GitNexus/gitnexus/test/unit/ai-context.test.ts
azizur100389 e262dda35b
fix(cli): only match <!-- gitnexus:* --> markers at section position (#1041) (#1042)
`upsertGitNexusSection` in ai-context.ts uses `indexOf` to locate the
bounds of the GitNexus section in CLAUDE.md / AGENTS.md before
replacement. `indexOf` matches the first occurrence of the marker
anywhere in the file, including inline prose references in backtick-
quoted fragments mid-sentence.

The shipped CLAUDE.md contains exactly such a reference ("See the
`<!-- gitnexus:start --> … <!-- gitnexus:end -->` block in AGENTS.md
for the canonical MCP tools..."). Running `gitnexus analyze` on a
fresh install matches those inline markers as section delimiters and
replaces the prose between them with the full ~100-line injected
block, breaking the backtick and corrupting markdown for every user.

Fix: new private `findSectionMarkerIndex` helper that only matches
markers occupying their own line — preceded by `\n` or start-of-file,
followed by `\n` / `\r` (CRLF files) / end-of-file. `\r` is explicit
so CRLF-terminated sections on Windows (core.autocrlf = true) still
match. The generator always emits markers alone on their line, so
every legitimate section continues to update in place; only inline
prose references now fall through to the append branch, which leaves
existing content untouched.

Two new unit tests:
- #1041 regression — seed CLAUDE.md with the shipped inline prose
  line, run analyze twice, assert inline prose preserved verbatim
  and marker counts stay at 2/2 (1 inline + 1 section-position)
- CRLF handling — seed a CRLF file with inline prose + legitimate
  section, run analyze, assert section replaced in place, inline
  prose preserved, stale stub content removed

No destructive ops, no bypass flags, no new deps. Behaviour change
is strictly narrowing — files that previously updated correctly
still do; files that previously got corrupted now fall through to
the safer append branch.

