work, maybe easier to submit
to
patchwork-fdo rather than upstream?
TrevorW:
maybe fdo would really like a superceded feature, they split from upstream for same reasons as us, they probably share the same

Quoted 8 lines

frustrations
Timo:
agreed, we should reach out to them
RP:
we should get a list first before reaching out
Timo:
we probably have about 30 patches on top of upstream, although most of them of them are probably just to add the supercede feature

Captured notes and recent research/discovery in the bug:
https://bugzilla.yoctoproject.org/show_bug.cgi?id=13684

We have 49 commits since forking from patchwork-fdo. Upstream patchwork-fdo has 166 commits since we forked.

Quoted 22 lines

PaulB:
is there anything in bugzilla? 13684 is a django update
PaulB:
to take action to capture this discussion in bugzilla
Timo:
Amber working on layerindex, but maybe we can have her work on this too (not a promise), she has a lot of django experience

Timo:
re perl dependency. there’s an existing oe-package-data util script, i’m using almost exact same code in a bitbake task, should this be refactored to lib/oe?
RP:
sounds good, but depends on the implementation, in general i’m in
favour
of strong APIs in lib/oe. e.g. “packagedata” is too generic. the history was that packagedata.py was created as a way of pulling code
out
of a bbclass