Skip to contentSkip to conversation

Preferred version for upgrading Glib

scarthgap Christian Eggers 2 participants 5 messages Nov 8, 2024

lore
Share & tools
lore
#0 lore raw

The current Glib version (2.78.6) in Scarthgap is affected by #3064 [1] which lead to crashes in NetworkManager on my system. The fixing merge request [2] is available starting from 2.79.1 or 2.80.0.

Currently we have the following Glib versions:
- scarthgap: 2.78.6
- styhead: 2.80.4
- master(-next): 2.82.1

I haven't found any information whether Glib has stable/unstable/lts releases.

Which would be the preferred version for upgrading Scarthgap?
1. Creating patches against 2.78.6?
-> I have not the expertise for backporting this
2. Take 2.80.4 from Styhead?
-> Avoid the Scarthgap overtakes a newer release
3. Take 2.82.1 from Master?
4. None of the above?

regards,
Christian

[1] https://gitlab.gnome.org/GNOME/glib/-/issues/3064
[2] https://gitlab.gnome.org/GNOME/glib/-/merge_requests/3769

_______________________________________________________
Christian
Eggers
Software Engineer
​
ARRI
Arnold & Richter Cine Technik GmbH & Co. Betriebs KG Arriweg 17,
83071
Stephanskirchen
www.arri.com

+49 8036 3009-3118
[email protected]

​

Arnold & Richter Cine Technik GmbH & Co. Betriebs KG
Sitz: München - Registergericht: Amtsgericht München - Handelsregisternummer: HRA 57918 Persönlich haftender Gesellschafter: Arnold & Richter Cine Technik GmbH
Sitz: München - Registergericht: Amtsgericht München - Handelsregisternummer: HRB 54477 Geschäftsführer: Stefan Mairiedl; Walter Trauninger

​

#1 lore raw

You can do anything you want in your private layers, but in public the options are:
- backport the fixes
- add a newer glib version to the meta-ltx-mixins layer.

Alex

Quoted 69 lines · Christian Eggers via lists.openembedded.org

The current Glib version (2.78.6) in Scarthgap is affected by #3064 [1] which lead to crashes in NetworkManager on my system. The fixing merge request [2] is available starting from 2.79.1 or 2.80.0.

Currently we have the following Glib versions:
- scarthgap: 2.78.6
- styhead: 2.80.4
- master(-next): 2.82.1

I haven't found any information whether Glib has stable/unstable/lts releases.

Which would be the preferred version for upgrading Scarthgap?
1. Creating patches against 2.78.6?
-> I have not the expertise for backporting this
2. Take 2.80.4 from Styhead?
-> Avoid the Scarthgap overtakes a newer release
3. Take 2.82.1 from Master?
4. None of the above?

regards,
Christian

[1] https://gitlab.gnome.org/GNOME/glib/-/issues/3064
[2] https://gitlab.gnome.org/GNOME/glib/-/merge_requests/3769

_______________________________________________________​​​​ Christian
Eggers
Software Engineer
​
ARRI
Arnold & Richter Cine Technik GmbH & Co. Betriebs KG Arriweg 17 ,
83071
Stephanskirchen
www.arri.com

<http://www.arri.com/>
+49 8036 3009-3118 <+49%208036%203009-3118>
[email protected]

​
[image: Ensō Prime lenses]
<https://www.arri.com/en/cine-lenses/enso-prime-lenses>

Arnold & Richter Cine Technik GmbH & Co. Betriebs KG

Sitz: München ‑ Registergericht: Amtsgericht München ‑ Handelsregisternummer: HRA 57918 Persönlich haftender Gesellschafter: Arnold & Richter Cine Technik GmbH

Sitz: München ‑ Registergericht: Amtsgericht München ‑ Handelsregisternummer: HRB 54477 Geschäftsführer: Stefan Mairiedl; Walter Trauninger

​

-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group. View/Reply Online (#206860):
https://lists.openembedded.org/g/openembedded-core/message/206860
Mute This Topic: https://lists.openembedded.org/mt/109461202/1686489
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [ [email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

#2 lore raw

Hi Alex,

Quoted 3 lines · Alexander Kanavin

You can do anything you want in your private layers, but in public the options are:
- backport the fixes

I tried to backport them, but beside the mentioned merge request I need a number of other patches which feels like creating a zombie. Due to the complexity I cannot prove whether the result is valid.

Quoted 1 line

- add a newer glib version to the meta-ltx-mixins layer.

Would this help other users of Scarthgap? I thought that the mixin layers are intended to bring new features to a stable layer. But in this case there is a serious bug (race condition in the memory management routines) of a major piece of software. Every multi threaded application based on Glib is potentially affected. On my system it's NetworkManager, but I also read about Firefox.

