The tabindex on this link doesn't actually do anything, because of the Safari
tab workaround. I added it anyway in case we remove that workaround later.
(imported from commit 11f31f2561907300b350c11732be88589d736f94)
This fixes a bug where you could click on the already-selected mode button, and
it would switch modes without changing which button is selected.
(imported from commit 263ee0b52ba844c52a3a60968bb1bbff73482412)
With the removal of process_compose_hotkey, the state machine now has only one
state. Everything else is based on things like "is a text box focused right
now", which is probably a better approach.
(imported from commit 0e39c03956d28e30d2bdbf3b285410ad0cacca3e)
load_old_messages cares whether it is acting on behalf of a narrow
(for_narrow) not whether it is acting on behalf of a button push.
(imported from commit 16c00e015478f94f0631e211a92a7066a38135a8)
During long-polling mode, we were not passing the server generation
number back to the client even when they requested it. This was
causing the web client to initiate reloads every time it got a
non-immediate return from a long-polling request.
(imported from commit 66c16bf5a1c18bdd975b09d672ebdb4db8d4755d)
This restores the time-travel functionality and fixes Waseem's laundry
list of problems with its original UI.
(imported from commit e30e02c25af994435adb815d26284b3669c945a4)
This was causing glitching where if you narrow so that zfilt has only
one message, and add messages on top (e.g. via get_old_messages), and
then add messages on bottom (e.g. by receiving one), we end up with
the bookend row missing. One can make this happen with narrowing,
but, this situation happens every time with time travelling.
(imported from commit 71d85980d8aa9431a17b33e9e5111fd3f76cecf3)
authenticated_api_view and authenticated_json_view call
update_user_activity with a client generated using
@has_request_variables with a default of e.g. get_client("website").
Because that get_client call only happens once on importing the
module, if those client objects didn't exist previously in the
fixtures, then the first test will generate objects with ids 2 and 3,
and then the second test will dump the database, restore from the
fixtures, and then eventually generate client objects with ids 4 and
5. But since the default values were only computed at module load,
we'll still end calling add_user_activity with client objects with ids
2 and 3, which don't exist in the newly restored database.
Fix this madness by just making sure those two client objects exist in
the database.
(imported from commit d940e129d077a560d9a0f96ec3daa2e16ce21c8b)
If the message doesn't exist, rows.get returns an empty jQuery object, not
'undefined'.
(imported from commit 40620f3e4853a662cb93939bf7f0695f81298777)
Previously, if the client was not interested in pointer updates, they
would still return early from get_updates() when the pointer was
updated.
(imported from commit ef9d8b5678b7e72f799840148577651ee10d47ad)
Realistically, if there is a pointer update, no other client is going
to have the same pointer. If they do, the client can ignore the
update.
(imported from commit ca2db60ed7a7c5e4b50a9fa8d350985460922a50)
This is so we only send a single pointer update when the user scrolls
through a bunch of messages at once.
(imported from commit 20d239d7179e5f57ada99968361a0f9b7b18c591)
The assertion-checking callbacks were never called before. Now they will at
least get called when the view invokes handler.finish() before returning, i.e.
the return_messages_immediately case.
I fixed up the assertion-checking callback for one of the test cases. For the
other two it wasn't clear what to do -- one has been known-broken for a long
time -- so I just removed them.
"Fixes" #277 in as much as that test case is gone now.
(imported from commit 82d1db26f36f82f24aa5b7ff9e5aa65ce24c9a8f)
Run this script on an existing realm to create or change default
streams, which new users will get upon account creation.
