This PR introduces `Nearest` scrolling strategy to `UniformList`. This
is now used in completions menu and the picker to choose the appropriate
scrolling strategy depending on movement direction. Previously,
selecting the next element after the last visible item caused the menu
to scroll with `ScrollStrategy::Top`, which scrolled the whole page and
placed the next element at the top. This behavior is inconsistent,
because using `ScrollStrategy::Top` when moving up only scrolls one
element, not the whole page.
https://github.com/user-attachments/assets/ccfb238f-8f76-4a18-a18d-bbcb63340c5a
The solution is to introduce the `Nearest` scrolling strategy which will
internally choose the scrolling strategy depending on whether the new
selected item is below or above currently visible items. This ensures a
single-item scroll regardless of movement direction.
https://github.com/user-attachments/assets/8502efb8-e2c0-4ab1-bd8d-93103841a9c4
I also noticed that some functions in the file have different logic
depending on `y_flipped`. This appears related to reversing the order of
elements in the list when the completion menu appears above the cursor.
This was a feature suggested in #11200 and implemented in #23446. It
looks like this feature was reverted in #27765 and there currently seem
to be no way to have `y_flipped` to be set to `true`.
My understanding is that the opposite scroll strategy should be used if
`y_flipped`, but since there is no way to enable this feature to test it
and I don't know if the feature is ever going to be reintroduced I
decided not to include it in this PR.
Release Notes:
- gpui: Add 'Nearest' scrolling strategy to 'UniformList'
This reverts commit 082b80ec89.
This broke clicking, e.g. in snippets like
```rs
let x = vec![
1, 2, //
3,
];
```
clicking between `2` and `,` is quite off now.
Release Notes:
- N/A
We've had several outages with a proximate cause of "vercel is
complicated",
and auto-update is considered a critical feature; so lets not use vercel
for
that.
Release Notes:
- Auto Updates (and remote server binaries) are now downloaded via
https://cloud.zed.dev instead of https://zed.dev. As before, these URLs
redirect to the GitHub release for actual downloads.
Applications can select a policy for when the app quits using the new
function `Application::with_quit_mode`:
- Only on explicit calls to `App::quit`
- When the last window is closed
- Platform default (former on macOS, latter everywhere else)
Release Notes:
- N/A
- Enable Alt+Numpad input
- For this to be effective, the default keybindings for Alt+{Number}
will need to be unbound. This won't be needed once we gain the ability
to differentiate numpad digit keys from alphanumeric digit keys.
- Fixes https://github.com/zed-industries/zed/issues/40699
- Fix a number of edge cases with dead keys
Release Notes:
- N/A
Technically these should not be possible to hit, but sentry says
otherwise. Turning these into errors should give us more information
than the abort due to unwinding across ffi boundaries.
Fixes ZED-321
Release Notes:
- N/A *or* Added/Fixed/Improved ...
Supersedes https://github.com/zed-industries/zed/pull/39910
At some point, these two (`index_for_x` and `closest_index_for_x`)
methods where separated out and some code paths used one, while other
code paths took the other. That said, their behavior is almost
identical:
- `index_for_x` computes the index behind the pixel offset, and returns
`None` if there's an overshoot
- `closest_index_for_x` computes the nearest index to the pixel offset,
taking into account whether the offset is over halfway through or not.
If there's an overshoot, it returns the length of the line.
Given these two behaviors, `closest_index_for_x` seems to be a more
useful API than `index_for_x`, and indeed the display map and other core
editor features use it extensively. So this PR is an experiment in
simply replacing one behavior with the other.
Release Notes:
- Improved the accuracy of mouse selections in Markdown
Release Notes:
- N/A
---
https://github.com/user-attachments/assets/d46b77ae-88ba-43da-93ad-3656a7fecaf9
The system menu is only support for macOS, so here just modify the macOS
platform special code.
The Windows, Linux used `ApplicationMenu`, I have already added
`checked` option to Zed's ContextMenu.
Then later when this PR merged, we can improve "View" menu to show check
state to panels (Project Panel, Outline Panel, ...).
Given that when we lose our devices unrecoverably we will panic anyways,
might as well do so eagerly which makes it clearer.
Additionally this PR replaces all uses of `ManuallyDrop` with `Option`,
as otherwise we need to do manual bookkeeping of what is and isn't
initialized when we try to recover devices as we can bail out halfway
while recovering. In other words, the code prior to this was fairly
unsound due to freely using `ManuallyDrop::drop`.
Fixes ZED-1SS
Release Notes:
- N/A *or* Added/Fixed/Improved ...
Improves formatting consistency across various docs, fixes some typos,
and adds a missing `debug_assert_paint` to `Window::on_action` and
`Window::on_action_when`.
Release Notes:
- N/A
Release Notes:
- N/A
When I was implementing Input, I often used `TextRun`, but `background`,
`underline` and `strikethrough` were often not used.
So make change to simplify it.
