Build: Run GitHub Action browser tests on Playwright WebKit
So far, we've been running browser tests on GitHub Actions in Chrome
and Firefox. Regular Safari is not available in GitHub Actions but
Playwright WebKit comes close to a dev version of Safari.
With this change, our GitHub CI & local test runs will invoke tests on
all actively developed browser engines on all PRs.
Also, our GitHub Actions browser tests are now running on Node.js 18.
Detection of the Playwright WebKit browser in support unit tests is done
by checking if the `test_browser` query parameter is set to `"Playwright"`;
this is a `karma-webkit-launcher` feature. Detecting that browser via
user agent as we normally do is hard as the UA on Linux is very similar
to a real Safari one but it actually uses a newer version of the engine.
In addition, we now allow to pass custom browsers when one needs it;
e.g., to run the tests in all three engines on Linux/macOS, run:
```
grunt && BROWSERS=ChromeHeadless,FirefoxHeadless,WebkitHeadless grunt karma:main
```
The `test/middleware-mockserver.js` file used to have the same ESLint
settings applied as other test files that are directly run in tested
browsers. Now it shares settings of other Node.js files.
The file is now also written using modern JS, leveraging ES2018.
Selector: Make selector lists work with `qSA` again
jQuery 3.6.2 started using `CSS.supports( "selector(SELECTOR)" )` before using
`querySelectorAll` on the selector. This was to solve gh-5098 - some selectors,
like `:has()`, now had their parameters parsed in a forgiving way, meaning
that `:has(:fakepseudo)` no longer throws but just returns 0 results, breaking
that jQuery mechanism.
A recent spec change made `CSS.supports( "selector(SELECTOR)" )` always use
non-forgiving parsing, allowing us to use this API for what we've used
`try-catch` before.
To solve the issue on the spec side for older jQuery versions, `:has()`
parameters are no longer using forgiving parsing in the latest spec update
but our new mechanism is more future-proof anyway.
However, the jQuery implementation has a bug - in
`CSS.supports( "selector(SELECTOR)" )`, `SELECTOR` needs to be
a `<complex-selector>` and not a `<complex-selector-list>`. Which means that
selector lists now skip `qSA` and go to the jQuery custom traversal:
```js
CSS.supports("selector(div:valid, span)"); // false
CSS.supports("selector(div:valid)"); // true
CSS.supports("selector(span)"); // true
```
To solve this, this commit wraps the selector list passed to
`CSS.supports( "selector(:is(SELECTOR))" )` with `:is`, making it a single
selector again.
Core:Selector: Move jQuery.contains from the selector to the core module
The `jQuery.contains` method is quite simple in jQuery 4+. On the other side,
it's a dependency of the core `isAttached` util which is not ideal; moving
it from the `selector` the `core` module resolves the issue.
Alex [Thu, 1 Dec 2022 13:23:17 +0000 (15:23 +0200)]
Build: Limit permissions for GitHub workflows
Add explicit permissions section[^1] to workflows. This is a security
best practice because by default workflows run with extended set
of permissions[^2] (except from `on: pull_request` from external forks[^3].
By specifying any permission explicitly all others are set to none. By using
the principle of least privilege the damage a compromised workflow can do
(because of an injection[^4] or compromised third party tool or action) is
restricted. It is recommended to have most strict permissions on the top
level[^5] and grant write permissions on job level[^6] on a case by case
basis.
Selector: Implement the `uniqueSort` chainable method
Some APIs, like `.prevAll()`, return elements in the reversed order, causing
confusing behavior when used with wrapping methods (see gh-5149 for more info)
To provide an easy workaround, this commit implements a chainable `uniqueSort`
method on jQuery objects, an equivalent of `jQuery.uniqueSort`.
Tests: Indicate Firefox 106+ passes the `cssSupportsSelector` test
Firefox 106 adjusted to the spec mandating that `CSS.supports("selector(...)")`
uses non-forgiving parsing which makes it pass the relevant support test.
Re-introduce the `selector-native` similar to the one on the `3.x-stable`
branch. One difference is since the `main` branch inlined Sizzle, some
selector utils can be shared between the main `selector` module and
`selector-native`.