Closes #1041.
2026-04-23 07:58:07 +01:00

279 lines
12 KiB
TypeScript

import { describe, it, expect, beforeAll, afterAll } from 'vitest';
import fs from 'fs/promises';
import path from 'path';
import os from 'os';
import { generateAIContextFiles } from '../../src/cli/ai-context.js';
describe('generateAIContextFiles', () => {
let tmpDir: string;
let storagePath: string;
beforeAll(async () => {
tmpDir = await fs.mkdtemp(path.join(os.tmpdir(), 'gn-ai-ctx-test-'));
storagePath = path.join(tmpDir, '.gitnexus');
await fs.mkdir(storagePath, { recursive: true });
});
afterAll(async () => {
try {
await fs.rm(tmpDir, { recursive: true, force: true });
} catch {
/* best-effort */
}
});
it('generates context files', async () => {
const stats = {
nodes: 100,
edges: 200,
processes: 10,
};
const result = await generateAIContextFiles(tmpDir, storagePath, 'TestProject', stats);
expect(result.files).toBeDefined();
expect(result.files.length).toBeGreaterThan(0);
});
it('creates or updates CLAUDE.md with GitNexus section', async () => {
const stats = { nodes: 50, edges: 100, processes: 5 };
await generateAIContextFiles(tmpDir, storagePath, 'TestProject', stats);
const claudeMdPath = path.join(tmpDir, 'CLAUDE.md');
const content = await fs.readFile(claudeMdPath, 'utf-8');
expect(content).toContain('gitnexus:start');
expect(content).toContain('gitnexus:end');
expect(content).toContain('TestProject');
});
it('keeps the load-bearing repo-specific sections in the CLAUDE.md block (#856)', async () => {
// The trimmed block must still contain everything that is genuinely
// unique per repo or load-bearing for the agent: the freshness warning,
// the Always Do / Never Do imperative lists, the Resources URI table
// (projectName-interpolated), and the skills routing table that tells
// the agent which skill file to read for each task.
const stats = { nodes: 50, edges: 100, processes: 5 };
await generateAIContextFiles(tmpDir, storagePath, 'TestProject', stats);
const content = await fs.readFile(path.join(tmpDir, 'CLAUDE.md'), 'utf-8');
expect(content).toContain('If any GitNexus tool warns the index is stale');
expect(content).toContain('## Always Do');
expect(content).toContain('## Never Do');
expect(content).toContain('## Resources');
expect(content).toContain('gitnexus://repo/TestProject/context');
expect(content).toContain('gitnexus-impact-analysis/SKILL.md');
expect(content).toContain('gitnexus-refactoring/SKILL.md');
expect(content).toContain('gitnexus-debugging/SKILL.md');
expect(content).toContain('gitnexus-cli/SKILL.md');
});
it('does not duplicate content that already lives in skill files (#856)', async () => {
// The six sections listed in issue #856 are redundant with the skill
// files shipped alongside the CLAUDE.md block (both are loaded into
// every Claude Code session). Their absence is the whole point of the
// trim — assert each header is gone so a future regression that pads
// the block back out fails here.
const stats = { nodes: 50, edges: 100, processes: 5 };
await generateAIContextFiles(tmpDir, storagePath, 'TestProject', stats);
const content = await fs.readFile(path.join(tmpDir, 'CLAUDE.md'), 'utf-8');
expect(content).not.toContain('## Tools Quick Reference');
expect(content).not.toContain('## Impact Risk Levels');
expect(content).not.toContain('## Self-Check Before Finishing');
expect(content).not.toContain('## When Debugging');
expect(content).not.toContain('## When Refactoring');
expect(content).not.toContain('## Keeping the Index Fresh');
});
it('keeps the CLAUDE.md GitNexus block under the token-cost budget (#856)', async () => {
// The pre-trim block was ~5465 chars. After #856 it's ~2580 — about a
// 52% reduction. 2700 is a soft ceiling that still leaves headroom for
// legitimate future additions but will fail loudly if the trim is
// reverted or someone pads the block back out toward the original size.
const stats = { nodes: 50, edges: 100, processes: 5 };
await generateAIContextFiles(tmpDir, storagePath, 'TestProject', stats);
const content = await fs.readFile(path.join(tmpDir, 'CLAUDE.md'), 'utf-8');
const block = content.slice(
content.indexOf('<!-- gitnexus:start -->'),
content.indexOf('<!-- gitnexus:end -->'),
);
expect(block.length).toBeLessThan(2700);
});
it('handles empty stats', async () => {
const stats = {};
const result = await generateAIContextFiles(tmpDir, storagePath, 'EmptyProject', stats);
expect(result.files).toBeDefined();
});
it('updates existing CLAUDE.md without duplicating', async () => {
const stats = { nodes: 10 };
// Run twice
await generateAIContextFiles(tmpDir, storagePath, 'TestProject', stats);
await generateAIContextFiles(tmpDir, storagePath, 'TestProject', stats);
const claudeMdPath = path.join(tmpDir, 'CLAUDE.md');
const content = await fs.readFile(claudeMdPath, 'utf-8');
// Should only have one gitnexus section
const starts = (content.match(/gitnexus:start/g) || []).length;
expect(starts).toBe(1);
});
it('installs skills files', async () => {
const stats = { nodes: 10 };
const result = await generateAIContextFiles(tmpDir, storagePath, 'TestProject', stats);
// Should have installed skill files
const skillsDir = path.join(tmpDir, '.claude', 'skills', 'gitnexus');
try {
const entries = await fs.readdir(skillsDir, { recursive: true });
expect(entries.length).toBeGreaterThan(0);
} catch {
// Skills dir may not be created if skills source doesn't exist in test context
}
});
it('preserves manual AGENTS.md and CLAUDE.md edits when skipAgentsMd is enabled', async () => {
const stats = { nodes: 42, edges: 84, processes: 3 };
const agentsPath = path.join(tmpDir, 'AGENTS.md');
const claudePath = path.join(tmpDir, 'CLAUDE.md');
const agentsContent = '# AGENTS\n\nCustom manual instructions only\n';
