Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

feat: Add --chromium-pref flag #2912

Open
wants to merge 17 commits into
base: master
Choose a base branch
from

Conversation

aklinker1
Copy link
Contributor

@aklinker1 aklinker1 commented Oct 15, 2023

Overview

This closes #2899.

Add a new parameter, --chromium-pref, which lets users configure some profile settings when running the web-ext run command. For default preferences, I set extensions.ui.developer_mode=true so you don't have to toggle the developer mode switch every time you run an extension on chromium.

const DEFAULT_PREFS = {
'extensions.ui.developer_mode': true,
};

Manual Testing

Here's the commands I ran to do manual testing:

# Default, dev mode should be enabled at chrome://extensions
node .\bin\web-ext.js run -s .\tests\fixtures\minimal-web-ext\ -t chromium
# Disable dev mode, change download dir (chrome://settings/downloads)
node .\bin\web-ext.js run -s .\tests\fixtures\minimal-web-ext\ -t chromium --chromium-pref extensions.ui.developer_mode=false --chromium-pref download.default_directory="C:\Users\Aaron\Downloads\test"

@@ -66,6 +66,9 @@ describe('util/extension-runners/chromium', async () => {
'--password-store=basic',
'--use-mock-keychain',
'--force-fieldtrials=*BackgroundTracing/default/',
'--disable-hang-monitor',
'--disable-prompt-on-repost',
'--disable-domain-reliability',
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

These changes were required after upgrading chrome-launcher, similar to #2825 (comment)

@aklinker1 aklinker1 changed the title Chromium Preferences feat: Add --chromium-pref flag Oct 15, 2023
@codecov
Copy link

codecov bot commented Oct 15, 2023

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.31%. Comparing base (17d1c4b) to head (b1779d9).

Current head b1779d9 differs from pull request most recent head 61e8a86

Please upload reports for the commit 61e8a86 to get more accurate results.

Additional details and impacted files
@@             Coverage Diff             @@
##           master    #2912       +/-   ##
===========================================
+ Coverage        0   99.31%   +99.31%     
===========================================
  Files           0       32       +32     
  Lines           0     1759     +1759     
===========================================
+ Hits            0     1747     +1747     
- Misses          0       12       +12     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

@aklinker1 aklinker1 marked this pull request as ready for review October 15, 2023 18:15
@aklinker1
Copy link
Contributor Author

Rebased and fixed merge conflicts

@aklinker1
Copy link
Contributor Author

Rebased and fixed conflicts

@VersoriumX
Copy link

No conflicts present with Branch

Copy link

@VersoriumX VersoriumX left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

add

@pftom
Copy link

pftom commented May 30, 2024

@aklinker1 why not merge this pr? I still encounter the content script source map issues

@aklinker1
Copy link
Contributor Author

aklinker1 commented May 30, 2024

Only maintainers can merge PRs, and this hasn't been approved by one.

Copy link

@VersoriumX VersoriumX left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks

@aklinker1
Copy link
Contributor Author

@VersoriumX I appreciate your approvals, but since you're not a maintainer, you approving this more than once doesn't really help

@Rob--W
Copy link
Member

Rob--W commented Jun 6, 2024

Sorry for the delay here - I've put this on my list of tasks to review. We are supportive of the requested capability here.

@aklinker1
Copy link
Contributor Author

@Rob--W no worries! Thanks for the update

@Rob--W Rob--W self-requested a review June 20, 2024 12:15
Copy link
Member

@Rob--W Rob--W left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This PR contains unrelated code (maybe partially due to a merge conflict).

Since you need to fix up the PR, could you split this PR in two parts:

  • chore: Update chrome-launcher (this can be merged quickly)
  • feat: Add --chromium-pref flag (can be reviewed and merged separately).

If it makes it easier, you could create a new branch branched off the main branch that updates chrome-launcher (and create a PR for that), and then rebase the current branch on top of that. Once the chrome-launcher PR is merged into main, this PR can be rebased on top of the main branch.

package.json Outdated
@@ -81,6 +81,7 @@
"open": "9.1.0",
"parse-json": "7.1.1",
"promise-toolbox": "0.21.0",
"set-value": "4.1.0",
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Let's not add the set-value module. It appears to be unmaintained (no responses from the project maintainer on several comments from the past year), and upon inspection I also see the lack of safeguards against pollution of built-in prototype members.

src/program.js Outdated
requiresArg: true,
type: 'array',
coerce: (arg) =>
arg != null ? coerceCLICustomPreference(arg) : undefined,
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

coerceCLICustomPreference is Firefox-specific, and includes Firefox-specific logic such as preferences that cannot be overridden. You should use a custom one for Chromium.

* "extensions.ui.developer_mode".
*/
getPrefs() {
return Object.entries({
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

To avoid the dependency of set-value and avoid overwriting built-in instance members, let's roll (y)our own. The expected input is a list of dotted paths + values, currently stored as a flat object (but originally a string of dotted.key=value pairs). The expected output is an object with nested objects.

Here is an example (not tested):

function mapToObject(map) {
  return Object.fromEntries(
    Array.from(
      map,
      ([k, v]) => [k, v instanceof Map ? mapToObject(v) : v]
    )
  );
}

// Logic for getPrefs(), given prefs = DEFAUL_PREFS etc.
const prefsMap = new Map();
for (let [key, value] of prefs) {
  let submap = prefsMap;
  const props = key.split(".");
  const lastProp = props.pop();
  for (let prop of props) {
    if (!submap.has(prop)) {
      submap.set(prop, new Map());
    }
    submap = submap.get(prop);
    if (!(submap instanceof Map)) {
      throw new Error(`Cannot set ${key} because a value already exists at ${prop}`);
    }
  }
  // TODO: Consider log.warn() if submap.has(lastProp) before overwriting.
  submap.set(lastProp, value);
}

return mapToObject(prefsMap);

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm going to base the function off this implementation. It's very compact and doesn't require using maps: https://youmightnotneed.com/lodash/#set

Will add some basic tests too.