(imported from commit 8938dcbd3520d97d25b4c6ca783d35c9aef52df0)
Ensure that every result has one of:
* find what you typed in
* search for what you typed in
* Narrow to a stream related to what you typed
* Narrow to a person related to what you type
(imported from commit 2178f17932f951a48f53d982ef660942562b55dc)
Modify the Bootstrap default sorter to sort by:
1. Starts-with, in a case-sensitive way
2. Starts-with, in a case-insensitive way
3. Matches anywhere, case-sensitive
4. Matches anywhere, case-insensitive
(This fixes the Keegan-reported issue of "Testing!" taking precedence
over "test" when he types in "tes")
(imported from commit b2a0127956fe7a8bf1cbf30752a6ddc2c49b7198)
(But don't actually make it work yet -- that's for the next commit.
This makes the diff a bit easier to review, since it's really just
code moving from composebox_typeahead to typeahead_helper, and the
appropriate changes in zephyr.js, rather than code moving AND
changing)
(imported from commit 5cb2e836eeb8807f4eb98424558702d44a0e2b70)
We'll probably eventually get fancier with this, suggesting various
default sets depending on the company or making it easy for one person
to create default subs for several people at once.
Until we know what we want, keep it simple.
(imported from commit 14319dd50c67fe33ac6f15288dba4916ce0a89ac)
If Javascript breaks or stops working, the previous hack means that
Tab-Enter stops working on Firefox and Chrome (since the tab key now
needs to select two divs before it gets to the Send button.)
By putting the one div *after* the Send button, we ensure that this
keeps working gracefully on Firefox and Chrome (and presumably IE), at
the expense of Safari in this specific case.
(imported from commit 9c9a613b1b1718ff8f0b9ef7497ebb13db0ddc64)
The main problem with this code is that the error message doesn't go
away until you reload. To fix this, we need to add a noblock option
on get_updates.
(imported from commit dc45af397bcf06a218bda5dd224ebd5fdf3462db)
The reproducer for the issue here is:
- Scroll to the top of your feed
- Click on a stream name
- Open a popover
- Click on a subject name
- Note that your old popover doesn't go away, and that you can open
a new one.
The problem was that when we narrow, we call jQuery.empty() on the
zfilt table. That not only removes nodes but also clears event
handlers and jquery data. Thus, even though we have a reference to
the old popover element in the ui module, the element has forgotten
it had a popover. When we call .popover("destroy") it actually
creates a new popover, but never shows it.
(imported from commit 9721d60c78549bd2362833590b304952f2bdef2d)
When we switched to delegated event handling, the bound handler for
all of our events was #main_div, but the floating recipient bar lives
outside of #main_div. Additionally, the bar needs to inherit the zid
from the target recipient bar since it is used for the narrow.
(imported from commit 7c18e16f2e98436888a8edb81fbbdd4d17abfe2a)
The problem seems to be caused by a bug in Firefox. We work around
the issue by adding empty table rows to the top of the table and
removing them when the user is idle.
This resolves trac #413
(imported from commit 2b15a4a2241bd7e813800a42608d650e0d4fa4f0)
I'd like to think about how to polish aspects of this a bit more,
but would like us to be able to deploy master at some point today.
This is basically the philosophical equivalent of reverting the
user-visible UI changes introduced in
b7b6794ad635ec63269a2043cd48b02749fbffda
(imported from commit edfaadf26741c47120c3acf6c410d33025c0a260)
Previously, it was impossible to narrow to a completely empty narrow.
Now it is possible, and the code needs to be tweaked in a few ways
not to break in this case.
(imported from commit e4dd4159ad52d003fc11d0b8b6531322c12a3de8)
Fixes#396.
We could display an error message, but jumping right to the login page seems
smoother and conveys the same information.
This will discard any message being composed, but preserving it would have
security consequences that we should consider further before implementing that.
Hopefully, users only get logged out by an explicit action, so they can't
complain too much (but see #217).
(imported from commit aaa23ecf46c73e514117ae1010fc44e133f2ba07)
Previously, we were adding those narrowed messages to message_dict the
first time, and thus totally skipping those messages each additional
time you tried to "load old messages" on that narrow.