I've already cherry picked the Styhead updates into my private repo, but I feel this should also be updated in the LTS tree somehow.

regards,
Christian

Quoted 4 lines

Alex

On Fri, 8 Nov 2024 at 11:43, Christian Eggers via lists.openembedded.org

_______________________________________________________
Christian
Eggers
Software Engineer
​
ARRI
Arnold & Richter Cine Technik GmbH & Co. Betriebs KG Arriweg 17,
83071
Stephanskirchen
www.arri.com

+49 8036 3009-3118
[email protected]

​

Arnold & Richter Cine Technik GmbH & Co. Betriebs KG
Sitz: München - Registergericht: Amtsgericht München - Handelsregisternummer: HRA 57918 Persönlich haftender Gesellschafter: Arnold & Richter Cine Technik GmbH
Sitz: München - Registergericht: Amtsgericht München - Handelsregisternummer: HRB 54477 Geschäftsführer: Stefan Mairiedl; Walter Trauninger

​

Quoted 31 lines
Quoted 29 lines · <[email protected]> wrote:

The current Glib version (2.78.6) in Scarthgap is affected by #3064 [1] which lead to crashes in NetworkManager on my system. The fixing merge request [2] is available starting from 2.79.1 or 2.80.0.

Currently we have the following Glib versions:
- scarthgap: 2.78.6
- styhead: 2.80.4
- master(-next): 2.82.1

I haven't found any information whether Glib has stable/unstable/lts releases.

Which would be the preferred version for upgrading Scarthgap?
1. Creating patches against 2.78.6?
-> I have not the expertise for backporting this
2. Take 2.80.4 from Styhead?
-> Avoid the Scarthgap overtakes a newer release
3. Take 2.82.1 from Master?
4. None of the above?

regards,
Christian

[1] https://gitlab.gnome.org/GNOME/glib/-/issues/3064
[2] https://gitlab.gnome.org/GNOME/glib/-/merge_requests/3769

_______________________________________________________​​​​

#3 lore raw
Quoted 11 lines · Christian Eggers

Hi Alex,

Quoted 3 lines · Alexander Kanavin

You can do anything you want in your private layers, but in public the options are:
- backport the fixes

I tried to backport them, but beside the mentioned merge request I need a number of other patches which feels like creating a zombie. Due to the complexity I cannot prove whether the result is valid.

Yes. This is the problem with LTS releases 'bugfixes only' policy that cannot be easily resolved. Do you *have* to use scarthgap, or can you transition to styhead?

Quoted 7 lines

Would this help other users of Scarthgap? I thought that the mixin layers are intended to bring new features to a stable layer. But in this case there is a serious bug (race condition in the memory management routines) of a major piece of software. Every multi threaded application based on Glib is potentially affected. On my system it's NetworkManager, but I also read about Firefox.

LTS mixins isn't about new features. It's more about making it possible to 'opt into' potentially disruptive version upgrades. If such an upgrade would land in oe-core scarthgap, users would have no choice. With LTS mixins they have a choice: take the upgrade, or not. If the issues don't affect them and they don't need new features, not taking the update is a valid choice.

Anyway, you are welcome to prepare such an update for meta-lts-mixins.

Alex

#4 lore raw

On Friday, 8 November 2024, 15:49:11 CET, Alexander Kanavin wrote:

_______________________________________________________
Christian
Eggers
Software Engineer
​
ARRI
Arnold & Richter Cine Technik GmbH & Co. Betriebs KG Arriweg 17,
83071
Stephanskirchen
www.arri.com

+49 8036 3009-3118
[email protected]

​

Arnold & Richter Cine Technik GmbH & Co. Betriebs KG
Sitz: München - Registergericht: Amtsgericht München - Handelsregisternummer: HRA 57918 Persönlich haftender Gesellschafter: Arnold & Richter Cine Technik GmbH
Sitz: München - Registergericht: Amtsgericht München - Handelsregisternummer: HRB 54477 Geschäftsführer: Stefan Mairiedl; Walter Trauninger

​

Quoted 9 lines · Christian Eggers
Quoted 8 lines

Hi Alex,

Quoted 3 lines · Alexander Kanavin

You can do anything you want in your private layers, but in public the options are:
- backport the fixes

Quoted 8 lines
Quoted 4 lines

I tried to backport them, but beside the mentioned merge request I need a number of other patches which feels like creating a zombie. Due to the complexity I cannot prove whether the result is valid.

Yes. This is the problem with LTS releases 'bugfixes only' policy that cannot be easily resolved. Do you *have* to use scarthgap, or can you transition to styhead?

Currently we are in the process of making a new release, so switching to a newer OE version is not possible. Switching "only" to a newer Glib release will let me sleep well tonight.

Quoted 8 lines
Quoted 7 lines

Would this help other users of Scarthgap? I thought that the mixin layers are intended to bring new features to a stable layer. But in this case there is a serious bug (race condition in the memory management routines) of a major piece of software. Every multi threaded application based on Glib is potentially affected. On my system it's NetworkManager, but I also read about Firefox.

Quoted 6 lines

LTS mixins isn't about new features. It's more about making it possible to 'opt into' potentially disruptive version upgrades. If such an upgrade would land in oe-core scarthgap, users would have no choice. With LTS mixins they have a choice: take the upgrade, or not. If the issues don't affect them and they don't need new features, not taking the update is a valid choice.

Thanks for the clarification.

Quoted 2 lines

Anyway, you are welcome to prepare such an update for meta-lts-mixins.

Cherry picking the commits from Styhead into our private repo was a no brainer, so anybody can do. I only thought about other people who may have the same (almost unfixable) problems with NetworkManager. Providing a mixin wouldn't really help them because probably nobody would look for it in such a case.

Quoted 3 lines

Alex

regards,
Christian