Explicitly stubbing i18n in 48 different files
is mostly busy work at this point, and it doesn't
provide much signal, since often it's invoked
only to satisfy transitive dependencies.
Let's say you have module hello.js like so:
// hello.js
const hello_world = i18n.t('Hello world');
exports.get_greeting = () => hello_world;
And then two modules like this:
// apple.js
const hello = require('hello');
exports.foo = () => {
show_greeting(hello.get_greeting());
};
// banana.js
const hello = require('hello');
exports.foo = () => {
display_greeting(hello.get_greeting());
};
The test for apple.js could look like this,
and it won't crash due to the stub:
set_global('i18n', {t: () => {}});
zrequire('hello');
zrequire('apple');
Now let's say your write this broken version
of a test for banana.js:
zrequire('hello');
zrequire('banana');
If you run `./tools/test-js-with-node`, the
"banana" test will pass, because while it
does require "hello", it won't actually
*execute* the code that happens at require
time for "hello", because it's already in
the cache. Here is the code that gets
skipped:
const hello_world = i18n.t('Hello world');
But then if you try to run the banana test
individually, the above line of code will
cause the test to crash. And it will crash
even before you actually try to test the
meaningful code here:
exports.foo = () => {
display_greeting(hello.get_greeting());
};
This commit fixes this leak scenario by just
aggressively clearing out things from the
require cache.
This slows tests down by about 10%, which I think
is worth the extra safety here.
This is not always a behavior-preserving translation: _.extend mutates
its first argument. However, the code does not always appear to have
been written to expect that.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
This is not always a behavior-preserving translation: $.extend mutates
its first argument. However, the code does not always appear to have
been written to expect that.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
This allows us to collect coverage for Handlebars templates, and also
improves the readability of Handlebars-related stack traces.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
This effectively reverts the following
commit from May 2019:
be527905ca
The implementation of closest() was a bit
buggy and complex. It's easy enough
to just stub the method yourself. We may
want to eventually re-implement it, but we
should follow the template of parent/set_parent.
If you fail to stub `closest` zjquery gives
a fairly helpful error message:
Error: You must create a stub for $("link-stub").closest
We now require all of our unit tests to handle
blueslip errors for warn/error/fatal. This
simplifies the zblueslip code to not have any
options passed in.
Most of the places changed here fell into two
categories:
- We were just missing a random piece of
setup data in a happy path test.
- We were testing error handling in just
a lazy way to ensure 100% coverage. Often
these error codepaths were fairly
contrived.
The one place where we especially lazy was
the stream_data tests, and those are now
more thorough.
This is relatively unobtrusive, and we don't send
anything to the server.
But any user can now enter blueslip.timings in the
console to see a map of how long things take in
milliseconds. We only record one timing per
event label (i.e. the most recent).
It's pretty easy to test this by just clicking
around. For 300 users/streams most things are
fast except for:
- initialize_everything
- manage streams (render_subscriptions)
Both do lots of nontrivial work, although
"manage streams" is a bit surprising, since
we're only measuring how long to build the
HTML from the templates (whereas the real
time is probably browser rendering costs).
We now require the actual tests to explicitly
to zrequire Dict, rather than magically adding this.
In one case, the use of Dict was clearly just for
the test (not the app), so I converted that an ordinary
JS object (see timerender.js).
ES6 and TS modules don’t insert themselves into `window`, so our tests
shouldn’t insert them either. Since the test `window` behaves like
`global` now, we can rely on legacy modules that do insert themselves
to do it themselves.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
This commit was originally automatically generated using `tools/lint
--only=eslint --fix`. It was then modified by tabbott to contain only
changes to a set of files that are unlikely to result in significant
merge conflicts with any open pull request, excluding about 20 files.