(imported from commit 77eef376e1165b86e3c599608a1b5089a09d51e0)
In get_display_recipient, the userprofile was selected incorrectly by
user_id instead of the userprofile_id. In production, this hasn't
resulted in a user-visible error because we use MySQL and user ids are
always equal to userprofile ids.
This does happen if you are using SQLite locally and run populate_db,
which adds a bunch of users in parallel in an insufficiently
transactional way.
(imported from commit c25a04b4919e3efdfc6996b03492f7714d9034e8)
This bug affected receiving messages while narrowed. When none of
the messages recieved matched the narrow predicate, we'd try to
render messages and then create a jQuery object out of a bunch of
newlines (the only thing that results from a render of 0 messages).
(imported from commit 81f5aa46fac06fe0e5a14a8757f245f90b5845cc)
This greatly reduces the number of events generated and prevents the
blue flashing on message arrival.
This also necessitated a change in how we looked up message rows to
add the 'next_is_same_sender' class, which led to a further
optimization where we don't have to do as many jQuery selections by
id.
(imported from commit 0bcd5688b483c560b6f3a29c6d36433da600e8ef)
And wire it up to our local copy of codehilite. This fixes highlighting in
fenced code blocks, e.g.
~~~~ .js
var x = function () {
return "hi";
};
~~~~
(imported from commit 0efb0c9b98a3acdf55e18bb1918af7960f3425be)
In particular:
* Taking a list of streams as arguments.
* Using the _backend model so that we can have an API version.
* Considering "not subscribed" to be a non-fatal error.
And of course the corresponding changes to subs.js.
(imported from commit fdb300c6aa6921c2c6b09c22bd1e64405c368809)
One way this can happen is if run-dev.py --test is already running. In that
case the one we launch will die with "Address already in use", and we should
alert the user that test results may not be accurate.
(imported from commit 078091cace2cff777b444668b03f96dc208f22fc)
This should address the catherio/tibbetts feedback of the name
breaking oddly across the middle of their name.
One notable change introduced by this commit is: If your name is very
long, e.g., "Waseemio Daherioian", it gets cut off. (On Firefox, it
gets rendered as "Waseemio Daherioia...", and on Chrome it gets no
ellipsis at all.)
The current behavior is that the long name actually overflows into
the main text area, which I think is worse.
(imported from commit 668cb30bc2326c255b229f4f19f29be473bdc1e8)
The existing code shortens the searchbox each time it receives focus.
Unfortunately, this means that if it receives focus twice in a row, it
shrinks twice in a row. (For some reason, the '/' hotkey does this).
So, instead, make it idempotent -- if we're already shrunk, don't
shrink us again.
(imported from commit 8179963bbd00822d15d92609d89f572d2de7800c)
This is nearly perfect, modulo two things:
1. If you have a search active and you resize the window, the search
box resize doesn't take effect until you exit the search.
2. In super-narrow windows (<380px), the searchbox overshoots
the message area slightly.
I don't regard either as huge issues -- I'll probably fix#1
eventually.
(imported from commit 4900fb9783cc9f447315b0892bd3505f5c31ce15)
This doesn't fully fix the problems related to not syncing
subscriptions to browser clients, but it does fix the instance that
everyone experiences.
(imported from commit be2bc31a7c4443c1678321f1a938496e2632c0d3)
This commit changes APIs and requires and update of all zephyr
mirroring bots to deploy properly.
(imported from commit 2672d2d07269379f7a865644aaeb6796d54183e1)
These tests don't have the same coverage as the json test on the
theory that the backend is shared by the two views and that
differences are mostly on the way into the backend functions.
(imported from commit ddd21135565122dae8cbe90846d1aee7e4a2f56e)
Adding a positional argument caused a problem when
@authenticated_api_view started using @has_request_variables
internally. The 'handler' argument used to be passed through
positionally to the wrapped function, but when using
@has_request_variables, the wrapper inside @authenticated_api_view
had to take additional arguments. The handler argument was then
assigned to one of those parameters instead of being passed through.
