Use fully resolvable request paths because we need to be able to refer
to third party modules, and to increase uniformity and explicitness.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
For most functions that we were using __Rewire__ for,
it's better to just use the override helper, which
use __Rewire__ under the hood, but also resets
the reference at the end of run_tests.
Another nice thing about override() is that it reports
when you never actually needed the mock, and this
commit fixes the instances found here.
I didn't replace every call to __Rewire__. The
remaining ones fall under these categories:
* I looked for ") =>" in my code sweep,
so I missed stuff like "noop" helpers.
* Sometimes we directly update something
in a module that's not a function. This
is generally evil, and we should use setters.
* Some tests have setup() helpers or similar
that complicated this code sweep, so I
simply punted.
* Somes modules rely on intra-test leaks. We
should fix those, but I just punted for the
main code sweep.
This is a deceptively ugly diff. It makes
the actual code way more tidy.
I basically inlined some calls to mock_module
and put some statements in lexical order.
We now just use a module._load hook to inject
stubs into our code.
For conversion purposes I temporarily maintain
the API of rewiremock, apart from the enable/disable
pieces, but I will make a better wrapper in an
upcoming commit.
We can detect when rewiremock is called after
zrequire now, and I fix all the violations in
this commit, mostly by using override.
We can also detect when a mock is needlessly
created, and I fix all the violations in this
commit.
The one minor nuisance that this commit introduces
is that you can only stub out modules in the Zulip
source tree, which is now static/js. This should
not really be a problem--there are usually better
techniques to deal with third party depenencies.
In the prior commit I show a typical workaround,
which is to create a one-line wrapper in your
test code. It's often the case that you can simply
use override(), as well.
In passing I kill off `reset_modules`, and I
eliminated the second argument to zrequire,
which dates back to pre-es6 days.
This warning was added in #6551. It’s not for any version of the
current Electron app, which we warn about on the server side with
DESKTOP_WARNING_VERSION, but rather some pre-Electron app so ancient I
don’t even know what it is. Apparently it communicated using the
window.bridge global, so eradicate that too.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
We now call $.clear_all_elements at the top
of run_test.
We have to exempt two modules from the new regime:
compose
settings_user_groups
Also, if modules do set_global("$", ...) we don't
try to call the non-existent function.
It's possible we'll want to move to something like
this, but we might want to clean up the two
sloppy_$ modules first:
// AVOID THIS:
// const $ = require("zjquery")
run_test("test widget", ({override, $}) => {
override(foo, "bar", ...);
$.create(...);
// do stuff
});
We no longer export make_zjquery().
We now instead have a singleton zjquery instance
that we attach to global.$ in index.js.
We call $.clear_all_elements() before each module.
(We will soon get even more aggressive about doing
it in run_test.)
Test functions can still override $ with set_global.
A good example of this is copy_and_paste using the
real jquery module.
We no longer exempt $ as a global variable, so
test modules that use the zjquery $ need to do:
const $ = require("../zjsunit/zjquery");
We still need to write to these globals with set_global because the
code being tested reads from them, but the tests themselves should
never need to read from them.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
The maybe_clear_subscribers() function was an artifact of
when we used to attach subscribers to the "sub" records in
stream_data.js. I think it was basically a refactoring
shim, and due to some other recent cleanup, it was only
used in test code.
We also change how we validate stream ids.
Going forward, peer_data just looks up stream_ids with the
normal stream_data API when it's trying to warn about
rogue stream_ids coming in. As I alluded to in an earlier
commit, some of the warning code here might be overly
defensive, but at least it's pretty self-contained.
We now use the same code in all places to
get the bucket of user_ids that correspond
to a stream, and we consistently treat
a stream as having zero subscribers, not
an undefined number of subscribers, in
the hypothetical case of us asking about
a stream that we're not tracking.
The behavior for untracked streams has
always been problematic, since if a
stream is untracked, all bets are off.
So now if we don't "track" the stream,
the subscriber count is zero. None of
our callers distinguish between undefined
and zero.
And we just consider the stream to be subscribed
by a user when add_subscriber is called,
even if we haven't been told by stream_data
to track the stream. (We also stop
returning true/false from add_subscriber,
since only test code was looking at it.)
We protect against the most likely source
of internal-to-the-frontend bugs by adding
the assert_number() call.
We generally have to assume that the server
is sending us sensible data at page load
time, or all bets are off.
And we have good protections in place
for unknown ids in our dispatch code
for peer_add/peer_remove events.