Why do you want to avoid overwriting existing values? None of the default prefs I set in this PR are necessary for running the extension, they're just nice defaults that someone could turn off if they wanted. I also can't think of a use-case in the future where that would be useful.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm going to base the function off this implementation. It's very compact and doesn't require using maps: https://youmightnotneed.com/lodash/#set

This snippet you've selected and lodash (_.set) are both vulnerable to a Prototype pollution vulnerability. The set-value library that you used before fixed that in response to a report (GHSA-4g88-fppr-53pp). Despite the fix, it is still possible to overwrite properties on inherit built-in prototypes. For example:

var setValue = require('set-value');
setValue({}, 'hasOwnProperty.call', 1);
// Expected: true. Actual: TypeError: Object.prototype.hasOwnProperty.call is not a function
console.log(Object.prototype.hasOwnProperty.call(Math, "min"));

Why do you want to avoid overwriting existing values?

Are you referring to "TODO: Consider log.warn() if submap.has(lastProp) before overwriting."? The purpose of that is to let the user know that a built-in value has been overwritten. Not that big of a deal. I'm also fine with omitting it. Overwriting the value when there is a Map at that point may be an error though, consider the following:

download.download_path=/path/to/somewhere
download=1

It is not meaningful to set "download" to 1, and a warning could be nice. Not required though.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I see now. I misunderstood lots of things about the original comment lol. But now that I've implemented and tested your original suggestion, we're on the same page.

Only comment is about throwing the error. See these test cases:

https://github.com/aklinker1/web-ext/blob/eefc4adeea0b3930adcdf69b7b92a751e81eddc7/tests/unit/test-util/test.expand-prefs.js#L39-L61

We throw an error on the first, but not the second. I think we should make both behave the same way, either throw an error or allow the later value to override the earlier. Thoughts?

@aklinker1
Copy link
Contributor Author

@Rob--W Thanks for the review! I'll update this branch and make the suggested changes once chrome-launcher is updated. I created a PR: #3200

@aklinker1 aklinker1 requested a review from Rob--W July 20, 2024 21:06
getPrefs() {
return expandPrefs({
...DEFAULT_PREFS,
...(this.params.customChromiumPrefs || {}),
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Spread syntax with ...undefined and ...null works, you don't need to fall back to || {}.

Suggested change
...(this.params.customChromiumPrefs || {}),
...this.params.customChromiumPrefs,


// Create an alias for --chromium-pref since it has been transformed into an
// object containing one or more preferences.
const customChromiumPrefs = { ...chromiumPref };
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I see that you basically copied the previous few lines. The logic there did not completely make sense, I've sent a PR to fix that up: #3218

Let's simplify to the following:

Suggested change
const customChromiumPrefs = { ...chromiumPref };
const customChromiumPrefs = chromiumPref;

};
const actual = expandPrefs(input);

assert.notEqual(Object.prototype.hasOwnProperty.call, call);
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Instead of notEqual, save the original value before modifying the test, then do a strict comparison. A unit test that tests for equality is stronger than one that tests for inequality.

assert.deepEqual(actual, expected);
});

it("doesn't pollute the object prototype", () => {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The test case here is a specific one that failed the original library that you proposed. Let's also prepend a test for a more obvious test case:

const input = {
  '__proto__.x': 1
};
// Using JSON.parse because an object literal with __proto__
// would be mistaken for an object with a different prototype.
const expected = JSON.parse('{"__proto__":{"x":1}}');

Besides the obvious comparison like above, you should also add a test Object.getPrototypeOf(actual) and confirm that it is Object.prototype.

And let's also have the same test, but with the following:

const input = JSON.parse('{"__proto__":[]}');
const expected = JSON.parse('{"__proto__":[]}');

params: {
customChromiumPrefs: {
'download.default_directory': '/some/directory',
'extensions.ui.developer_mode': false,
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Let's split the test in two: that a custom pref can be passed, and that a default pref can be overridden.

We want to make sure that custom prefs are merged with the default prefs, and not changed. The current implementation is correct, but the test as constructed would still pass if the implementation failed to merge, which is a sign of an incomplete test.

import { UsageError } from '../errors.js';

export function coerceCLICustomChromiumPreference(cliPrefs) {
const customPrefs = {};
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
const customPrefs = {};
const customPrefs = new Map();

Suggesting a few changes to avoid prototype pollution.

value = value === 'true';
}

customPrefs[`${key}`] = value;
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
customPrefs[`${key}`] = value;
customPrefs.set(`${key}`, value);

customPrefs[`${key}`] = value;
}

return customPrefs;
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
return customPrefs;
return Object.fromEntries(customPrefs);

const key = prefsAry[0];
let value = prefsAry.slice(1).join('=');

if (/[^\w_.]/.test(key)) {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is there any point in this check or can it be dropped?

\w means [A-Za-z0-9_], so what your code here does is rejecting anything other than alphanumeric characters and period.

It is stricter than the actual values permitted by Chromium's Preferences keys, which is quite permissive actually. For example it can contain URLs (including : characters which are rejected by your current code), e.g. as seen at:

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Pass in custom chromium prefs (enable developer mode by default)
4 participants