(imported from commit 66240bd465c803ddcbf4a603509051fca7381468)
If we don't do this, we get all kinds of nasty shadowing where
references to 'search.whatever' seem to be references to the
HTML input element, rather than our search.js module.
(imported from commit 4e4b562ddf895baea9619316d9fab27ae5e9fc4e)
We have a lot of forged users that have bad fullnames due to
historical versions of our fullname computations; this function will
clean those up.
Also, we have a bunch of users with emails like foo|mit.edu@mit.edu
that were the result of a mirroring bug that we want to get rid of
from autocomplete -- putting them in a useless realm name will do.
(imported from commit 6e305093653ca9d327e9e28491636e99d16cfe1d)
This should fix the problem where only one of pairs of identical
messages sent to two different zephyr classes by bots will make it
over.
(imported from commit 37005417e2e1f737501c9524b95b044eefbfe235)
Fixes a message forgery bug (#335).
This works because Django will not generate a new CSRF token if a valid token
cookie is already present (see django/middleware/csrf.py).
(imported from commit 23222cb0bb62ae8a2f8ac7fb3f24bbc866103454)
These were lists of pairs because we were going to repeat keys, but that didn't
work anyway.
(imported from commit 687b3f7b8a2821d057719c725f1f39db3992ae5c)
This was causing Zephyr mirroring to break because
create_mirrored_message_users was returning False due to the
same_realm_email check failing.
(imported from commit e6a63160f34ec056461038650b5f8027718e6c63)
Previously we bypassed the big buttons on the left bar and clicked on
the anchor tag inside the compose box. This bypassed the
compose.start() call.
(imported from commit 2b627825596c8d3c21441e58db895b8e488e624b)
Personals are now just private messages between two people (which
sometimes manifests as a private message with one recipient). The
new message type on the send path is 'private'. Note that the receive
path still has 'personal' and 'huddle' message types.
(imported from commit 97a438ef5c0b3db4eb3e6db674ea38a081265dd3)
It now takes an anchor message id, a number of messages before, and a
number of messages after. The result always contains the anchor
message.
(imported from commit 84d070dc8091161c86d4bbeafbdc299493890a2a)
We need a deterministic order for the client test suite, and it seems like a
good idea generally.
(imported from commit cc8fc555611f2d2f1b21e63ce6860d446baa3410)
We had this problem where clicking a hyperlink bubbles up and causes a
click on the message, which causes the composebox to open.
We "fixed" this by setting cancelBubble (or, even better, calling
stopPropagation()).
Unfortunately, on Firefox, this fix breaks Ctrl-click and Shift-click,
because those are (apparently) implemented by adding an event listener
on link clicks, and stopPropagation prevents them from being called.
We instead work around this by handling this case in the click handler
of the parent element. (This allows the normal URL click AND Firefox's
bound event handlers for Ctrl and Shift to run.)
This resolves Trac #374.
(imported from commit 16fb3aa6fc582f1fba5009812e0b1178ce7c5bb7)
Mixing these two in this file is bound to lead to a world of hurt (and
has, historically). At some point I'd like to do this across the
entire codebase.
(imported from commit 9ff029597587f9c37a0bd9f32c25a769aa1a7a20)
This makes the "handle hotkeys" code path a lot simpler, and also
fixes the "copy not working" issue we were seeing on Firefox 17.
(imported from commit 8ab96d12895da2876f60da58f373372612f4ba32)
So, in Firefox, $(window).width() does not include the width of the
scrollbar. However, the CSS media-query max-width DOES include the
width of the scrollbar -- so the Bootstrap change and our change do
not happen at the same time.
window.innerWidth does take into account the width of the scrollbar,
though, and seems to have reasonable cross-browser support, so we use
that instead.
