Hi,

On Fri, Mar 17, 2023 at 03:25:10PM +0000, Alexander Kanavin wrote:
> On Fri 17. Mar 2023 at 12.01, Mikko Rapeli <[email protected]> wrote:
> 
> > Hi,
> >
> > On Fri, Mar 17, 2023 at 11:39:23AM +0100, Alexander Kanavin wrote:
> > > This doesn’t seem right. We should not give special privileges to tooling
> > > from any particular host distro. If you need to manipulate images, build
> > > e2fsprogs-native and run that.
> >
> > This is about ext4 images breaking compatibility with older versions of
> > tooling used manipulate the images. For Debian the upstream maintainers
> > and Debian package maintainers agreed to keep the backwards
> > compatibility for longer time. I think that makes sense for yocto too.
> >
> > While I agree that using e2fsprogs-native is a solution too, I also
> > agree that a lot of users are using non-yocto e2fsprogs when e.g.
> > flashing images to target HW. Breaking compatibility can quite simply be
> > avoided and upstream will maintain this branch. Win win from my point of
> > view.
> 
> But then we should put e2fsprogs into buildtools-tarball(-extended). It’s
> not any different to providing compatible gcc, when one from the host
> distro is too old.

I think this one is a bit different. At flashing/install step people
frequently manipulate rootfs images coming out of yocto builds to match
what HW and various boards actually have, like extending the file system
to fill full EMMC/SD cards. Like dd the .wic out of a build but extend
the file system for the full disk before first boot on the system.

> Taking a custom branch creates a maintenance problem. What if it stops
> being maintained? Will it follow proper upstream closely? Can we ever drop
> this backwards compatibility? It creates a long term headache that I do not
> want.

This branch is from upstream developers who are also the Debian package
maintainers. Of course anything can break and no support can be
guaranteed.

https://git.kernel.org/pub/scm/fs/ext2/e2fsprogs.git/commit/?h=debian/master&id=1181f164c48574a4813bfa203dbd7b4123154021

"author Theodore Ts'o <[email protected]>   2023-03-04 22:04:29 -0500

Disable metadata_csum_seed and orphan_file by default for Debian
There are a lot of debian packages that assume that mkfs.ext4 for
Debian N will work on Debian N-1 (and sometimes N-2, N-3, etc.).  This
is a bad assumption, but it's too late to fix this for Debian
Bookworm.

So revert making these feature a default for now.  We will enable it
again post Bookworm.

Signed-off-by: Theodore Ts'o <[email protected]>"

That explains the timeline a bit. I think this is sensible to follow. If the
branch dies and there are no more updates to CVEs etc triggers, then switching
to master branch is ok. Right now the branch is actively maintained.

Cheers,

-Mikko
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#178799): 
https://lists.openembedded.org/g/openembedded-core/message/178799
Mute This Topic: https://lists.openembedded.org/mt/97669707/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub 
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to