The main `selector` module can be disabled in favor of `selector-native`
via:
grunt custom:-selector
Other changes:
* Tests: Fix Safari detection - Chrome Headless has a different user
agent than Safari and a browser check in selector tests didn't take
that into account.
* Tests: Run selector-native tests in `npm test`
* Selector: Fix querying on document fragments
Selector:Manipulation: Fix DOM manip within template contents
The `<template/>` element `contents` property is a document fragment that may
have a `null` `documentElement`. In Safari 16 this happens in more cases due
to recent spec changes - in particular, even if that document fragment is
explicitly adopted into an outer document. We're testing both of those cases
now.
The crash used to happen in `jQuery.contains`. As it turns out, we don't need
to query the supposed container `documentElement` if it has the
`Node.DOCUMENT_NODE` (9) `nodeType`; we can call `.contains()` directly on
the `document`. That avoids the crash.
Selector: Drop support for legacy pseudos, test custom pseudos
This backports custom pseudos tests from Sizzle; they were missed in original
test backports. Also, the support for legacy custom pseudos has been dropped.
The `jQuery.expr` test cleanup has been wrapped in `try-finally` for cleaner
test isolation in case anything goes wrong.
We've already had `buildFragment` extracted to a separate file long ago.
`domManip` is quite a complex & crucial API and so far it has existed within
the `manipulation.js` module. Extracting it makes the module shorter and easier
to understand.
A few comments / messages in tests have also been updated to not suggest there's
a public `jQuery.domManip` API - it's been private since 3.0.0.
This will resolve the following security issues:
* Path Traversal in Grunt: https://github.com/advisories/GHSA-j383-35pm-c5h4
* Race Condition in Grunt: https://github.com/advisories/GHSA-rm36-94g8-835r
The previous details were showing their age, e.g. mentions about browsers
not supporting ES2015. The story with ES modules is more complex as it's also
about loaders but to keep the README simple, let's just make it more up to date
with typical usage.
Tests: Remove a workaround for a Firefox XML parsing issue
Firefox 96-100 used to report the column number smaller by 2 than it should
in the `parsererror` element generated for invalid XML documents. Since that
version range is unsupported now and it includes no ESR versions, the workaround
can now be dropped.
CSS: Return `undefined` for whitespace-only CSS variable values (#5120)
The spec requires that CSS variable values are trimmed. In browsers that do
this - mainly, Safari, but also Firefox if the value only has leading
whitespace - we currently return undefined; in other browsers, we return
an empty string as the logic to fall back to undefined happens before
trimming.
This commit adds another explicit callback to `undefined` to have it consistent
across browsers.
Also, more explicit comments about behaviors we need to work around in various
browsers have been added.
Selector: Use jQuery `:has` if `CSS.supports(selector(...))` non-compliant
jQuery has followed the following logic for selector handling for ages:
1. Modify the selector to adhere to scoping rules jQuery mandates.
2. Try `qSA` on the modified selector. If it succeeds, use the results.
3. If `qSA` threw an error, run the jQuery custom traversal instead.
It worked fine so far but now CSS has a concept of forgiving selector lists that
some selectors like `:is()` & `:has()` use. That means providing unrecognized
selectors as parameters to `:is()` & `:has()` no longer throws an error, it will
just return no results. That made browsers with native `:has()` support break
selectors using jQuery extensions inside, e.g. `:has(:contains("Item"))`.
Detecting support for selectors can also be done via:
```js
CSS.supports( "selector(SELECTOR_TO_BE_TESTED)" )
```
which returns a boolean. There was a recent spec change requiring this API to
always use non-forgiving parsing:
https://github.com/w3c/csswg-drafts/issues/7280#issuecomment-1143852187
However, no browsers have implemented this change so far.
To solve this, two changes are being made:
1. In browsers supports the new spec change to `CSS.supports( "selector()" )`,
use it before trying `qSA`.
2. Otherwise, add `:has` to the buggy selectors list.
The third parameter of `jQuery.fn.init` - `root` - was just needed to support
`jQuery.sub`. Since this API has been removed in jQuery 1.9.0 and Migrate 3.x
is not filling it in, this parameter is no longer needed.