His plan is to merge the remaining changes with more precise care,
potentially involving merging parts of conflicting pull requests
before running the `eslint --fix` operation.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
ESLint won’t convert these automatically because it can’t rule out a
behavior difference arising from an access to a self-referential var
before it’s initialized:
> var x = (f => f())(() => x);
undefined
> let y = (f => f())(() => y);
Thrown:
ReferenceError: Cannot access 'y' before initialization
at repl:1:26
at repl:1:15
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
Because of the separate declarations, ESLint would convert them to
`let` and then trigger the `prefer-const` error.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
With webpack, variables declared in each file are already file-local
(Global variables need to be explicitly exported), so these IIFEs are
no longer needed.
Signed-off-by: Anders Kaseorg <andersk@mit.edu>
This adds the general machinery required, and sets it up for the file
`typing_status.js` as a first use case.
Co-authored-by: Anders Kaseorg <anders@zulipchat.com>
Adds a electron_bridge event that takes in message id and reply recived from
the notification reply and sends a message. We do this in webapp so desktop
doesn't have to depend on narrow and channel modules.
We also modify zjunit to reset window.electron_bridge after every run
to avoid leaking it.
Not all our errors actually happen in the contexts we were
wrapping (e.g. `setTimeout` and `_.throttle`). Also this fixes the
neat Firefox inspector feature that shows you where your event
handlers for a given DOM element actually live.
Using this "semi-modern" browser event means that Safari 9 and older
and IE10 and older may not have our browser error reporting active;
that seems fine giving the vanishing market share of those browsers.
https://blog.sentry.io/2016/01/04/client-javascript-reporting-window-onerror
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
It seems like the de facto standard ES polyfill library these days,
and we already depend on it through simplebar.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
The minimal syntactic sugar it might provide isn’t worth the
unexpected side effects (including side effects on third party
modules).
For now, we allow zrequire to emulate the previous syntax in the Node
test suite, even though stealing part of the NPM namespace is
confusing.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
We now use a Proxy to wrap zjquery elements, so
that we can detect callers trying to invoke methods
(or access attributes) that do not exist. We try
to give useful error messages in those cases.
The main impact here is that we force lots of tests
to explicitly stub `length`.
Also, we can't do equality checks on zjquery
objects any more due to the proxy object, but the
easy workaround is to compare selectors. (This
is generally an unnecessary technique, anyway.)
The proxy wrapper is fairly straightforward, and
we just have a few special cases for things like
"inspect" that happen when you try to print out
objects.
We no longer store handlers as an array of functions,
and instead we assume that code will only ever set up
one handler per sel/event or sel/event/child. This is
almost always a sane policy for the app itself.
We also try to improve error handling when devs write
incorrect tests.
The only tests that required changes here are the
activity tests, which were a little careless about how
data got reset between tests.
Apparently, we didn't have one of these, and thus had a moderate
number of generally very old violations in the codebase. Fix this and
clear the ones that exist..
Adds box-shadow to `#searchbox` when either `#search_query` or any
of the pills have focus. Uses jquery instead of pure css as the
`:focus` event occurs on `#search_query`, while we want to add
box-shadow to `#searchbox`. This could have been done with
`:focus-within` CSS selector, but it is not supported in IE or Opera.
`#search_query` already had an onfocus/focusout listener, adding
listeners to `#searchbox.pills` for those events wouldn't have worked
as you don't want the focusout event to fire when the focus shifts
from input to pill.
Also adds `focusin`, `focusout` and `css()` to zjquery. `css` is
same as `val`, except it returns an empty object in case of no value
instead of an empty string. I don't think `css()` is valid syntax
in actual jquery.
The reason to add this api is that many times some elements are
already used/cached and then their value interfere/exists in
other tests which gives false results.
This is preparation for our migration of our JS pipeline to webpack,
which includes as part of the process a hack of exporting globals via
the window object.
We have less urgency to test all templates now. The
most common error is probably unbalanced tags, and our
python-based template checker catches those problems
pretty well.
It's still possible to create bad templates, of course,
but the node tests have never been super deep at finding
semantic errors.
This also updates node_tests to use new constructor which is uppercase,
and some properties that are changed to be more clear now, like
jsdom().defaultView which is meant to the window object is now called window.
