test: session compaction — observation mask and arg truncation

[anandgupta42]

Summary

What does this PR do?

1. SessionCompaction.createObservationMask() — src/session/compaction.ts (10 new tests)

This function generates the replacement text that substitutes pruned tool outputs during session compaction. When a long session triggers pruning, old tool call results are replaced with a compact summary like [Tool output cleared — bash(command: "git status") returned 3 lines, 45 B — "On branch main"]. Zero tests existed for this function or its internal helpers (truncateArgs, formatBytes).

Why it matters: Every long-running session that triggers compaction relies on these masks to preserve context about what tools were previously called. A malformed mask means the model loses track of prior work — causing it to re-read files, re-run commands, or lose context about the task.

Scenarios covered:

• Full mask format verification (tool name, args, line count, byte size, fingerprint)
• Empty output produces no fingerprint suffix
• Pending status falls through to empty args path (not completed/running/error)
• Multi-line output counts lines correctly
• Long args are truncated at 80 chars with "…" suffix
• Circular/unserializable input gracefully returns [unserializable]
• KB formatting for ~2KB outputs
• MB formatting for ~1.5MB outputs
• Multi-byte UTF-8 characters (CJK) counted correctly in bytes
• Fingerprint capped at 80 characters from first output line

Type of change

• New feature (non-breaking change which adds functionality)

Issue for this PR

N/A — proactive test coverage discovered via test-discovery rotation (session area)

How did you verify your code works?

bun test test/session/compaction-mask.test.ts   # 10 pass, 26 expect() calls

Checklist

• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes

Summary by CodeRabbit

• Tests
  • Added comprehensive test suite for observation mask generation functionality, covering various scenarios including tool execution handling, output formatting, size calculations, and edge cases.

[claude]

Claude Code Review

This repository is configured for manual code reviews. Comment @claude review to trigger a review and subscribe this PR to future pushes, or @claude review once for a one-time review.

_{Tip: disable this comment in your organization's Code Review settings.}

[coderabbitai]

📝 Walkthrough

Walkthrough

A new Bun test suite for SessionCompaction.createObservationMask was added, covering mask generation for tool executions, fingerprint handling, output formatting, multi-line content, byte-size encoding, UTF-8 character counting, and unserializable input handling.

Changes

Cohort / File(s)	Summary
New test suite packages/opencode/test/session/compaction-mask.test.ts	Added comprehensive test coverage for SessionCompaction.createObservationMask, validating mask generation for completed and pending tool executions, fingerprint behavior with/without output, line/byte-size formatting, UTF-8 byte counting, and truncation handling.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Suggested labels

contributor

Poem

🐰 A test suite hops along with care,
Checking masks both fair and square,
Fingerprints and bytes align,
UTF-8 counts so fine,
Validations everywhere! ✨

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 50.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and specifically summarizes the main change: adding tests for session compaction observation mask and argument truncation functionality.
Description check	✅ Passed	The description comprehensively covers all template sections: Summary details what changed and why, Test Plan documents verification, and Checklist confirms tests were added and pass locally.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch test/hourly-20260328-0108

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

🧹 Nitpick comments (1)

packages/opencode/test/session/compaction-mask.test.ts (1)

78-87: Add running/error status assertions to close the remaining branch gap.

You validate pending’s empty-args path, but createObservationMask also has explicit running/error arg handling via part.state.input. A small table-driven test would lock this down.

✅ Suggested test addition

+  test.each(["running", "error"] as const)(
+    "uses input args for %s status",
+    (status) => {
+      const part = {
+        ...makeCompletedPart({ tool: "bash", input: { command: "pwd" }, output: "" }),
+        state:
+          status === "running"
+            ? {
+                status: "running",
+                input: { command: "pwd" },
+                output: "",
+                title: "test",
+                metadata: {},
+                time: { start: 1000 },
+              }
+            : {
+                status: "error",
+                input: { command: "pwd" },
+                error: "boom",
+                title: "test",
+                metadata: {},
+                time: { start: 1000, end: 2000 },
+              },
+      } as MessageV2.ToolPart
+
+      const mask = SessionCompaction.createObservationMask(part)
+      expect(mask).toContain('bash(command: "pwd")')
+      expect(mask).toContain("1 lines")
+      expect(mask).toContain("0 B")
+    },
+  )

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/opencode/test/session/compaction-mask.test.ts` around lines 78 - 87,
Test coverage currently verifies the "pending" branch for createObservationMask
but misses the explicit "running" and "error" branches that take args from
part.state.input; add table-driven assertions in the same test file to create
parts with state "running" and "error" (use makePendingPart or a similar helper
to set part.state.input and part.state.status) and assert that
SessionCompaction.createObservationMask(part) contains the expected
"tool(args...)" string derived from part.state.input plus the usual "lines" and
"B" size outputs, so both running and error code paths are exercised.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Nitpick comments:
In `@packages/opencode/test/session/compaction-mask.test.ts`:
- Around line 78-87: Test coverage currently verifies the "pending" branch for
createObservationMask but misses the explicit "running" and "error" branches
that take args from part.state.input; add table-driven assertions in the same
test file to create parts with state "running" and "error" (use makePendingPart
or a similar helper to set part.state.input and part.state.status) and assert
that SessionCompaction.createObservationMask(part) contains the expected
"tool(args...)" string derived from part.state.input plus the usual "lines" and
"B" size outputs, so both running and error code paths are exercised.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: 16862d0d-f223-42eb-95de-d8ef900cc2cb

📥 Commits

Reviewing files that changed from the base of the PR and between 9ba2114 and 5cc85f7.

📒 Files selected for processing (1)

• packages/opencode/test/session/compaction-mask.test.ts

[anandgupta42]

Consolidated into #545