Closes#36206
Disclaimer: I did use AI for help to end up with this proposed solution.
😅
## Observed behavior of native apps on macOS (like Safari)
I first did a quick research on how Safari behaves on macOS, and here's
what I have found:
1. Safari seems to position new windows with an offset based on the
currently active window
2. It keeps opening new windows with an offset until the new window
cannot fit the display bounds horizontally, vertically or both.
3. When it cannot fit horizontally, the new window opens at x=0
(y=active window's y)
4. When it cannot fit vertically, the new window opens at y=0 (x=active
window's x)
5. When it cannot fit both horizontally and vertically, the new window
opens at x=0 and y=0 (top left).
6. At any moment if I activate a different Safari window, the next new
window is offset off of that
7. If I resize the active window and open a new window, the new window
has the same size as the active window
So, I implemented the changes based on those observations.
I am not sure if touching `gpui/src/window.rs` is the way to go. I am
open to feedback and direction here.
I am also not sure if making my changes platform (macOS) specific, is
the right thing to do. I reckoned that Linux and Windows have different
default behaviors, and the original issue mentioned macOS. But,
likewise, I am open to take a different approach.
## Tests
I haven't included tests for such change, as it seems to me a bit
difficult to properly test this, other than just doing a manual
integration test. But if you would want them for such a change, happy to
try including them.
## Alternative approach
I also did some research on macOS native APIs that we could use instead
of trying to make the calculations ourselves, and I found
`NSWindow.cascadeTopLeftFromPoint` which seems to be doing exactly what
we want, and more. It probably takes more things into consideration and
thus it is more robust. We could go down that road, and add it to
`gpui/src/platform/mac/window.rs` and then use it for new window
creation. Again, if that's what you would do yourselves, let me know and
I can either change the implementation here, or open a new pull request
and let you decide which one would you would like to pursue.
## Video showing the behavior
https://github.com/user-attachments/assets/f802a864-7504-47ee-8c6b-8d9b55474899🙇♂️
Release Notes:
- Improved macOS new window stacking
- On macOS, pasting now inserts the actual file path when the clipboard
contains a file URL (public.file-url/public.url)
- Terminal paste remains text-only; no temp files or data URLs are
created. If only raw image bytes exist on the clipboard, paste is a
no-op.
- Scope: macOS only; no dependency changes.
- Added a test (test_file_url_converts_to_path) that verifies URL→path
conversion using a unique pasteboard.
Release Notes:
- Improved pasting on macOS: now inserts the actual file path when the
clipboard contains a file URL (enables image paste support for Claude
Code)
On macOS, the Function key is reserved for system use and should not be
used in application code.
This commit updated keystroke matching and key event handling to ignore
the Function key modifier while users are typing or pressing
keybindings.
For some keyboards with compact layout (like my 65% keyboard), there is
no separated backtick key. Esc and it shares the same physical key. To
input backtick, users may press `Fn-Esc`. However, macOS will still
deliver events with Fn key modifier to applications. Cocoa framework can
handle this correctly, which typically ignore the Fn directly. GPUI
should also follow the same rule, otherwise, the backtick key on those
keyboards won't work.
Release Notes:
- Fixed a bug where typing fn-\` on macOS would not insert a `.
The Alt+Numpad buffer that's maintained by the input stack is getting
corrupted, leading to garbage characters being inserted on keystrokes
like Alt+Up. Disable the automatic handling of Alt+Numpad for now until
the cause of this corruption is understood. The Alt+Numpad input did not
work anyway, so this does not regress anything.
Release Notes:
- windows: Fixed corrupted characters being inserted when Alt is pressed
(preview only)
Works around an issue, which can be reproduced in the following program:
```rs
use windows::Win32::UI::Input::KeyboardAndMouse::{GetKeyboardState, VK_CONTROL};
fn main() {
let mut keyboard_state = [0u8; 256];
unsafe {
GetKeyboardState(&mut keyboard_state).unwrap();
}
let ctrl_down = (keyboard_state[VK_CONTROL.0 as usize] & 0x80) != 0;
println!("Is Ctrl down: {ctrl_down}");
}
```
In debug mode, this program prints the correct answer. In release mode,
it always prints false. The optimizer appears to think that
`keyboard_state` isn't mutated and remains zeroed, and folds the
`modifier_down` comparisons to `false`.
Release Notes:
- N/A
### Summary
This PR changes `gpui/build.rs` to look up the Windows SDK directory in
the registry instead of falling back to a hard-coded path.
---
### Problem
Currently, building `gpui` on Windows requires `fxc.exe` to be in `PATH`
or at a predefined location (unless `GPUI_FXC_PATH` is set). This
requires to maintain a certain build environment with proper paths/vars
or to install the specific SDK version.
It is possible to find the SDK automatically using the registry keys it
creates upon installation. Specifically in
`SOFTWARE\\WOW6432Node\\Microsoft\\Microsoft SDKs\\Windows\\v10.0`
branch there are:
* `InstallationFolder` telling the SDK installation location;
* `ProductVersion` telling the SDK version in use.