Ref: https://github.com/jsdom/jsdom/blob/master/Changelog.md
This run_test helper sets up a convention that allows
us to give really short tracebacks for errors, and
eventually we can have more control over running
individual tests. (The latter goal has some
complications, since we often intentionally leak
setup in tests.)
This introduces a generic class called list_cursor to handle the
main details of navigating the buddy list and wires it into
activity.js. It replaces some fairly complicated code that
was coupled to stream_list and used lots of jQuery.
The new code interacts with the buddy_list API instead of jQuery
directly. It also persists the key across redraws, so we don't
lose our place when a focus ping happens or we type more characters.
Note that we no longer cycle to the top when we hit the bottom, or
vice versa. Cycling can be kind of an anti-feature when you want to
just lay on the arrow keys until they hit the end.
The changes to stream_list.js here do not affect the left sidebar;
they only remove code that was used for the right sidebar.
There was really no reason for this to be a nested function, since
we weren't closing on any variables. Flatter is better. Also, it
is plausible that folks will want more control over creating
individual jQuery elements (but still want this helper).
I don't think anybody ever really used this feature, which I
developed but don't even use myself. It kind of runs counter
to the minimalist approach of the rest of node tests.
I would eventually like to re-think the template tests altogether.
They're slow, and we could solve that somewhat by replacing
jsdon/jquery with an HTML parser library to verify structural
things.
It's also possible that we can just rely on our template linters
to catch the biggest class of errors (malformed tags) and let
code review do the rest.
And it's also possible that we should make a second attempt to
ramp up tooling on making it easy to verify templates, but it
doesn't have to be part of the node tests. If we did that, we
would also potentially use tooling for Python-side templates.
We flatten the code a bit by removing a check that type is object,
and we replace it later with a check that type is string.
We also no longer allow document-like objects to be wrapped based
on the location-attribute-is-present hack. Instead, we want the
tests to just set document to 'document-stub'.
We can now extend zjquery using the $.fn mechanism. This isn't
necessarily recommended for test code (since you can just stub
individual objects directly), but some of our real code does this.
Added support for passing a filename without `.js` suffix.
This then fixed the issue of no complaints for invalid test
files. Now, throws an error for invalid test files.
Fixes#8579.
Clicking the cancel button removes all the changes and the user
group returns back to the original state. Saved button is showed
once the changes are saved on blur.
This restructures organization settings and permissions to be
more accurately grouped and for the permissions page to not be too
long.
CHANGES:
PROFILE:
(this was split out)
organization-profile-admin.handlebars:
form #1:
name
description
(SUBMIT)
avatar:
(UPLOAD)
(DELETE)
SETTINGS:
organization-settings-admin.handlebars:
language (mostly untouched)
message editing:
time limit/history/retention
message feed:
mandatory-topics
preview images
preview websites
PERMISSIONS:
organization-permissions-admin.handlebars
(mostly stuff was removed)
Joining:
restrict domains
require invite
User Identity:
name changes
email changes
Streams/Emoji:
creating streams:
waiting period (ADDED)
adding emojis
(SUBMIT) for whole panel
The profile group (name, description, avatar) were split into a new
page that did not previously exist, and the permissions was stripped
of message settings (message editing, message feed), but keeping the
"waiting period" input and putting it in the "Streams & custom emoji"
section.
Fixes: #5844.
We are phasing out the following in tests:
add_dependencies - this is just kind of a clunky UI
require - normal JS requires cause test leaks
In order to plug require leaks, we are effectively doing what
we always have done inside of add_dependencies, which is to
keep track of which modules we have done `require` on, and
these get cleared between tests.
Now we just use `zrequire` every time we want to pull in real
code to our global namespace.
This allow the webbpack dev server to properly reload JavaScript modules
while running in dev without restarting the server. We need to connect
to webpack-dev-server directly because SockJS doesn't support more than
one connection on the same host/port.
This commit add $.create(), which allows you to create a
jQuery object that just has a name to identify it, as opposed
to some selector or HTML fragment. It's useful for things that
are really used as stubs.
This also fixes a bunch of the existing tests to use $.create().
