Skip to contentSkip to conversation

Dunfell Update Status

DISCUSSION Andrew Geissler 3 participants 5 messages Apr 28, 2020 – May 11, 2020

lore
Share & tools
lore
#0 lore raw

A fairly quick update on where we are with getting back onto upstream yocto.

For those that don’t know, dunfell is the name of the next yocto release. As Brad pointed out in a previous email, if you want your meta layers to keep working, you’ll need to update your layer.conf with it.

The move from python2 to python3 is mostly behind us, thanks everyone, especially Patrick (stwcx) for all the help there.

All of the CI system layers have been updated to now support dunfell.

We ran into a phosphor-webui issue that Gunnar has a fix for:
https://gerrit.openbmc-project.xyz/c/openbmc/phosphor-webui/+/31736

IBM systems pull in an external tool, cronus, which needs python3 updates:
https://github.com/open-power/eCMD/issues/350

The two unknown issues I'm currently stuck on are these:
https://github.com/openbmc/u-boot/issues/24 (u-boot compile fail)
https://github.com/openbmc/openbmc/issues/3657 (weird sed issue)

The u-boot one I just don't understand and am hoping to get things in a state where others can recreate and take a look.

The sed issue isn't related to dunfell, I've been seeing it without the upstream bumps. I think the upstream bumps just make us recompile gcc which seems to aggravate the issue.

Here are the yocto bumps for anyone interested in pulling them down:
https://gerrit.openbmc-project.xyz/c/openbmc/openbmc/+/31715
https://gerrit.openbmc-project.xyz/c/openbmc/openbmc/+/31716

It would be great to get over this hump and back on upstream yocto/OE so any help or ideas appreciated with the remaining issues.

Some other issues that we ran into but are now fixed in upstream:
https://patchwork.openembedded.org/patch/172082/ (boost)
https://github.com/openembedded/meta-openembedded/pull/213 (new recipe)
https://github.com/openembedded/openembedded-core/commit/7bae99cd6c82fff6fa56bbddeb936e27b28e79ff

Thanks,
Andrew

#1 lore raw

An update on the email below. Upstream yocto (including the new Dunfell release) went into openbmc master this morning with the following commit:

https://github.com/openbmc/openbmc/commit/82c905dc58a36aeae40b1b273a12f63fb1973cf4

Our hardware CI and CT buckets looked good with it.

It does seem to bring some new restrictions on the minimum gcc you can have installed natively on your system so some updates may be required.

Let me know if anyone runs into issues.

Now that we’re past the python3 migration issue, we’ll be doing rebases on upstream yocto once a week again (which should prevent needing big painful updates like this).

Andrew

Quoted 44 lines · Andrew Geissler

A fairly quick update on where we are with getting back onto upstream yocto.

For those that don’t know, dunfell is the name of the next yocto release. As Brad pointed out in a previous email, if you want your meta layers to keep working, you’ll need to update your layer.conf with it.

The move from python2 to python3 is mostly behind us, thanks everyone, especially Patrick (stwcx) for all the help there.

All of the CI system layers have been updated to now support dunfell.

We ran into a phosphor-webui issue that Gunnar has a fix for:
https://gerrit.openbmc-project.xyz/c/openbmc/phosphor-webui/+/31736

IBM systems pull in an external tool, cronus, which needs python3 updates:
https://github.com/open-power/eCMD/issues/350

The two unknown issues I'm currently stuck on are these:
https://github.com/openbmc/u-boot/issues/24 (u-boot compile fail)
https://github.com/openbmc/openbmc/issues/3657 (weird sed issue)

The u-boot one I just don't understand and am hoping to get things in a state where others can recreate and take a look.

The sed issue isn't related to dunfell, I've been seeing it without the upstream bumps. I think the upstream bumps just make us recompile gcc which seems to aggravate the issue.

Here are the yocto bumps for anyone interested in pulling them down:
https://gerrit.openbmc-project.xyz/c/openbmc/openbmc/+/31715
https://gerrit.openbmc-project.xyz/c/openbmc/openbmc/+/31716

It would be great to get over this hump and back on upstream yocto/OE so any help or ideas appreciated with the remaining issues.