const claudeContent = '# CLAUDE\n\nCustom manual instructions only\n';
await fs.writeFile(agentsPath, agentsContent, 'utf-8');
await fs.writeFile(claudePath, claudeContent, 'utf-8');
const result = await generateAIContextFiles(
tmpDir,
storagePath,
'TestProject',
stats,
undefined,
{ skipAgentsMd: true },
);
expect(result.files).toContain('AGENTS.md (skipped via --skip-agents-md)');
expect(result.files).toContain('CLAUDE.md (skipped via --skip-agents-md)');
const agentsAfter = await fs.readFile(agentsPath, 'utf-8');
const claudeAfter = await fs.readFile(claudePath, 'utf-8');
expect(agentsAfter).toBe(agentsContent);
expect(claudeAfter).toBe(claudeContent);
});
it('preserves inline marker references in prose and does not corrupt markdown (#1041)', async () => {
// Regression guard for #1041. The shipped CLAUDE.md ships with a
// prose paragraph referencing the marker pair inline — wrapped in a
// backtick-quoted fragment mid-sentence. `indexOf` (the pre-fix
// matcher) would match both of those inline markers and replace the
// content between them with the full injected block, destroying the
// sentence and leaving the backtick unclosed.
//
// Per-test tmpdir so we start from a known clean slate — the shared
// `tmpDir` from beforeAll may already contain CLAUDE.md from earlier
// tests in this describe block.
const bugDir = await fs.mkdtemp(path.join(os.tmpdir(), 'gn-ai-ctx-1041-'));
const bugStorage = path.join(bugDir, '.gitnexus');
await fs.mkdir(bugStorage, { recursive: true });
const inlineProseLine =
'See the `<!-- gitnexus:start --> … <!-- gitnexus:end -->` block in **[AGENTS.md](AGENTS.md)** for the canonical MCP tools, impact analysis rules, and index instructions.';
const originalContent = `# Claude Code Rules\n\nLast reviewed: 2026-04-21\n\n## GitNexus rules\n\n${inlineProseLine}\n`;
const claudeMd = path.join(bugDir, 'CLAUDE.md');
await fs.writeFile(claudeMd, originalContent, 'utf-8');
try {
const stats = { nodes: 50, edges: 100, processes: 5 };
// First run — no section-position markers exist yet, so the
// injector must append a fresh section at end. The inline prose
// must be preserved verbatim; if it disappears or gets altered,
// the bug has recurred.
await generateAIContextFiles(bugDir, bugStorage, 'TestProject', stats);
let contentAfter = await fs.readFile(claudeMd, 'utf-8');
expect(contentAfter, 'inline prose line must survive the first run verbatim').toContain(
inlineProseLine,
);
// Exactly 2 start markers total: 1 inline (in prose) + 1
// section-position (appended by the injector). The pre-fix
// behaviour would have only 1 — the inline pair having been
// consumed as if they were section delimiters.
expect((contentAfter.match(/<!-- gitnexus:start -->/g) || []).length).toBe(2);
expect((contentAfter.match(/<!-- gitnexus:end -->/g) || []).length).toBe(2);
// Second run — the section from run 1 is now at section position,
// so the injector must UPDATE in place (not re-append). Inline
// prose stays preserved; marker counts unchanged.
await generateAIContextFiles(bugDir, bugStorage, 'TestProject', stats);
contentAfter = await fs.readFile(claudeMd, 'utf-8');
expect(contentAfter, 'inline prose line must survive the second run verbatim').toContain(
inlineProseLine,
);
expect((contentAfter.match(/<!-- gitnexus:start -->/g) || []).length).toBe(2);
expect((contentAfter.match(/<!-- gitnexus:end -->/g) || []).length).toBe(2);
} finally {
await fs.rm(bugDir, { recursive: true, force: true });
}
});
it('matches section markers on files with CRLF line endings (#1041 cross-platform)', async () => {
// Locks in the CRLF leg of the section-position matcher. Git on
// Windows may store files with `\r\n` line endings depending on
// `core.autocrlf`; when a section line ends `<!-- gitnexus:start
// -->\r\n`, the byte at `endPos` is `\r` (not `\n`). A `\n`-only
// line-end check would reject the real section, fall through to
// "append", and duplicate the block every run.
const crlfDir = await fs.mkdtemp(path.join(os.tmpdir(), 'gn-ai-ctx-crlf-'));
const crlfStorage = path.join(crlfDir, '.gitnexus');
await fs.mkdir(crlfStorage, { recursive: true });
// Inline reference carries BOTH markers in a backtick-quoted
// fragment — matches the shape of the shipped CLAUDE.md line
// that triggered #1041 so the regression guard is meaningful.
const inlineProseLine =
'See the `<!-- gitnexus:start --> … <!-- gitnexus:end -->` block in **[AGENTS.md](AGENTS.md)** for more.';
const seeded = [
'# Claude Code Rules',
'',
'## GitNexus rules',
'',
inlineProseLine,
'',
'<!-- gitnexus:start -->',
'# GitNexus — Code Intelligence (stale stub)',
'<!-- gitnexus:end -->',
'',
].join('\r\n');
const claudeMd = path.join(crlfDir, 'CLAUDE.md');
await fs.writeFile(claudeMd, seeded, 'utf-8');
try {
const stats = { nodes: 50, edges: 100, processes: 5 };
await generateAIContextFiles(crlfDir, crlfStorage, 'TestProject', stats);
const content = await fs.readFile(claudeMd, 'utf-8');
// Inline prose survives verbatim — no corruption of CRLF bytes.
expect(content).toContain(inlineProseLine);
// Exactly 2 start markers total (1 inline + 1 section-position).
// If CRLF handling broke, the inline marker would be (incorrectly)
// matched as a section start, OR the real section would be
// appended duplicated — either way we'd see !== 2.
expect((content.match(/<!-- gitnexus:start -->/g) || []).length).toBe(2);
expect((content.match(/<!-- gitnexus:end -->/g) || []).length).toBe(2);
// Stale stub content must be gone — proves the section was
// REPLACED (not appended as a duplicate), which requires the
// CRLF-ending markers to have been matched.
expect(content).not.toContain('# GitNexus — Code Intelligence (stale stub)');
} finally {
await fs.rm(crlfDir, { recursive: true, force: true });
}
});
});