fix(droid): harden exec passthrough and add integration coverage

This commit is contained in:
Tam Nhu Tran committed 2026-02-26 23:28:36 +07:00
1 parent 5963ba973e
commit ce44d2ece0
6 files changed
+441 -42

No files matched your search

@@ -1,5 +1,8 @@
import { describe, expect, it } from 'bun:test';
import { routeDroidCommandArgs } from '../../../src/targets/droid-command-router';
import {
DroidCommandRouterError,
routeDroidCommandArgs,
} from '../../../src/targets/droid-command-router';
describe('droid-command-router', () => {
it('keeps interactive mode for plain profile launches', () => {
@@ -8,6 +11,7 @@ describe('droid-command-router', () => {
expect(route.mode).toBe('interactive');
expect(route.argsForDroid).toEqual([]);
expect(route.autoPrependedExec).toBe(false);
expect(route.duplicateReasoningDisplays).toEqual([]);
});
it('keeps explicit droid subcommands untouched', () => {
@@ -17,6 +21,7 @@ describe('droid-command-router', () => {
expect(route.command).toBe('mcp');
expect(route.argsForDroid).toEqual(['mcp', '--help']);
expect(route.autoPrependedExec).toBe(false);
expect(route.duplicateReasoningDisplays).toEqual([]);
});
it('auto-prepends exec for exec-only flags provided after profile', () => {
@@ -26,6 +31,7 @@ describe('droid-command-router', () => {
expect(route.command).toBe('exec');
expect(route.argsForDroid).toEqual(['exec', '--skip-permissions-unsafe']);
expect(route.autoPrependedExec).toBe(true);
expect(route.duplicateReasoningDisplays).toEqual([]);
});
it('does not auto-prepend exec for root help flag', () => {
@@ -34,6 +40,7 @@ describe('droid-command-router', () => {
expect(route.mode).toBe('interactive');
expect(route.argsForDroid).toEqual(['--help']);
expect(route.autoPrependedExec).toBe(false);
expect(route.duplicateReasoningDisplays).toEqual([]);
});
it('normalizes --effort alias to --reasoning-effort for explicit exec', () => {
@@ -41,12 +48,8 @@ describe('droid-command-router', () => {
expect(route.mode).toBe('command');
expect(route.command).toBe('exec');
expect(route.argsForDroid).toEqual([
'exec',
'--reasoning-effort',
'xhigh',
'fix test flake',
]);
expect(route.argsForDroid).toEqual(['exec', '--reasoning-effort', 'xhigh', 'fix test flake']);
expect(route.reasoningSourceDisplay).toBe('--effort xhigh');
});
it('normalizes --thinking alias when exec is auto-prepended', () => {
@@ -58,10 +61,12 @@ describe('droid-command-router', () => {
'exec',
'--auto',
'high',
'--reasoning-effort=medium',
'--reasoning-effort',
'medium',
'summarize logs',
]);
expect(route.autoPrependedExec).toBe(true);
expect(route.reasoningSourceDisplay).toBe('--thinking=medium');
});
it('still auto-prepends exec when --effort appears before exec-only flags', () => {
@@ -82,5 +87,80 @@ describe('droid-command-router', () => {
'fix flaky test',
]);
expect(route.autoPrependedExec).toBe(true);
expect(route.reasoningSourceDisplay).toBe('--effort xhigh');
});
it('auto-prepends exec for non-ambiguous short exec flags', () => {
const route = routeDroidCommandArgs(['-m', 'custom:gpt-5.3-codex', 'fix flaky test']);
expect(route.mode).toBe('command');
expect(route.command).toBe('exec');
expect(route.argsForDroid).toEqual(['exec', '-m', 'custom:gpt-5.3-codex', 'fix flaky test']);
expect(route.autoPrependedExec).toBe(true);
});
it('routes -r to exec when value matches reasoning effort level', () => {
const route = routeDroidCommandArgs(['-r', 'high', 'summarize logs']);
expect(route.mode).toBe('command');
expect(route.command).toBe('exec');
expect(route.argsForDroid).toEqual(['exec', '--reasoning-effort', 'high', 'summarize logs']);
expect(route.autoPrependedExec).toBe(true);
});
it('keeps interactive mode for ambiguous -r resume-style usage', () => {
const route = routeDroidCommandArgs(['-r', 'session-1234']);
expect(route.mode).toBe('interactive');
expect(route.argsForDroid).toEqual(['-r', 'session-1234']);
expect(route.autoPrependedExec).toBe(false);
});
it('dedupes mixed reasoning flags with first occurrence precedence', () => {
const route = routeDroidCommandArgs([
'exec',
'--reasoning-effort',
'high',
'--thinking',
'low',
'--effort=xhigh',
'summarize logs',
]);
expect(route.mode).toBe('command');
expect(route.command).toBe('exec');
expect(route.argsForDroid).toEqual(['exec', '--reasoning-effort', 'high', 'summarize logs']);
expect(route.reasoningSourceDisplay).toBe('--reasoning-effort high');
expect(route.duplicateReasoningDisplays).toEqual(['--thinking low', '--effort=xhigh']);
});
it('throws for missing reasoning value in command mode (alias)', () => {
expect(() => routeDroidCommandArgs(['exec', '--effort'])).toThrow(DroidCommandRouterError);
});
it('throws for missing reasoning value in command mode (native)', () => {