This parameter has never been documented but it's safer to remove it in a major
update.
Tests: Exclude tests based on compilation flags, not API presence
Introduces a new test API, `includesModule`. The method returns whether
a particular module like "ajax" or "deprecated" is included in the current
jQuery build; it handles the slim build as well. The util was created so that
we don't treat presence of particular APIs to decide whether to run a test as
then if we accidentally remove an API, the tests would still not fail.
Build: Test on Node 17, update Grunt & `karma-*` packages
This adds testing on Node.js 17 in addition to the currently tested 10, 12, 14
and 16 versions.
Also, update Grunt & `karma-*` packages.
Testing in Karma on jsdom is broken in Node 17 at the moment; until we find
a fix, this change disables such testing on Node 17 or newer.
Node smoke tests & promises aplus tests are disabled on Node.js 10 as they
depend on jsdom and the latest jsdom version doesn't run properly on Node 10.
`jQuery.fx.interval` has been deprecated since jQuery 3.0.0 but it has been
still used in jQuery code until this change. This commit removes the definition
and explicitly uses the `13` number in its place.
CSS: Skip falsy values in `addClass( array )`, compress code
This change makes jQuery skip falsy values in `addClass( array )`
& `removeClass( array )` instead of stopping iteration when the first falsy
value is detected. This makes code like:
```js
elem.addClass( [ "a", "", "b" ] );
```
add both the `a` & `b` classes.
The code was also optimized for size a bit so it doesn't increase the
minified gzipped size.
Bruno PIERRE [Mon, 24 Jan 2022 17:55:16 +0000 (18:55 +0100)]
Core: Don't rely on splice being present on input
Without this fix calling `jQuery.uniqueSort` on an array-like can result in:
TypeError: results.splice is not a function
at Function.jQuery.uniqueSort (https://code.jquery.com/jquery-git.js:664:12)
at jQuery.fn.init.find (https://code.jquery.com/jquery-git.js:2394:27)
at gocusihafe.js:3:4
PR gh-4993 changed a few too many issue references to `trac-NUMBER` ones. This
change fixes them. It also fixes a typo in one Trac issue number in selector
tests.
Neither of the removed links is crucial; one of them refers to a site that has
since started being malicious; while the Web Archive links remain safe, some
scanners warn about such links. Removing them is the safest thing to do.
Docs: Replace `#NUMBER` Trac issue references with `trac-NUMBER`
The GitHub UI treats `#NUMBER` as referring to its own issues which is confusing
when in jQuery source it's usually referring to the old deprecated Trac instance
at https://bugs.jquery.com. This change replaces all such Trac references with
`trac-NUMBER`.
A few of the references came with the Sizzle integration and referred to the
Sizzle GitHub bug tracker. Those have been replaced with full links instead.
A new entry describing issue reference conventions has been added to README.
Richard Gibson [Mon, 3 Jan 2022 12:28:49 +0000 (07:28 -0500)]
CSS: Justify use of rtrim on CSS property values
CSS does not acknowledge carriage return or form feed characters
as whitespace but it does replace them with whitespace, making it
acceptable to use `rtrim`.
TestSwarm is now proxied via Cloudflare which cuts out headers relevant for
ETag tests, failing them. We're still running those tests in Karma on Chrome
& Firefox (including Firefox ESR).
Tests: Allow statusText to be "success" in AJAX tests
In HTTP/2, status message is not supported and whatever is reported as
statusText differs between browsers. In Chrome & Safari it's "success", in
Firefox & IE it's "OK". So far "success" wasn't allowed. This made the tests
pass locally if you're running an HTTP/1.1 server but on TestSwarm which is
now proxied via an HTTP/2-equipped Cloudflare, the relevant test started failing
in Chrome & Safari.
Docs: Update the URL to the latest jQuery build in CONTRIBUTING.md
It used to say https://code.jquery.com/jquery.js but that's a frozen URL
to jQuery 1.11.1. Let's switch that to the URL to the Git build, i.e.
https://releases.jquery.com/git/jquery-git.js.