Before this fix, you could actually just do $('some-stub'), but
now we enforce that the input to $() looks like a valid selector
or HTML fragment, and we make some exceptions for things like
window-stub and document-stub.
Hopefully this will make it more explicit that zjquery does
not truly simulate DOM, but it instead allows you to dynamically
set what you want the results of $('foo').find(some_selector)
to be.
Before this commit, we were erroneously setting up parents
as part of add_child() calls, but it's not necessarily the
case that those children are immediate children, and therefore
the first object is not necessarily the immediate parent.
jQuery's behavior in methods that, because of their nature, don't need to
return anything is to return the element itself in the jQuery object form.
Now the zjquery element is returned when one of these methods is called.
There are some cases when the jQuery dollar
function is called with an element as argument.
If such element has already been created using
zjquery, we should simply return it.
The node tests have purged modules from cache that were
included via things like set_global(), but calling require
directly would leak modules into the next test, which made
a couple tests only work when you ran the whole suite. I
fixed those tests to work standalone. And then I now make
dependencies explicitly clear the require cache before we
require them in namespace.js.
We now make it so that $('foo').addClass(whatever) and similar
functions properly return the wrapper object for chaining
purposes. We may eventually want to change the wrapper object
to automatically dispatch to the first child object, but this
should work for now.
The with_overrides() function no longer works at the module
level, so you can temporarily override one function in a module
without breaking more permanent monkey patches.
This makes us slightly more vulnerable to unintentional test
leakages, but it solves the problem where you often need lots
of temporary overrides for writer functions but you want to
keep a more permanent override for some sort of reader function.
The function with_overrides() uses the logic from the old
run() function in dispatch.js to allow you to call a test
function and override parts of the global namespace only
for the duration of when test_function runs.
- Remove `underscore.js` from `static/third` and fetch it from `npm`.
- Upgrade `underscore.js` to 1.8.3.
- Bump up the `PROVISION_VERSION` to 4.2.
Part of #1709
- Remove `codepointat` from `static/third` and fetch it from `npm`.
- Upgrade `codepointat` to 0.2.0.
- Bump up the `PROVISION_VERSION` to 4.1.
Part of #1709.
We now sort lists of users ids deterministically, and we also
sort list of emails deterministically and without regard to case.
This probably fixes the bug #2343, although I never got a great
repro on that.
Now we just have two methods of importance:
compile_template
render_template
The compile_template() function will automatically
(and recursively) compile any partials it depends
on and mark those as compiled.
When the render function is run now, it uses the partial_finder
function to search recursively through files for partials and add them
so that test writers don’t have to.
This means that we no longer have to do any manual work to maintain
the templates.js check that all handlebars templates are rendered by
the node tests.
This separates the display settings module from the
settings_table.handlebars template.
Additionally, it fixes the node tests to search in the
static/templates/settings directory and initialize any templates in there
while running tests on the settings templates.
This makes sure we are explicit about partials in
individual test modules. Eventually, we should figure
out a way to make partials automatically compile as
part of the node tests.
Create `media.css` using media queries that had been at the bottom
of `zulip.css`, then update miscellaneous setttings/docs files.
I also add `.screen-medium-show` and `.screen-narrow-show` to
`media.css`, as they seem to be an important part of our
responsive design.
Fixes#1532.
Some node tests used to pass as long as prior tests ran,
but then they would fail if you ran them standalone. Now
we are more aggressive about cleaning up node's require
cache after each individual test runs.
This introduces a very minor different in behavior if you specify
an invalid filename as a command line argument. We now show
warnings for those *before* running the rest of the tests.
Steve Howell
Stefan Weil
Anders Kaseorg
Greg Price
Tim Abbott
Priyank Patel
Thomas Ip
Vishnu Ks
Puneeth Chaganti
Cynthia Lin
Shubham Padia
Rohitt Vashishtha
Shubham Dhama
Preston Hansen
Joshua Pan
Robert Hönig
Andy Perez
rht
Brock Whittaker
Pweaver (Paul Weaver)
Steve Howell
Aditya Bansal
Yago González
Rafid Aslam
Sahil Dua