(If we wanted to be slicker, we could use a media query a la
https://developer.mozilla.org/en-US/docs/DOM/window.matchMedia ,
but that's not supported in IE <10.)
This resolves Trac #35.
(imported from commit ca35321c02d5e79e4f9c439a662805c016a333ed)
Old browsers might not have the global JSON object, so we may have to
include something like https://github.com/douglascrockford/JSON-js
for old browsers in the future.
(imported from commit e30a291d1212f2a00b543551b3a77082c7406eec)
This was causing our tests to fail and would have also
affected API users not using our Python bindings
(imported from commit 2d81496892e9042e328279edea94be8ee4d21c1b)
return_messages_immediately's return value is not returned, so the
argument validation in @has_request_variables didn't work correctly.
@has_request_variables would return a json_error, but
send_with_safety_check expects a dict.
(imported from commit 86b6bccb7861dbf523c06b606b87374e339059a4)
The original check has become too broad now that we have more buttons,
and specifically this lets you use the search hotkey to start a new
search after you've been searching up and down.
(imported from commit 0e691ff55ff9d4be8d406d1eb47fc2062758d28b)
This change substantially increases the number of view functions where
the API and JSON versions are actually identical code.
(imported from commit 2eee55a8943cf9a684bec2ba1f6d7afcb2b91948)
Functions with the @has_request_variables decorator can have some of
their arguments extracted from the HTTP request. For each such
argument, its default value should be an instance of the POST class.
The arguments to the POST constructor control the request variable
name that the function parameter should be populated from (it
defaults to the same as the parameter name), whether the value should
be converted before being passed, and whether a default value should
be supplied if the parameter is missing from the request.
(imported from commit ba1c25d73ba3980e44abec1458e6496807fcdaa4)
No need for an 'if' if we're just returning a boolean. And using
QuerySet.exists() should be a little more efficient.
(imported from commit 69ec3cc9f2fe904ec40ea3b8a8687a06cd03f3f3)
login_required_api_view is misleadingly named. It accepts neither a Django
login session nor login credentials (username / password). The intent here is
authentication, whether stateful (login) or stateless (API key).
(imported from commit 7e9be552168396b399116737655bd7267fd5c1a3)
Within 'except', 'raise' re-raises the current exception. But outside, it produces
TypeError: exceptions must be old-style classes or derived from BaseException, not NoneType
which is pretty confusing as a generic "something has gone wrong" exception.
(imported from commit 9fcd003a952b82df67726c26161dced079978a32)
We've had multiple requests from MIT zephyr users to allow
non-alphanumeric stream names, and we haven't decided what we want to
allow, so for now allow everything.
Note that the web client and mirror script limit stream names to 30
characters, which is our database limit.
(imported from commit 2acb5ee04e5ee7c40031ac831e12d09d04bbb2e6)
From the Google JavaScript Style Guide:
If you need a map/hash use Object instead of Array in these cases because
the features that you want are actually features of Object and not of Array.
Array just happens to extend Object (like any other object in JS and
therefore you might as well have used Date, RegExp or String).
(imported from commit 048e7a640137f3919c0097a421b7b6c366b65cfe)
This clarifies that clicking on any of those three pieces of
information will pop up the user info tooltip.
(imported from commit 1e57550d66acbb2e8d5d244d2997bbd394c334c3)
This was causing us to log some requests twice, and might have more serious
consequences as well.
(imported from commit 0bb2d7207ee3e4e04679215a7f5ae637cd26aa19)
The zero-port case never actually worked, because addrport wasn't an optional
parameter in run_one. And multiple ports was implemented using the
multiprocessing library, which is just bad news. Since we have no need for
this, remove it before it can cause trouble.
(imported from commit 9d913924701f30d23ebe878b76c8f1f0da2800e2)
Keegan McAllister
Zev Benjamin
Jessica McKellar
Jeff Arnold
Waseem Daher
Tim Abbott
Luke Faraone