Skip to contentSkip to conversation

Poky doesn't play nice with meta-oe

DISCUSSION Gary Thomas 4 participants 6 messages Mar 10, 2012 – Apr 20, 2012

lore
Share & tools
lore
#0 lore raw

I wanted to do some experimenting with midori (a webkit based browser that is [arguably] better than web2) in my familiar Poky environment. I've run midori in the past by cobbling in the recipes and dependencies, but I wanted to see if I could just use the whole meta-oe layer, as is. Here's what I tried:

* Build core-image-sato for qemuarm from latest master Everything works as expected

* Add in meta-oe layers, in particular (I used my up-to-date checkout of Angstrom for this):

    /local/new-angstrom/sources/meta-openembedded/meta-xfce \
    /local/new-angstrom/sources/meta-openembedded/meta-multimedia \
    /local/new-angstrom/sources/meta-openembedded/meta-gpe \
    /local/new-angstrom/sources/meta-openembedded/meta-initramfs \
    /local/new-angstrom/sources/meta-openembedded/meta-efl \
    /local/new-angstrom/sources/meta-openembedded/meta-oe \
    /local/new-angstrom/sources/meta-openembedded/meta-gnome \

* Rebuild core-image-sato
Lots of packages get rebuilt, down to GCC which has been updated to r27 from r22. I thought the poky meta "core" was tracking meta-oe and openembedded-core more closely than this.

Image will boot, but most things are broken, no X, no network, ...

It looks like the evil systemd has pervaded meta-oe (which IMO should be at least a DISTRO decision).

I guess the only way I'll be able to test this in Poky, if at all, will be to only import the midori recipes and its dependencies, which there seem to be many.

Oh well, so much for my "simple" test.

Signature · 4 lines
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------
#1 lore raw
Quoted 8 lines · On Saturday 10 March 2012 08:31:50 Gary Thomas wrote:

I wanted to do some experimenting with midori (a webkit based browser that is [arguably] better than web2) in my familiar Poky environment.
...
* Rebuild core-image-sato
Lots of packages get rebuilt, down to GCC which has been updated to r27 from r22. I thought the poky meta "core" was tracking meta-oe and openembedded-core more closely than this.

I agree, this is not particularly desirable. We've discussed this in the OE TSC and the agreement was that the toolchain bits currently in the meta-oe layer itself will be moved out to a separate layer.

Quoted 1 line

Image will boot, but most things are broken, no X, no network, ...

That's not good :(

Quoted 2 lines

It looks like the evil systemd has pervaded meta-oe (which IMO should be at least a DISTRO decision).

I agree, but there are others who think it should be switchable on a per-image basis. I don't think it's going to be long before that's impossible however.

Quoted 3 lines

I guess the only way I'll be able to test this in Poky, if at all, will be to only import the midori recipes and its dependencies, which there seem to be many.

That's not an unreasonable approach for now. However we do really want to tidy up meta-oe such that it doesn't do surprising things when you enable it. Hopefully we will get a chance to make progress on this over the next few months.

Cheers,
Paul

Signature · 2 lines
Paul Eggleton
Intel Open Source Technology Centre
#2 lore raw

Hi! I'm resurrecting this thread because it's becoming somewhat interesting for me.

Quoted 24 lines · Paul Eggleton
Quoted 8 lines · On Saturday 10 March 2012 08:31:50 Gary Thomas wrote:

I wanted to do some experimenting with midori (a webkit based browser that is [arguably] better than web2) in my familiar Poky environment.
...
* Rebuild core-image-sato
   Lots of packages get rebuilt, down to GCC which has been updated to r27 from r22.  I thought the poky meta "core" was tracking meta-oe and openembedded-core more closely than this.

I agree, this is not particularly desirable. We've discussed this in the OE TSC and the agreement was that the toolchain bits currently in the meta-oe layer itself will be moved out to a separate layer.

