Clean up branches in TrenchBoot repositories

krystian-hebel commented 1 month ago

Affected component(s) or functionality (if applicable) All of them.

Brief summary Analyze and reduce the number of branches on code repositories (mainly grub, xen and linux as they are modified most frequently). Ideally, users and developers shouldn't have to think about which branch is the main one that should be the target of pull requests.

Additional context All of the following is open for discussion. I'm not strongly opposed to changing the rules, but we need something to start with, hence this list. Also, keeping the repositories clean is indefinite process, not a one-time task, so some rules will have to be added to documentation about expectations discussed here.

Upstream branches are to be left as-is. They usually are for specific versions of component, keeping them may make backports easier.
Branches may be part of PR or draft, in such cases reviewers and/or authors should be pinged to get such PRs merged or closed if they are no longer applicable.
In the end there should be no more than one branch per upstream version that branch is based on. Exceptions:
- Branches that are part of an open PR are allowed only if they target that versioned branch based on upstream version, or a chain that starts with such branch.
- Branches that are part of active upstreaming effort (e.g. https://github.com/TrenchBoot/linux/tree/linux-sl-master-5-16-24-v9), with all previous iterations for reference, until the feature gets merged.
There should be no branches for PoCs and demos presented on conferences after it has concluded, unless it is to be merged (in which case PR or draft should be opened). If the specific code must be kept because of external references, it should be tagged instead.

SergiiDmytruk commented 3 weeks ago

I'll post comments as I go through the repos for feedback and to not have a single huge comment at the end.

Xen

Dropped:

aem-phase3-rebase (was d6111adaacf523b2c74b519373f46a23ea44ab90).

Pinged:

@krystian-hebel, absolute majority of branches come from upstream and I think we're only interested in stable-* ones. OK to remove all others? The removed would include staging-*, coverity/*, *-shim-* and smoke, which all seem useless for interacting with upstream.

SergiiDmytruk commented 3 weeks ago

GRUB

Dropped:

several tb-2.12-57* that were created recently for temporary purposes
test_gitlabci (https://github.com/TrenchBoot/grub/commit/eeba8e69cda0ab13abdd9483af27e656bbd33c98) seems to be an old branch name for https://github.com/TrenchBoot/grub/pull/1
aem_p3_upstreaming (https://github.com/TrenchBoot/grub/commit/66b9d72c3119ab93b7529074010aefb3ad3d8bb4) looked outdated

Renamed:

tb-2.12-57-linux-amd to tb-2.12-57, because the old name was reflecting contents of top patches only

Pinged:

Questionable:

these don't seem to appear in 3mdeb's blog posts but maybe they are referenced somewhere else, safe to drop?
- https://github.com/TrenchBoot/grub/tree/intel-txt-aem
- https://github.com/TrenchBoot/grub/tree/intel-txt
these are older ones and can potentially contain something useful by @dpsmith

SergiiDmytruk commented 3 weeks ago

Linux

Upstream:

https://github.com/TrenchBoot/linux/tree/master

Upstreaming:

Versions for different kernels (convert to tags?):

Unique commits?:

Temporary branches?:

Also, I'm not sure how to interpret [revision] for {2ba13df} commits.

SergiiDmytruk commented 3 weeks ago

trenchboot-sdk

Pinged:

SergiiDmytruk commented 3 weeks ago

TrenchBoot.github.io

Dropped:

3mdeb-presentations (was https://github.com/TrenchBoot/TrenchBoot.github.io/commit/75953c7d3e6a9447951afc52f043ae762dd18ebc) because the branch was merged 3 years ago
old-site-archive is pointing at a parent of current master, I've made a tag for reference to this older version before removing the branch
hugo looked weird as it's mkdocs-based version, it looked like a test of https://github.com/TrenchBoot/TrenchBoot.github.io/pull/16

The rest of the branches are from Dependabot which seems to be ignored and it won't even send any more PRs because of that. Should either use it (if its suggestions are even useful) or disable and remove all PRs and corresponding branches. Given that https://github.com/TrenchBoot/TrenchBoot.github.io/pull/31 removes most of the files, we won't need those dependencies in the future?

slexec

One extra branch pre_slrt, convert to a tag?

trenchboot-issues

2 branches, pushed to https://github.com/TrenchBoot/trenchboot-issues/pull/7 to get it merged and drop the second branch.

Other repos

Have only one branch.

krystian-hebel commented 2 weeks ago

Xen:

@krystian-hebel, absolute majority of branches come from upstream and I think we're only interested in stable- ones. OK to remove all others? The removed would include staging-, coverity/, -shim-* and smoke, which all seem useless for interacting with upstream.

Sure, let's remove them.

GRUB:

these don't seem to appear in 3mdeb's blog posts but maybe they are referenced somewhere else, safe to drop?

https://github.com/TrenchBoot/grub/tree/intel-txt-aem

https://github.com/TrenchBoot/grub/tree/intel-txt

Can be removed, all of the code was migrated to newer branches. IIRC the first one was used as a demo for one of the conferences, but we weren't able to reproduce it even few days after, most likely to different SKL (or LZ at that time).

Linux:

The *-amd branches may be used to help with implementation of the upcoming phase, I'd leave them until it is done.

Also, I'm not sure how to interpret [revision] for {2ba13df} commits.

I think this is some kind of patch queue, it allows adding changes as a new commit and squashing them later. @rossphilipson would probably know more, maybe we could use it somewhere.

trenchboot-sdk:

I've closed Optiplex PR and removed its branch. As for Qubes packages, on the one hand we have CI that does it automatically, on the other hand, it already got outdated several times. Perhaps manual instructions would allow to proceed in those cases using common sense, but it would have to be reworked to match current state of the builder(s).

SergiiDmytruk commented 2 weeks ago

Xen:

Sure, let's remove them.

-15 branches there.

GRUB:

Can be removed, all of the code was migrated to newer branches. IIRC the first one was used as a demo for one of the conferences, but we weren't able to reproduce it even few days after, most likely to different SKL (or LZ at that time).

-2 branches.

SergiiDmytruk commented 2 weeks ago

Went through the repos again to assess the updated state.

Xen

Restored https://github.com/TrenchBoot/xen/pull/17 PR. Also there was aem based on 4.17.2, so I renamed it to aem-4.17.2. Renamed aem-phase4-rebase to aem-4.17.4 because that's what it is.

There are still a bunch of stable-4.* branches from upstream. Remove all but stable-4.17?

GRUB2

Removed intel-txt-aem-base (was 7e0310c) which was used as a target of https://github.com/TrenchBoot/grub/pull/16.

Usefulness of the following old branches by @dpsmith is unclear:

Turn them into tags?

Linux

Nothing was removed, https://github.com/TrenchBoot/trenchboot-issues/issues/48#issuecomment-2299841619 still applies (but see https://github.com/TrenchBoot/trenchboot-issues/issues/48#issuecomment-2310155351).

trenchboot-sdk

On https://github.com/TrenchBoot/trenchboot-sdk/pull/7 from https://github.com/TrenchBoot/trenchboot-issues/issues/48#issuecomment-2310155351:

As for Qubes packages, on the one hand we have CI that does it automatically, on the other hand, it already got outdated several times. Perhaps manual instructions would allow to proceed in those cases using common sense, but it would have to be reworked to match current state of the builder(s).

There are very few branches there, so keeping it isn't a big deal.

TrenchBoot / trenchboot-issues