Move conflicted commit detection from headers to commit message markers

[mtsgrd]

Summary

The gitbutler-conflicted extra header is stripped when commits are rebased outside of GitButler, silently losing conflict state. This replaces it with commit message markers that survive rebasing and are visible to developers in git log:

• A [conflict] prefix on the subject line for fast visual detection
• A GitButler-Conflict: multi-line git trailer whose value explains the conflict tree layout

Using a standard git trailer means the description is embedded as the trailer value with indented continuation lines, so git interpret-trailers and other standard tools can parse and manipulate it. The resulting commit message looks like:

[conflict] <original subject>

<original body>

<existing trailers, if any>
GitButler-Conflict: This is a GitButler-managed conflicted
   commit. Files are auto-resolved using the "ours" side.
   The commit tree contains additional directories:
     .conflict-side-0 — our tree
     .conflict-side-1 — their tree
     ...
   To manually resolve, check out this commit, remove the
   directories listed above, resolve the conflicts, and
   amend the commit.

The trailer is appended after any existing trailers in the message, preserving Change-Id, Signed-off-by, and similar lines. When stripping conflict markers (on resolution, display, or commit matching), the prefix, trailer, and all continuation lines are removed to restore the original message.

New commits only write message markers (no header). Reading checks the message first, then falls back to the header for backward compatibility with existing conflicted commits.

The CONFLICT-README.txt tree blob is removed since the trailer now serves the same explanatory purpose with better visibility.

The conflicted file count (Option<u64>) stored in the old header was only ever used as a boolean, so the new approach uses a simple presence check with no count.

Test plan

• All Rust unit/integration tests pass (including 10 new conflict marker round-trip tests)
• All 45 Playwright e2e tests pass
• cargo clippy --workspace --all-targets -- -D warnings clean
• cargo fmt --all clean
• Conflicted commits survive rebase outside GitButler (message markers preserved)
• Old header-based conflicted commits still detected (backward compat)
• Existing trailers (Change-Id, Signed-off-by) preserved in correct order

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
gitbutler-web	Ignored	Preview	Apr 14, 2026 0:50am

[Copilot]

Pull request overview

This PR changes GitButler’s “conflicted commit” detection from relying on a custom commit header (gitbutler-conflicted, which can be lost during external rebases) to commit-message markers ([conflict] prefix + GitButler-Conflict: true footer), while keeping legacy header fallback for existing commits.

Changes:

• Add/strip/check conflict markers in commit messages and update conflict detection to prefer message markers with legacy-header fallback.
• Update rebase/cherry-pick/edit-mode flows to add markers when creating conflicted commits and strip markers when conflicts are cleared.
• Adjust UI/workspace projections and snapshots/tests to hide markers in displayed messages while preserving has_conflicts.

Reviewed changes

Copilot reviewed 14 out of 14 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
crates/gitbutler-edit-mode/src/lib.rs	Strips conflict markers when saving edited commits back to workspace.
crates/gitbutler-commit/src/commit_ext.rs	Updates is_conflicted() to check message markers first, then legacy headers.
crates/gitbutler-branch-actions/tests/branch-actions/reorder.rs	Normalizes commit-message comparisons by stripping markers in tests.
crates/but/tests/but/command/rub.rs	Updates snapshot due to changed CLI IDs in output.
crates/but-workspace/tests/workspace/commit/squash_commits.rs	Updates snapshots to reflect [conflict] prefix on conflicted commits.
crates/but-workspace/src/ref_info.rs	Strips markers before exposing commit messages to UI models when conflicted.
crates/but-workspace/src/branch_details.rs	Strips markers from local commit messages while keeping has_conflicts.
crates/but-rebase/tests/rebase/main.rs	Updates snapshots for new message-marker format and removal of CONFLICT-README.txt.
crates/but-rebase/tests/rebase/graph_rebase/conflictable_restriction.rs	Updates snapshots for new message-marker format.
crates/but-rebase/tests/rebase/graph_rebase/cherry_pick.rs	Updates snapshots for new message-marker format and tree layout.
crates/but-rebase/src/lib.rs	Preserves markers on reword/squash operations when commit remains conflicted.
crates/but-rebase/src/graph_rebase/cherry_pick.rs	Adds markers on conflicted commit creation; strips markers/legacy header when unconflicting.
crates/but-rebase/src/cherry_pick.rs	Same as graph_rebase path; removes legacy conflicted-count writing helpers.
crates/but-core/src/commit.rs	Implements marker constants + add/strip/detect helpers; updates Commit::is_conflicted().

[mtsgrd]

@krlvi this is what we spoke about earlier, let me know what you think?

[Copilot]

Pull request overview

Copilot reviewed 17 out of 17 changed files in this pull request and generated 3 comments.

[Copilot]

Pull request overview

Copilot reviewed 20 out of 20 changed files in this pull request and generated 2 comments.

[Byron]

I like it!

Just a question regarding the format:

GitButler-Conflict: true
This is a GitButler-managed conflicted commit. Files are auto-resolved
using the "ours" side. The commit tree contains additional directories:
  .conflict-side-0  — our tree
  .conflict-side-1  — their tree
  .conflict-base-0  — the merge base tree
  .auto-resolution  — the auto-resolved tree
  .conflict-files   — metadata about conflicted files
To manually resolve, check out this commit, remove the directories
listed above, resolve the conflicts, and amend the commit.

GitButler-Conflict: true looks like a trailer, but what follows doesn't look like it will be part of the value. While this might be intentional, I'd prefer it if trailers aren't interleaved with non-trailers. You can also try if gitoxide can parse these correctly in its interleaved form, should be commit.message().trailers().

[mtsgrd]

Cool cool, I'm afk for most of the day but I'll pick this back up sometime soon.

[mtsgrd]

Actually, I think I am happy with it in principle, but feel the need to refactor. Can I take it from here?

Absolutely! Thank you.

[Byron]

Tasks

• first review
• structure in but-core
• use latest gix trailer parsing facilities
  • don't normalize line-endings!
• can unconditionally strip in ref_info.rs
• try to unify is_conflicted() checks, it taking out the conflicted trailer in many places now. Check for but_core::commit::message_is_conflicted to find them.
• add_markers_if_needed()
• persist synthetic change-ids when it's obvious enough. doesn't belong here and just increases complexity

[Byron]

I am force-pushing based on the most recent changes that were , erm, force-pushed on top of mine.
The changes in @mtsgrd commit are only the whitespace that was added to Cargo.toml for no reason. Maybe it's a formatter thing.

My refactors are on top and I will merge (if I can) in fear of another force-push 😅.

@Caleb-T-Owens To my mind, the changes to but-rebase are Ok but you might want to take another look nonetheless. It would also be good to find a way to tackle the 'persist synthetic change-ids' story that is still lingering.

[Copilot]

Pull request overview

Copilot reviewed 21 out of 21 changed files in this pull request and generated 4 comments.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 4bc063068d

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[Byron]

Hitting auto-merge as I spent some time with it and think it's good.

It's notable that for GUI purposes, we strip the [conflict] prefix from commit messages in types that also have a has_conflict field. This introduces a mismatch between the titles people see in git log and friends, and in our clients. So that's definitely to watch out for and maybe we change it.

[Copilot]

Pull request overview

Copilot reviewed 23 out of 23 changed files in this pull request and generated 1 comment.