[OpenWrt-Devel] OpenWrt 20.X release plans

[Hauke Mehrtens]

On 1/16/20 12:48 PM, Petr Štetiar wrote:
> Peter Geis <pgwipeout at gmail.com> [2020-01-15 21:15:41]:
> 
> Hi,

Hi,

> tl;dr I'm going to vote in favor of skipping release with 4.19 and focus on
> 5.4 kernel.
> 
> The 19.07 release was delayed by a few months, so this has affected the
> subsequent release as well. The plan is to have stable release every 6 months
> as it is easier to support, allows us iterating faster and breaking stuff more
> often :-)

The biggest disadvantage of using kernel 5.4 is that we would do the
next release probably 3 months later compared to a release with kernel
4.19, but I am fine with that.

> It's clear, that this fast release cycle needs some time to settle, but I
> believe, that in the end everybody is going to benefit from that. Users by
> having modern kernel and userland, developers less backports to maintain and
> everybody predictable releases (to some extent).

Yes especially when you want to get your kernel patches also into
upstream Linux it is much easier when OpenWrt uses a recent Linux kernel
as there is less stuff to backport. So this would reduce the workload.

> Once mastered, this faster release pace might be one more very good reason for
> the vendors/companies to adopt (or at least strongly consider) "upstream
> first" workflow :-)

For my experience there are a lot of factors out of our control. A more
predictable release scheduled should help companies doing upstream work
to plan when they will see this in a production release.

The SoC vendors normally update the kernel major version only every few
years, so there is some patching to OpenWrt needed to add support for
more recent or older version anyway, like it is currently the case for
prplWrt using OpenWrt 19.07 with kernel 4.9.
If a company targets to get their stuff into upstream kernel, they
anyway have to support recent kernel versions.

> If that's not doable for various obstacles, then let's
> hope, that they will hopefully engage and consider sponsoring some kind of
> OpenWrt (Long Term Support) LTS release effort.

Often the time to market for a product takes over a year, so if a
product uses OpenWrt it is normally shipped with a version which is end
of life already. I would support it, if some companies sponsor a long
term support of OpenWrt, I think the community can live with the current
support times.

>> I think it is wise to target 4.19 as it has been in the nightly
>> releases, and as such has been in testing.
> 
> Yeah, we're all aware about that. If we decide to skip 4.19, we're partialy
> throwing away a lot of work, but on the other hand there was a lot of stuff
> upstreamed/fixed, so making the 5.4 better. It's more then clear, that staying
> closer to the upstream brings mostly only benefits.

I think most of this work will be reused for kernel 5.4, we just will
get more challenges with 5.4. ;-) I would not consider the integration
of kernel 4.19 a wast.

> So I'm going to support focus on 5.4 with my vote, thus suggest, that we skip
> release with 4.19.  Even that I know, that we're going to violate our 6 month
> release cycle goal one more time, but hopefully for the last time :-) Lets
> bite the bullet and move forward in that 5.4 direction ASAP.
> 
> BTW people doing the kernel development are essential element for this
> project, so we should try to make their work easier as much as possible. Or at
> least don't put any obstacles in their development direction.
> 
> In other words, if I've to choose where to spend the finite and scarce amount
> of resources, whether it should go into the development of 5.4 or into the
> maintenance of 4.19, I'm wholeheartedly selecting 5.4. 
> 
>> It would be nice to have lxd support in the full release, instead of
>> needing to custom build to enable support.
> 
> Feel free to follow https://openwrt.org/submitting-patches and make it reality
> :-) Thanks!

I think Daniel already did this:
https://git.openwrt.org/fcb41decf6c622482b20af45a77e62db8d95046e
https://git.openwrt.org/7cc22d72e96093c816292e1b640adff770c7fe77

Hauke

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <http://lists.infradead.org/pipermail/openwrt-devel/attachments/20200119/0d01571d/attachment.sig>
-------------- next part --------------
_______________________________________________
openwrt-devel mailing list
openwrt-devel at lists.openwrt.org
https://lists.openwrt.org/mailman/listinfo/openwrt-devel