These keys provide enough information to locate the SDK binaries, with
added robustness:
* handles non-standard SDK installation path;
* deterministically selects the latest SDK when multiple versions are
present.
---
### Changes Made
* **Updated `crates/gpui/build.rs`**:
* added dependency on `winreg`
* introduced `find_latest_windows_sdk_binary()` helper
* updated fallback logic to use registry lookup
This PR only changes the fallback location, and does not touch the
established environment-based workflow.
Release Notes:
- N/A
---
### Impact
Reduces manual configuration needed to build GPUI on Windows.
---------
Co-authored-by: John Tur <john-tur@outlook.com>
`ICustomDestinationList::AppendCategory` rejects an empty `IObjectArray`
and returns an `E_INVALIDARG` error. Error propagated and caused an
early-return from `update_jump_list()`.
<img width="1628" height="540" alt="image"
src="https://github.com/user-attachments/assets/f8143297-c71e-42a1-a505-66cd77dfa599"
/>
Release Notes:
- N/A
A user on Discord requested this feature:
https://discord.com/channels/869392257814519848/1434188637389717556/1434188637389717556
I added a scrollbar setting called `completion_menu_scrollbar` to the
completion menu and defaulted it to "Never" to match past behavior.
Release Notes:
- editor: Add `editor.completion_menu_scrollbar` setting to show a
scrollbar in the completion menu
Closes#40360
This PR added heuristics to determine what variable/breakpoint list
entry has the longest width when rendered. I added this in so the
uniform list would correctly determine which item has the longest width
and use that to calculate the scrollbar size.
The heuristic can be off if a non-mono space font is used in the UI; in
most cases, it's more than accurate enough though.
Release Notes:
- debugger: Add horizontal scroll bars to variable list, memory view,
and breakpoint list
---------
Co-authored-by: MrSubidubi <dev@bahn.sh>
When doing a project wide search in zed on windows for `hang`, zed
starts to freeze for a couple seconds ultimately starting to error with
`Not enough quota is available to process this command.` when
dispatching windows messages. The cause for this is that we simply
overload the windows message pump due to the sheer amount of foreground
tasks we spawn when we populate the project search.
This PR is an attempt at reducing this.
Release Notes:
- Reduced hangs and stutters in large project file searches
Closes https://github.com/zed-industries/zed/issues/39104
This fixes an issue where the preview would not work for remote buffers
in the process.
Release Notes:
- Fixed an issue where the SVG preview would not work in remote
scenarios.
- The SVG preview will now rerender on every keypress instead of only on
saves.
Closes#41180
When using the fallback prompt renderer (default on Wayland), clicks
would bleed through into underlying windows. When the click happens to
hit a button that creates a prompt, it drops the
`RenderablePromptHandle` which is contained within `Window`, causing the
`Receiver` which returns the index of the clicked `PromptButton` to
return `Err(Canceled)` even though a button was pressed.
This bug appears in the GPUI `window.rs` example, which can be ran using
`cargo run -p gpui --example window`. MacOS has a native
`PromptRenderer` and thus needs additional code to be adjusted to be
able to reproduce the issue.
Release Notes:
- N/A *or* Added/Fixed/Improved ...
This reintroduces `layer_shell` support after #32651 was reverted. On
top of that, it allows setting options for the created surface,
restricts the enum variant to the `wayland` feature, and adds an example
that renders a clock widget using the protocol.
I've renamed the `WindowKind` variant to `LayerShell` from `Overlay`,
since the protocol can also be used to render wallpapers and such, which
doesn't really fit with the word.
Things I'm still unsure of:
- We need to get the layer options types to the user somehow, but
nothing from the `platform::linux` crate was exported, I'm assuming
intentionally. I've kept the types inside the module (instead of doing
`pub use layer_shell::*` to not pollute the global namespace with
generic words like `Anchor` or `Layer` Let me know if you want to do
this differently.
- I've added the options to the `WindowKind` variant. That's the only
clean way I see to supply them when the window is created. This makes
the kind no longer implement `Copy`.
- The options don't have setter methods yet and can only be defined on
window creation. We'd have to make fallible functions for setting them,
which only work if the underlying surface is a `layer_shell` surface.
That feels un-rust-y.
CC @zeroeightysix
Thanks to @wuliuqii, whose layer-shell implementation I've also looked
at while putting this together.
Release Notes:
- Add support for the `layer_shell` protocol on wayland
---------
Co-authored-by: Ridan Vandenbergh <ridanvandenbergh@gmail.com>
This happens quite often with cargo based diagnostics which may spawn
several lines (sometimes the entire screen), forcing the user to scroll
up to the start of the diagnostic just to see the hover message is not
great.
Release Notes:
- Fixed diagnostics hovers not working if the diagnostic spans out of
view
