Ensure we `panic()` instead of crash on graphics memory buffer overflow
Also bump the buffer size to 32Mb from 8Mb to make this rarer (but still
possible)
Release Notes: Fixes some crahes due to lack of graphics buffer spacae
This PR cleans up some TODOs we had in the themes.
We won't be addressing these in the immediate term, so no need to have
them show up when folks are looking for TODOs to burn down before
launch.
I added some general notes as signposts until we're ready to revisit
this.
Release Notes:
- N/A
This release call was added during the conversion to gpui2. I think it
is probably valid, but want to remove it on the off chance that it is
causing the crash that we're seeing in the `livekit.multicast` thread
when leaving a room.
Most likely, this is not going to fix anything, and is just introducing
a small memory leak, but it is a step back to how the app worked with
gpui 1.
This release call was added during the conversion to gpui2. I think
it is probably valid, but want to remove it on the off chance that
it is causing the crash that we're seeing in the `livekit.multicast`
thread when leaving a room.
During layout of EditorElement we use 'm' character from current font to calculate sizes, panicking with fonts that do not have that character (e.g. Arabic fonts). It's not really EditorElement's fault, as it assumes that the font it's dealing with is gonna have that character available.
To prevent a crash, I added validation while loading a family that a given font contains the glyphs we're gonna use down the line.
An old fix was ported over from gpui1, and the two fixes could not exist
side-by-side. Delete this code and let the keymap handle it
Release Notes:
- (Added|Fixed|Improved) ...
([#<public_issue_number_if_exists>](https://github.com/zed-industries/community/issues/<public_issue_number_if_exists>)).
Before this change the views would continue to update in the background
of the "disconnected" dialogue, which was disconcerting.
[[PR Description]]
Release Notes:
- Fixed an edge-case where following didn't handle unshare correctly
Editors WrapMap could become desynchronised if user had an invalid font
specified in their config. Compared to Zed1, WrapMap ignored the
resolution failure instead of panicking. Now, if there's an invalid font
in the user config, we just fall back to an arbitrary default.
Release Notes:
- Fixed the editor panic in presence of invalid font name in the config
(fixes https://github.com/zed-industries/community/issues/2397)
---------
Co-authored-by: Conrad <conrad@zed.dev>
Co-authored-by: Conrad Irwin <conrad.irwin@gmail.com>
When logging the edit environment, we were logging the newest environment being sent into the EventCoalescer on the latest activity log, when we should've been logging the environment that was associated with the ended period within the EventCoalescer.
The goal of this PR is to send up events that contain the duration of
keyboard activity within Zed. I built the `EventCoalescer`, which gets
called called each time a key is pressed, within any environment (in the
case of this PR, within "editor" and "terminal). The" `EventCoalescer`
holds a start and end DateTime and adjusts the end instant as events
come in, until a timeout from the last event is hit.
I did my best to keep this mechanism efficient (avoiding vectors to
store moments in time, avoiding spawning timers threads, etc.), because
I know this is going to be per keystroke, but its behind a lock on the
telemetry struck, since it has to know when the environment changes, in
which point, it will automatically end the activity period, even if
there is no timeout. Because we have to have access to it from different
parts of the system, we have to go through the lock, which worried me a
bit (@mikayla-maki's intuition is that it should be fine).
As for the event, I take the time between the durations reported by the
event coalescer and send that up to zed - the indention is to
reconstruct the start and end times on zed.dev, in the same way we use
the event offset times and the time on zed.dev to get the official event
timestamp, and avoid the issue of the user having their system clocks
set wrong.
I'd really appreciate it if @nathansobo or @maxbrunsfeld could check
this out.
Release Notes:
- N/A
