We remove the StackOverflow link because it is now so dated as to be
irrelevant -- it does not use `self.ident`, and cargo-cults the return
value of PyThreadState_SetAsyncExc.
As noted in the docstring for this function, the timeout is
best-effort only -- if the thread is blocked in a syscall, it will not
service the exception until it returns. It can also choose to catch
and ignore the TimeoutExpired; in either case it will still be running
even after the `timeout()` function returns.
Raising a vare TimeoutExpired it still somewhat accurate, but obscures
that the backend thread may still be running along merrily. Notice
such cases, and log a warning about them.
Having just thrown an exception into the thread, it is often useful to
know _what_ was the slow code that we interrupted. Raising a bare
TimeoutExpired here obscures that information, as any `exc_info` will
end there.
Examine the thread for any exception information, and use that to
re-raise. This exception information is not guaranteed to exist -- if
the thread didn't respond to the exception in time, or caught it, for
instance.
The quote in question originates in python/cpython@b8b6d0c2c6, when
the code was added. However, the code stopped having that comment,
and was no longer able to return anything but 1 or 0, starting in
python/cpython@4643c2fda1 -- Python 2.5.
Remove the block.
There is no guarantee that the code passed into parse_unicode_emoji_code
is valid unicode. In the case that it is not, it might be better to
return undefined instead of throwing an exception: to represent a
non-parseable code.
For context, mobile currently returns custom emojis as emojis with
string names in their code property, instead of actual unicode.
Previously, these buttons were centered via flex, which meant that in
the rare case that a long list of private message recipients caused
the recipient area to line-wrap, these icons would be incorrectly
placed at the vertical center of the now multi-line block.
Fix this by setting an auto bottom-marging.
Fixes#21693.
Reformats two events (`reaction op: add` and `reaction op:remove`)
to follow the general format of events in the OpenAPI that are
returned by the `/get-events` endpoint.
Removes unneeded reference to `EmojiBase` schema in `user_status`
return value for the `/register-queue` endpoint. Also, clarifies
the text about the `user_status` object and fields being returned.
Change the logic for rendering PM threads in PM section to
be in the same as that of topics view --
In default view, only recent 5 PM threads would be shown
and append the active conversation as the 6th one at last
if not present in those 5.
In PM section with unreads, a maximum of 8 conversations
would be shown and rest of them would be hidden behind
the 'more conversations' li-item, clicking on which takes
to the zoomedIn view of PM section where all the present
PM threads would be visible.
Co-authored-by: Tim Abbott <tabbott@zulip.com>
This PR implements checking for a literal emoji match in emoji
typeaheads. In other words, if you paste or type panda face into an
emoji typeahead, panda face should be presented as an option to choose
from.
This behavior is currently present in the mobile app, adding it to
shared will enable both platforms to utilize this logic.
On the Debian 10 -> 11 upgrade, the server is running Zulip 4.x, which
lets us pass `--audit-fts-indexes` to `upgrade-zulip-stage-2` rather
than run the command as a separate step.
The reindex-textual-data tool needs the venv to be cable to run;
switch the order of the last two steps, making them now match the
Debian 9 -> 10 and 10 -> upgrades.
Ref #21296.
The old link here broke once we introduced separate APKs per ABI,
in zulip/zulip-mobile#5296.
We could make a direct link to app-armeabi-v7a-release.apk , the one
that's compatible with almost all devices. But perhaps better is to
just go back to linking to the release page, where the user can
choose the best APK for their device. (If they're in the habit of
downloading APKs manually to install on their device, then probably
that means they're going to be used to choosing the right one.)
User report and discussion:
https://chat.zulip.org/#narrow/stream/48-mobile/topic/Direct.20apk.20download.20link.20is.20404/near/1358758
marked.js provides a helpful error message asking for bugs to be
reported upstream, but since we're running a fork, we should redirect
such support requests to us.
We can triage as necessary.
This makes parse() more re-entrant.
This also drives out a change to the linkifiers
test, where I no longer couple the linkifiers
logic to markdown concerns. I probably should have
done this in an earlier commit, but better late
than never. I didn't bother to split out a commit
for the test stuff, since it's just tests and
the commit is still fairly atomic in nature.
e6e6010, 41ddf29 and other commits migrating components to Micromodal
have removed references to these divs. Did a global search and
removed these divs.