expect(() => routeDroidCommandArgs(['exec', '--reasoning-effort'])).toThrow(
DroidCommandRouterError
);
});
it('records malformed duplicate reasoning flags when first value is already selected', () => {
const route = routeDroidCommandArgs([
'exec',
'--thinking',
'medium',
'--reasoning-effort',
'--skip-permissions-unsafe',
'summarize logs',
]);
expect(route.argsForDroid).toEqual([
'exec',
'--reasoning-effort',
'medium',
'--skip-permissions-unsafe',
'summarize logs',
]);
expect(route.duplicateReasoningDisplays).toEqual(['--reasoning-effort <missing-value>']);
});
});
@@ -0,0 +1,173 @@
import { afterEach, beforeEach, describe, expect, it } from 'bun:test';
import { spawnSync } from 'child_process';
import * as fs from 'fs';
import * as os from 'os';
import * as path from 'path';
interface RunResult {
status: number | null;
stdout: string;
stderr: string;
}
function runCcs(args: string[], env: NodeJS.ProcessEnv): RunResult {
const ccsEntry = path.join(process.cwd(), 'src', 'ccs.ts');
const result = spawnSync(process.execPath, [ccsEntry, ...args], {
encoding: 'utf8',
env,
timeout: 20000,
});
return {
status: result.status,
stdout: result.stdout || '',
stderr: result.stderr || '',
};
}
describe('droid command routing integration', () => {
let tmpHome: string;
let ccsDir: string;
let settingsPath: string;
let configPath: string;
let fakeDroidPath: string;
let droidArgsLogPath: string;
let baseEnv: NodeJS.ProcessEnv;
beforeEach(() => {
if (process.platform === 'win32') {
return;
}
tmpHome = fs.mkdtempSync(path.join(os.tmpdir(), 'ccs-droid-route-it-'));
ccsDir = path.join(tmpHome, '.ccs');
fs.mkdirSync(ccsDir, { recursive: true });
settingsPath = path.join(ccsDir, 'myglm.settings.json');
configPath = path.join(ccsDir, 'config.json');
fakeDroidPath = path.join(tmpHome, 'fake-droid.js');
droidArgsLogPath = path.join(tmpHome, 'droid-args.json');
fs.writeFileSync(
settingsPath,
JSON.stringify(
{
env: {
ANTHROPIC_BASE_URL: 'https://example.invalid/anthropic',
ANTHROPIC_AUTH_TOKEN: 'test-token',
ANTHROPIC_MODEL: 'gpt-5.3-codex',
CCS_DROID_PROVIDER: 'openai',
},
},
null,
2
)
);
fs.writeFileSync(
configPath,
JSON.stringify(
{
profiles: {
myglm: settingsPath,
},
},
null,
2
)
);
fs.writeFileSync(
fakeDroidPath,
`#!/usr/bin/env node
const fs = require('fs');
const out = process.env.CCS_TEST_DROID_ARGS_OUT;
if (!out) process.exit(2);
fs.writeFileSync(out, JSON.stringify(process.argv.slice(2)));
process.exit(0);
`,
{ encoding: 'utf8', mode: 0o755 }
);
fs.chmodSync(fakeDroidPath, 0o755);
baseEnv = {
...process.env,
CI: '1',
NO_COLOR: '1',
CCS_HOME: tmpHome,
CCS_DROID_PATH: fakeDroidPath,
CCS_TEST_DROID_ARGS_OUT: droidArgsLogPath,
};
});
afterEach(() => {
if (process.platform === 'win32') {
return;
}
fs.rmSync(tmpHome, { recursive: true, force: true });
});
it('auto-routes exec-only long flags to droid exec from main ccs flow', () => {
if (process.platform === 'win32') return;
const result = runCcs(
['myglm', '--target', 'droid', '--skip-permissions-unsafe', 'fix failing tests'],
baseEnv
);
expect(result.status).toBe(0);
const routedArgs = JSON.parse(fs.readFileSync(droidArgsLogPath, 'utf8')) as string[];
expect(routedArgs).toEqual(['exec', '--skip-permissions-unsafe', 'fix failing tests']);
});
it('auto-routes non-ambiguous short exec flags', () => {
if (process.platform === 'win32') return;
const result = runCcs(
['myglm', '--target', 'droid', '-m', 'custom:gpt-5.3-codex', 'fix failing tests'],
baseEnv
);
expect(result.status).toBe(0);
const routedArgs = JSON.parse(fs.readFileSync(droidArgsLogPath, 'utf8')) as string[];
expect(routedArgs).toEqual(['exec', '-m', 'custom:gpt-5.3-codex', 'fix failing tests']);
});
it('dedupes reasoning flags with first occurrence precedence in exec mode', () => {
if (process.platform === 'win32') return;
const result = runCcs(
[
'myglm',
'--target',
'droid',
'exec',
'--reasoning-effort',
'high',
'--thinking',
'low',
'summarize logs',
],
baseEnv
);
expect(result.status).toBe(0);
expect(result.stderr).toContain('Multiple reasoning flags detected');
const routedArgs = JSON.parse(fs.readFileSync(droidArgsLogPath, 'utf8')) as string[];
expect(routedArgs).toEqual(['exec', '--reasoning-effort', 'high', 'summarize logs']);
});
it('fails fast for malformed reasoning alias in command mode', () => {
if (process.platform === 'win32') return;
const result = runCcs(['myglm', '--target', 'droid', 'exec', '--effort'], baseEnv);
expect(result.status).toBe(1);
expect(result.stderr).toContain('--effort requires a value');
expect(fs.existsSync(droidArgsLogPath)).toBe(true);
const probeArgs = JSON.parse(fs.readFileSync(droidArgsLogPath, 'utf8')) as string[];
// Droid binary is still invoked once for version preflight (`--version`) before routing.
expect(probeArgs).toEqual(['--version']);
});
});