Attributes: Don't stringify attributes in the setter
Stringifying attributes in the setter was needed for IE <=9 but it breaks
trusted types enforcement when setting a script `src` attribute.
Note that this doesn't mean script execution works. Since jQuery disables all
scripts by changing their type and then executes them by creating fresh script
tags with proper `src` & possibly other attributes, this unwraps any trusted
`src` wrappers, making the script not execute under strict CSP settings.
We might try to fix it in the future in a separate change.
Tests: Make Karma browser timeout larger than the QUnit one
Since the default Karma browser no activity timeout was lower than the QUnit
timeout, a single timing out test was interrupting the whole test run of
a browser.
The QUnit timeout is set to 1 minute so I set the Karma one to 2 minutes.
Build: Update ESLint & eslint-plugin-import, fixing the build
Latest `main` started failing the build after some transitive dependencies
got updated, incorrectly recognizing some files with default exports as unused.
Since the new ESLint no longer supports Node 10 which we have to build on due
to use in our CI, skip ESLint in Node 10.
Docs: Remove the CLA checkbox in the pull request template
The EasyCLA status check is required so this won't get missed. The old JSF CLA
is dead, the provided link doesn't return meaningful information. There's no
good replacement link for the old CLA; PR authors are just supposed to sign the
new CLA by clicking on a link posted by the EasyCLA bot when they submit their
first PR since EasyCLA was enabled for the repo.
This ensures HTML wrapped in TrustedHTML can be used as an input to jQuery
manipulation methods in a way that doesn't violate the
`require-trusted-types-for` Content Security Policy directive.
This commit builds on previous work needed for trusted types support, including
gh-4642 and gh-4724.
One restriction is that while any TrustedHTML wrapper should work as input
for jQuery methods like `.html()` or `.append()`, for passing directly to the
`jQuery` factory the string must start with `<` and end with `>`; no trailing
or leading whitespaces are allowed. This is necessary as we cannot parse out
a part of the input for further construction; that would violate the CSP rule -
and that's what's done to HTML input not matching these constraints.
No trusted types API is used explicitly in source; the majority of the work is
ensuring we don't pass the input converted to string to APIs that would
eventually assign it to `innerHTML`. This extra cautiousness is caused by the
API being Blink-only, at least for now.
The ban on passing strings to `innerHTML` means support tests relying on such
assignments are impossible. We don't currently have such tests on the `main`
branch but we used to have many of them in the 3.x & older lines. If there's
a need to re-add such a test, we'll need an escape hatch to skip them for apps
needing CSP-enforced TrustedHTML.
See https://web.dev/trusted-types/ for more information about TrustedHTML.
Manipulation: Don't remove HTML comments from scripts
When evaluating scripts, jQuery strips out the possible wrapping HTML comment
and a CDATA section. However, all supported browsers are already doing that
when loading JS via appending a script tag to the DOM which is how we've been
doing `jQuery.globalEval` since jQuery 3.0.0. jQuery logic was imperfect, e.g.
it just stripped the `<!--` and `-->` markers, respectively at the beginning or
the end of the script contents. However, browsers are also stripping everything
following those markers in the same line, treating them as single-line comments
delimiters; this is now also mandated by ECMAScript 2015 in Annex B. Instead
of fixing the jQuery logic, just let the browser do its thing.
We also used to strip CDATA sections. However, this shouldn't be needed as in
XML documents they're already not visible when inspecting element contents and
in HTML documents they have no meaning. We've preserved that behavior for
backwards compatibility in 3.x but we're removing it for 4.0.
Event: Don't break focus triggering after `.on(focus).off(focus)`
The `_default` function in the special event settings for focus/blur has
always returned `true` since gh-4813 as the event was already being fired
from `leverageNative`. However, that only works if there's an active handler
on that element; this made a quick consecutive call:
make subsequent `.trigger( "focus" )` calls to not do any triggering.
The solution, already used in a similar `_default` method for the `click` event,
is to check for the `dataPriv` entry on the element for the focus event
(similarly for blur).
Tests: Strip untypical callback parameter characters from mock.php
Only allow alphanumeric characters & underscores for callback parameters.