Adds a non-endpoint specific page to the API documentation about
organization-level roles and permissions for users in order to
highlight important and useful information for clients and API
users.
Also, adds links to new documentation page in related areas
of the API documentation.
It has always been pretty arbitrary what we did inside
of setup() vs. parse(), and we want to avoid unpredictable
results from other platforms neglecting to call setup().
On my machine you can parse a simple message in about
25 microseconds, based on a trial of a million messages
with the content of "**bold**". Whatever portion of
that time is related to setup-related things like
compiling regexes should be negligible from the user's
perspective, since we never run parse() in a loop.
We only need to loop through the preprocessors
once, and we should use the options passed
in to the parser, not the default options
from the original setOptions call.
The first loop here was doing nothing.
Our sub (i.e stream) and user_group objects have a bunch
of fields that aren't relevant to markdown parsing, so
we create narrow types that make it easier for us to
share code with mobile in the future.
I considered working purely in id space, but the problem
there is that user-entered stream names and user group
names need to be canonicalized.
The abstract_map() helper clarifies that our code
doesn't need a concrete Map object from JS. This
change is possibly premature until we get a little
bit closer on integrating with mobile, since it
solves kinda the same problem that we might handle
more elegantly with TypeScript or Flow.
OTOH I find it to be pretty non-intrusive for the
webapp.
These are the low-hanging-fruit places where we
can avoid using the helpers global.
The long term goal here is to make the markdown
code truly re-entrant, but some challenges still
remain.
Before this change, we would use **some** options relating
to parsing messages, but not all of them. The reason for
this was completely unintentional.
It's mostly a moot point, since the server sends back pretty
generic messages when you do something like invoke the
"/dark" command, and the message would parse the same way
whether or not the parser was looking for things like user
mentions or stream links.
In order to make this code predictable, I had to decide
whether we do a completely vanilla parse or a full message
parse. My decision now is mostly tactical. It's a trivial
one-line change to just use all the options for message
parsing, whereas it requires a major overhaul to allow a
vanilla parse.
I also predict that we will eventually want to parse these
server responses as if they were messages. I doubt the
zcommand responses would ever take advantage of it, but I
could imagine things like nag messages wanting to use user
mentions.
Even if my predictions are wrong, my decisions here are
pretty easy to reverse once we learn more.
For the particular case of zcommands, it is puzzling to me
why the server doesn't just send back HTML, but I don't want
to open that can of worms yet, as that would technically be
an API change.
For now I am happy with the one-line fix.
The zcommand code was calling directly into the "marked"
library, which was extremely misleading, since you don't
get a vanilla parse of the markdown due to the fact
that markdown.js calls setOptions at initialize time.
This commit shifts the responsibility to markdown.js
as well as adding a bit of test coverage, but it is
otherwise just a pure code-move refactoring.
The next commit will tweak things further.
This is mostly done for correctness reasons--it is
easiest from a logical standpoint to set the realm
emojis at the end of the function, since we do not want
them to be overwritten by normal emojis. The code
worked before this change, but it involved a clunky
check to map.has().
There is also probably a very minor performance
improvement insofar as N (the number of normal
emojis) is typically greater than R (the number
of realm emojis), and we eliminate N calls to
map.has in return for R calls to map.set. Even
if R is quite large, the readability advantages
probably far outweigh any performance considerations,
since we are using native map calls.
Thanks to Austin Riba for this suggestion.
The mobile app was never able to use the shared
version of emoji.js, because, among other problems
with our code organization, the emoji.js module
is strongly based on a mutate-the-data paradigm
that doesn't play nice with React. The way
that we mutate data and violate encapsuation
here is something that we would mostly want to fix
without even trying to shared code with mobile, so
subsequent commits will try to extract some pure
functions into a shared module.
The backend validates that URL inputs are RFC valid URLs (with no
specific length limit), but the frontend appears to have a maximum
length of 50 specified, likely because of a copy-paste error.
Increase the HTML maxlength for this input to 2048, which is a length
supported for URLs by all major browsers.
Fixes#21633
Alex Vandiver
Austin Riba
Alya Abbott
Sdrummolo
Lauryn Menard
jai2201
Tim Abbott
Aman Agrawal
Greg Price
Steve Howell
Greg Price
Dinesh
Lauryn Menard
Heidi Ahlberg
Archisman Das