Some other issues that we ran into but are now fixed in upstream:
https://patchwork.openembedded.org/patch/172082/ (boost)
https://github.com/openembedded/meta-openembedded/pull/213 (new recipe)
https://github.com/openembedded/openembedded-core/commit/7bae99cd6c82fff6fa56bbddeb936e27b28e79ff

Thanks,
Andrew

#2 lore raw
Quoted 9 lines · Andrew Geissler

An update on the email below. Upstream yocto (including the new Dunfell release) went into openbmc master this morning with the following commit:

https://github.com/openbmc/openbmc/commit/82c905dc58a36aeae40b1b273a12f63fb1973cf4

Now that we’re past the python3 migration issue, we’ll be doing rebases on upstream yocto once a week again (which should prevent needing big painful updates like this).

Awesome!

Quoted 9 lines
Quoted 9 lines · Andrew Geissler

A fairly quick update on where we are with getting back onto upstream yocto.

For those that don’t know, dunfell is the name of the next yocto release.

The move from python2 to python3 is mostly behind us, thanks everyone, especially Patrick (stwcx) for all the help there.

Thank you Andrew for driving this and to everyone who helped!

#3 lore raw

Do we need to install or upgrade any package in our build system to build this new dunfel. I am seeing lots of build error starting from sanity checker. My build machine is centos CentOS Linux release 7.7.1908.

On 5/5/20, 1:06 PM, "openbmc on behalf of Gunnar Mills" <[email protected] on behalf of [email protected]> wrote:

On 5/5/2020 12:38 PM, Andrew Geissler wrote:
> An update on the email below. Upstream yocto (including the new Dunfell release) > went into openbmc master this morning with the following commit:
>
> https://github.com/openbmc/openbmc/commit/82c905dc58a36aeae40b1b273a12f63fb1973cf4
>
>
> Now that we’re past the python3 migration issue, we’ll be doing rebases > on upstream yocto once a week again (which should prevent needing
> big painful updates like this).
Awesome!

>> On Apr 28, 2020, at 2:05 PM, Andrew Geissler <[email protected]> wrote:
>>
>> A fairly quick update on where we are with getting back onto upstream yocto. >>
>> For those that don’t know, dunfell is the name of the next yocto release. >>
>> The move from python2 to python3 is mostly behind us, thanks everyone, >> especially Patrick (stwcx) for all the help there.
>>
Thank you Andrew for driving this and to everyone who helped!

#4 lore raw
Quoted 7 lines · Vijay Khemka

Do we need to install or upgrade any package in our build system to build this new dunfel. I am seeing lots of build error starting from sanity checker. My build machine is centos CentOS Linux release 7.7.1908.

The main thing that bit us was we needed gcc 6 or greater but that was a pretty obvious error bitbake spit out before it started building anything.

If you have docker installed you could run the build-setup.sh script that CI uses to verify all builds well with that.
https://github.com/openbmc/openbmc-build-scripts/blob/master/build-setup.sh

Quoted 27 lines

On 5/5/20, 1:06 PM, "openbmc on behalf of Gunnar Mills" <[email protected] on behalf of [email protected]> wrote:

Quoted 9 lines · Andrew Geissler

An update on the email below. Upstream yocto (including the new Dunfell release) went into openbmc master this morning with the following commit:

https://github.com/openbmc/openbmc/commit/82c905dc58a36aeae40b1b273a12f63fb1973cf4

Now that we’re past the python3 migration issue, we’ll be doing rebases on upstream yocto once a week again (which should prevent needing big painful updates like this).

Awesome!

Quoted 9 lines
Quoted 9 lines · Andrew Geissler

A fairly quick update on where we are with getting back onto upstream yocto.

For those that don’t know, dunfell is the name of the next yocto release.

The move from python2 to python3 is mostly behind us, thanks everyone, especially Patrick (stwcx) for all the help there.

Thank you Andrew for driving this and to everyone who helped!