The change is done both for the PHP server as well as the Node.js-based version.
This is only test code so we're not fixing any security issue but it happens
often enough that the whole jQuery repository directory structure is deployed
onto the server with PHP enabled that it makes is easy to introduce security
issues if this cleanup is not done.
Tests: Make more tests run natively in Chrome & Firefox
Chrome & Firefox now support complex `:not()` selectors so those test can run
in them even without custom jQuery selector code. In the past, it was only
possible in Safari, now we only need to exclude IE.
Build: Take core-js-bundle from the external directory as well
That package was missed in gh-4865 as it only broke browsers needing the
polyfill which is just IE at the moment. Thus, it broke Core tests in IE only.
In gh-4466, we removed the `external` directory in favor of loading some files
directly from `node_modules`. This works fine locally but when deploying code
for tests, this makes it impossible to not deploy `node_modules` as well. To
avoid the issue, this change restores usage of the `external` directory.
One change is that we no longer commit this directory to the repository, its
only purpose is to have clear isolation from `node_modules`.
PR gh-2588 made jQuery stop auto-execute cross-domain scripts unless
`dataType: "script"` was explicitly provided; this change landed in jQuery
3.0.0. This change extends that logic same-domain scripts as well.
After this change, to request a script under a provided URL to be evaluated,
you need to provide `dataType: "script` in `jQuery.ajax` options or to use
`jQuery.getScript`.
Tests: Fix tests for not auto-executing scripts without dataType
Two issues are fixed in testing for responses with a script Content-Type not
getting auto-executed unless an explicit `dataType: "script"` is provided:
* the test is now using a correct "text/javascript" Content-Type; it was using
"text/html" until now which doesn't really check if the fix works
* the Node.js based version of the tests didn't account for an empty `header`
query string parameter
Timmy Willison [Mon, 11 Jan 2021 16:56:08 +0000 (11:56 -0500)]
Dimensions: Add offset prop fallback to FF for unreliable TR dimensions
Firefox incorrectly (or perhaps correctly) includes table borders in computed
dimensions, but they are the only one. Workaround this by testing for it and
falling back to offset properties
due to their synchronous nature everywhere outside of IE the hack added in
gh-4279 to leverage native events causes the native `.focus()` method to be
called last for the initial element, making it steal the focus back. Since
the native method is already being called in `leverageNative`, we can skip that
final call.
This aligns with changes to the `_default` method for the `click` event that
were added when `leverageNative` was introduced there.
A side effect of this change is that now `focusin` will only propagate to the
document for the last focused element. This is a change in behavior but it also
aligns us better with how this works with native methods.
Event: Don't crash if an element is removed on blur
In Chrome, if an element having a `focusout` handler is blurred by
clicking outside of it, it invokes the handler synchronously. If
that handler calls `.remove()` on the element, the data is cleared,
leaving private data undefined. We're reading a property from that
data so we need to guard against this.
Build: Explicitly exclude the queue module from the slim build
The queue module is not present in the slim build as it depends on deferred
and our Gruntfile specifies excluding deferred should also exclude queue:
https://github.com/jquery/jquery/blob/3.5.1/Gruntfile.js#L66
This commit makes this exclusion explicit so that the queue module never
accidentally gets re-included in the slim build if it stopped importing from
the deferred module directly.
Core: Drop support for Edge Legacy (i.e. non-Chromium Microsoft Edge)
Drop support for Edge Legacy: the non-Chromium, EdgeHTML-based Microsoft
Edge version. Also, restrict some workarounds that were applied
unconditionally in all browsers to run only in IE now. This slightly
increases the size but reduces the performance burden on modern browsers
that don't need the workarounds.
Also, clean up some comments & remove some obsolete workarounds.
Tests: Recognize callbacks with dots in the Node.js mock server
This aligns the Node.js server with the previous PHP one in sending `mock.php`
as a callback if there's no `callback` parameter in the query string which is
triggered by a recently added test. This prevents the request crashing on that
Node.js server and printing a JS error:
```
TypeError: Cannot read property '1' of null
```
Build: Make the import/no-unused-modules ESLint rule work in WebStorm
When run via WebStorm, the root path against which paths in the config of the
`import/no-unused-modules` ESLint rule are resolved is the path where the ESLint
config file that defines the rule lies, i.e. `src`. When run via the command
line, it's usually the root folder of the jQuery repository. This pattern
intends to catch both.
