Re: [PATCH] layers: set LAYERSERIES_COMPAT with honister


Martin Jansa
 

Then I believe layer index should be fixed to support longer values, there are layers which are compatible with multiple releases and it works fine with bitbake/oe-core, layer index parsing shouldn't be more strict than the mechanism which is using these variables.

On Fri, Jun 4, 2021 at 11:26 AM Kai <kai.kang@...> wrote:
On 6/4/21 4:52 PM, Martin Jansa wrote:
On Fri, Jun 4, 2021 at 3:44 AM kai <kai.kang@...> wrote:
From: Kai Kang <kai.kang@...>

Set LAYERSERIES_COMPAT with 'honister' which align with oe-core.


I think it's bad practice to claim honister support this early in its development cycle. It won't happen with well-maintained layers like meta-oe, but if some less frequently updated layer sets the compatibility with honister now and then doesn't get necessary updates to stay compatible with honister in next 4 months, then whole purpose of LAYERSERIES_COMPAT will get broken just by claiming the compatibility too soon (without even knowing what final honister will contain).

Changing it after release freeze in oe-core seems much better time to do it.

The same does apply on the meta-qt5 PR you've sent today https://github.com/meta-qt5/meta-qt5/pull/416/commits

Cheers,

Got it. Thanks very much.

I've comment on the meta-qt5 PR that too long strings in LAYERSERIES_COMPAT will cause layer index complains:

WARNING: YPCompatibleVersion.name: dunfell gatesgarth hardknott: length 28 exceeds maximum (25), truncating

So what is policy to keep old codenames in LAYERSERIES_COMPAT? Is  that ok to keep latest 2(hardknott and honister), please?

Regards,

-- 
Kai Kang
Wind River Linux

Join openembedded-devel@lists.openembedded.org to automatically receive all group messages.