Quoted 1 line

Image will boot, but most things are broken, no X, no network, ...

That's not good :(

Quoted 2 lines

It looks like the evil systemd has pervaded meta-oe (which IMO should be at least a DISTRO decision).

I agree, but there are others who think it should be switchable on a per-image basis. I don't think it's going to be long before that's impossible however.

Quoted 8 lines
Quoted 3 lines

I guess the only way I'll be able to test this in Poky, if at all, will be to only import the midori recipes and its dependencies, which there seem to be many.

That's not an unreasonable approach for now. However we do really want to tidy up meta-oe such that it doesn't do surprising things when you enable it. Hopefully we will get a chance to make progress on this over the next few months.

Before starting to mess up things I'm asking if there is any relevant improvement on this side? I need a few recipes that are in meta-oe but not in poky/yocto (python-twisted, gpsd and a few more). Enabling meta-oe over my current edison environment would be nice to avoid manually importing those recipes with dependencies. Maybe master branch and upcoming 1.2 are expected to work better on this side?

#3 lore raw
Quoted 7 lines · On Friday 20 April 2012 18:01:54 Andrea Galbusera wrote:
Quoted 4 lines · Paul Eggleton

That's not an unreasonable approach for now. However we do really want to tidy up meta-oe such that it doesn't do surprising things when you enable it. Hopefully we will get a chance to make progress on this over the next few months.

Quoted 6 lines

Before starting to mess up things I'm asking if there is any relevant improvement on this side? I need a few recipes that are in meta-oe but not in poky/yocto (python-twisted, gpsd and a few more). Enabling meta-oe over my current edison environment would be nice to avoid manually importing those recipes with dependencies. Maybe master branch and upcoming 1.2 are expected to work better on this side?

No major progress as yet, I've been busy with things leading up to the 1.2 release; however once that's out the door there should be more time available for improving meta-oe.

Cheers,
Paul

Signature · 2 lines
Paul Eggleton
Intel Open Source Technology Centre
#4 lore raw

Paul,

Quoted 15 lines · Paul Eggleton
Quoted 7 lines · On Friday 20 April 2012 18:01:54 Andrea Galbusera wrote:
Quoted 4 lines · Paul Eggleton

That's not an unreasonable approach for now. However we do really want to tidy up meta-oe such that it doesn't do surprising things when you enable it. Hopefully we will get a chance to make progress on this over the next few months.

Quoted 6 lines

Before starting to mess up things I'm asking if there is any relevant improvement on this side? I need a few recipes that are in meta-oe but not in poky/yocto (python-twisted, gpsd and a few more). Enabling meta-oe over my current edison environment would be nice to avoid manually importing those recipes with dependencies. Maybe master branch and upcoming 1.2 are expected to work better on this side?

Quoted 3 lines

No major progress as yet, I've been busy with things leading up to the 1.2 release; however once that's out the door there should be more time available for improving meta-oe.

Thank you for updating on this. I'm looking forward to hopefully see progress in this area in the future. For the time being, I'll go on manually importing those recipes. By the way, one of them (python-twisted) is triggering some problems when RDEPENDing on it: I have to figure out if they are yocto specific or not. If not, I'll report to meta-oe people.

#5 lore raw
In reply to Andrea Galbusera
Quoted 6 lines · Andrea Galbusera

Before starting to mess up things I'm asking if there is any relevant improvement on this side? I need a few recipes that are in meta-oe but not in poky/yocto (python-twisted, gpsd and a few more). Enabling meta-oe over my current edison environment would be nice to avoid manually importing those recipes with dependencies. Maybe master branch and upcoming 1.2 are expected to work better on this side?

I have tried combining Yocto with meta-oe about a week ago, and it does not work at all at the moment; ended up with images that were very broken. I think systemd is the principal culprit, though as the meta-oe layer has higher priority than Yocto, not the only problem (though forcing a lower priority on it, does not much help).

Tomas