Note that we cannot specify two patterns here:
```js
[ "src/*.js", "*.js" ]
```
as they're analyzed individually and the rule crashes if a pattern cannot be
matched.
Ajax: Make responseJSON work for erroneous same-domain JSONP requests
Don't use a script tag for JSONP requests unless for cross-domain requests
or if scriptAttrs are provided. This makes the `responseJSON` property available
in JSONP error callbacks.
This fixes a regression from jQuery 3.5.0 introduced in gh-4379 which made
erroneous script responses to not be executed to follow native behavior.
The 3.x-stable branch doesn't need this fix as it doesn't use script tags for
regular async requests.
Dallas Fraser [Tue, 25 Aug 2020 19:41:06 +0000 (15:41 -0400)]
Ajax: Execute JSONP error script responses
Issue gh-4379 was meant to be a bug fix but the JSONP case is a bit special:
under the hood it's a script but it simulates JSON responses in an environment
without a CORS setup and sending JSON payloads on error responses is quite
typical there.
This commit makes JSONP error responses still execute the payload. The regular
script error responses continue to be skipped.
Ajax: Avoid CSP errors in the script transport for async requests
Until now, the AJAX script transport only used a script tag to load scripts
for cross-domain requests or ones with `scriptAttrs` set. This commit makes
it also used for all async requests to avoid CSP errors arising from usage
of inline scripts. This also makes `jQuery.getScript` not trigger CSP errors
as it uses the AJAX script transport under the hood.
For sync requests such a change is impossible and that's what `jQuery._evalUrl`
uses. Fixing that is tracked in gh-1895.
The commit also makes other type of requests using the script tag version of the
script transport set its type to "GET", namely async scripts & ones with
`scriptAttrs` set in addition to the existing cross-domain ones.
Previously, `jQuery.ajax` with `dataType: 'json'` with a provided callback was
automatically converted to a jsonp request unless one also specified
`jsonp: false`. Today the preferred way of interacting with a cross-domain
backend is CORS which works in all browsers jQuery 4 will support.
Auto-promoting JSON requests to JSONP ones introduces a security issue as the
developer may be unaware they're not just downloading data but executing code
from a remote domain.
This commit disables the auto-promoting logic.
BREAKING CHANGE: to trigger a JSONP request, it's now required to specify
`dataType: "jsonp"`; previously some requests with `dataType: "json"` were
auto-promoted to JSONP.
This also resolves a security warning from GitHub about a vulnerable `request`
version - the new `testswarm` package version depends on a fixed `request`.
Manipulation: Avoid concatenating strings in buildFragment
Concatenating HTML strings in buildFragment is a possible security risk as it
creates an opportunity of escaping the concatenated wrapper. It also makes it
impossible to support secure HTML wrappers like
[trusted types](https://web.dev/trusted-types/). It's safer to create wrapper
elements using `document.createElement` & `appendChild`.
The previous way was needed in jQuery <4 because IE <10 doesn't accept table
parts set via `innerHTML`, even if the element which contents are set is
a proper table element, e.g.:
```js
tr.innerHTML = "<td></td>";
```
The whole structure needs to be passed in one HTML string. jQuery 4 drops
support for IE <11 so this is no longer an issue; in older version we'd have
to duplicate the code paths.
IE <10 needed to have `<option>` elements wrapped in
`<select multiple="multiple">` but we no longer need that on master which
makes the `document.createElement` way shorter as we don't have to call
`setAttribute`.
All these improvements, apart from making logic more secure, decrease the
gzipped size by 58 bytes.
Docs: Update Frequently Reported Issues in the GitHub issue template
The issue about selectors with '#' being broken is old and no longer
frequently reported so this commit removes it from the list. On the other
hand, we're now getting lots of reports about the security fix in jQuery 3.5.0
that was also a breaking change: gh-4642. This one is now mentioned in the
list.
projects / jquery.git / log