[gentoo] preper for usrmerge #1173
Conversation
This comment was marked as resolved.
This comment was marked as resolved.
257
force-pushed
the
gentoo
branch
5 times, most recently
from
dd1d006 to
b389359
Compare
|
What's the status here? |
|
i will rework it this weekend. |
actually it passes the test. alert are outdated/fixed. let me know if you see otherwise. |
|
Any idea when we'll be getting the merged-usr stage3? |
|
soon i hope. i'm hoping that gentoo-releng's uses |
by the EOY, supposidly. |
|
There's still a CI failure in the gentoo gpt ext4 CI run, so I don't think we can merge this just yet |
|
i will have a look over the weekend. yes, please hold off merging this. |
possibly fails on builds here, stage3 images upstream are autobuilt. i even thought of freezing the build that works (today it's my understanding is that there is an effort upstream to stanardise the way these stage3 release work. i understand this breaks what do you think? |
hmm, actually image boots we fail on ssh odd! |
|
seems to be a bug with portage and binary packages, fixed upstream. how do i force the actions to run again without pushing to the branch? |
|
I restarted the failed job |
|
@257 Can you take another look, something is borked with ssh |
i did while back i couldn't find a way to run the test locally. any pointers? |
257
force-pushed
the
gentoo
branch
6 times, most recently
from
78220b2 to
5b12058
Compare
|
np, working on it |
|
@DaanDeMeyer which version of |
257
force-pushed
the
gentoo
branch
5 times, most recently
from
251ba93 to
4f80ba7
Compare
ready with possible exception of ssh. please review. |
ssh is also fixed. probably a side-effect of usrmerge. |
|
not sure why centos is failing |
Unrelated to this PR |
| MkosiPrinter.print_step("Invoking emerge(1) inside stage3" | ||
| f"{self.root}") | ||
| run_workspace_command(self.state, cmd, network=True, env=self.emerge_vars, | ||
| nspawn_params=self.DEFAULT_NSPAWN_PARAMS, | ||
| check=check) |
There was a problem hiding this comment.
Why does this have to run inside the image? We're trying to move all the package manager stuff to run outside of the image instead of inside the image
There was a problem hiding this comment.
we ran into number of problems using host's portage in the past. only initial sync_portage_tree() uses host's portage and i'm tempted to even remove that.
also we still do make some changes related to systemd and we also install kernel related packages for bootable cases, see pkgs['boot'] especially under update_stage3().
also any extra packages asked by user also needs this, see merge_user_pkgs().
There was a problem hiding this comment.
I would prefer if we use host's portage if possible. We do the same for all other distributions. But we can leave that for another PR since this one is already been open for a long time.
There was a problem hiding this comment.
i started out using host's portage and its python interface in face, which turned out to be horrible. even @poettering mentioned that we probably do the same for dnf...
yeah changing to host's would be a big big change with lots of test. although with usermerge in place we're dropping lots of code and complexity. i will have a go at it once day job gets a bit lighter.
There was a problem hiding this comment.
keeping this unresolved for future ref.
257
force-pushed
the
gentoo
branch
3 times, most recently
from
e0fa0ec to
40829ef
Compare
|
once this is merged a mini-release would be nice since gentoo people insist on it for some reason! |
upstream finally has merged-usr profiles/stage3s [1]. this makes `no-multilib/systemd/merged-usr`, and related stage3, default for us. [1] https://bugs.gentoo.org/843818#c2 Signed-off-by: Paymon MARANDI <[email protected]>
upstream has added merged-usr profile: https://bugs.gentoo.org/843818#c2
this makes that profile default for us. I expect a stage3 for merged-usr
sometime soon which should result in signicant code-drop at our end.
Signed-off-by: Paymon MARANDI [email protected]