Without this fix, the zed.dev server doesn't recognize the version with
metadata and redirects to /releases, which
defaults to the Stable channel, so Preview users end up seeing Stable
release notes.
Closes#47539
Changes:
- Strip pre and build metadata from the semver version in
release_notes_url before constructing the path, so the URL
becomes /releases/preview/0.218.0 instead of
/releases/preview/0.218.0-beta.1+preview.131.68e98a53.
- Standardized the ordering of pre/build clearing to pre-first across
all 4 call sites (2 were build-first).
Open question: With this fix, the URL now points to the specific version
you're running (e.g.
/releases/preview/0.218.0). An alternative would be to link to
/releases/preview (no version) so users can see all
Preview releases, including ones newer than what they have installed. I
went with per-version linking since it matches
the existing intent of the code, but happy to change it if the team
prefers the other approach.
Release Notes:
- Fixed "View Release Notes" in Preview/Stable opening the wrong release
channel page due to build metadata in the URL.
Closes#48256
It appears that the current git remotes support was implemented in
#42819, following the design described in
https://github.com/zed-industries/zed/pull/42486#issuecomment-3524092306.
That design does not include an interaction for deleting remotes, but
the delete remote branch action was mistakenly implemented as deleting
the remote itself.
After this PR, there should be no code paths that use `remove_remote`
anymore. I've kept it for now though, as it may be useful when we
introduce the corresponding interaction in the future.
Release Notes:
- Fixed a bug where deleting a remote branch from the branch picker
would incorrectly remove the entire remote configuration
solution for #43004
Release Notes:
- Added workspace::FocusCenterPane
The `editor::ToggleFocus` action not working really drives me bonkers,
and it also doesn't really make sense, because it doesn't toggle
anything. it moves focus, but doesn't move focus back anywhere if you
run it again so 'ToggleFocus' seems misleading.
Here we have added an action that just moves focus to the most recent
center pane: `workspace::FocusCenterPane`
this action correctly moves the focus to the center pane from any dock,
even if the center pane is the project diff, branch diff, terminal, or
keymap editor, etc.
i'm not sure if this would be considered a 'replacement' for
`editor::ToggleFocus` but i do think this is probably more in line with
what most users would consider the expected behavior.
This PR updates the pet dependency to incorporate performance
improvements from upstream.
Related Upstream Changes:
- Issue: microsoft/python-environment-tools#369
- PR: microsoft/python-environment-tools#370
Updating pet reduces toolchain discovery time in WSL environments,
specifically when Conda is installed on the host Windows system. In my
testing, the time from clicking the toolchain button to the list
appearing was reduced by approximately 2 seconds.
While this is a notable improvement, discovery in WSL remains slightly
slower than in native development or remote Linux environments, where
toolchains are typically loaded near-instantaneously.
Before you mark this PR as ready for review, make sure that you have:
- [x] Added a solid test coverage and/or screenshots from doing manual
testing (no need)
- [x] Done a self-review taking into account security and performance
aspects
- [x] Aligned any UI changes with the [UI
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
Release Notes:
- Improved python toolchain discovery under WSL environment
With a large number of files in the git status, pressing stage/unstage
could cause a noticeable freeze. The async performance profiler showed
that the task spawned inside `change_all_files_stage` blocked the main
thread for 300ms+, resulting in a clear and visible UI stall. The main
cause was the need to traverse all entries to determine what to stage or
unstage, which becomes expensive for large change sets.
Same repro as: #47800
Release Notes:
- Improved latency of stage/unstage operations
Remove old multiworkspace commands and fix a bug where multiworkspaces
wouldn't be deserialized from the DB
Before you mark this PR as ready for review, make sure that you have:
- [x] Added a solid test coverage and/or screenshots from doing manual
testing
- [x] Done a self-review taking into account security and performance
aspects
- [x] Aligned any UI changes with the [UI
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
Release Notes:
- N/A
---------
Co-authored-by: eric <eric@zed.dev>
- In the project header, arrow keys left and right expand/collapse the
group
- Enter and space in any thread list item selects the item
- Arrow key down from the search editor moves focus to the the list
- Arrow key up from the first list item puts focus back to the editor
- At any moment while focus on the list, cmd-f moves focus back to the
search editor
- Made the `FocusWorkspaceSidebar` action always reset focus back to the
search editor
Release Notes:
- N/A
- Selection/focus improvements (reduces flickers, move selection more
correctly throughout the list)
- Adds open folder button in the sidebar header
- Fixes sidebar header design, including the thread view one, too
- Fixes behavior of cmd-n when focused in the sidebar
- Changes the design for the "new thread" button in the sidebar and adds
a preview of the prompt on it
- Rename items in the "start thread in" dropdown
Release Notes:
- N/A
---------
Co-authored-by: cameron <cameron.studdstreet@gmail.com>
Switch python_env_kernel_specifications to a buffered stream (using
buffer_unordered(4) and filter_map) to run discovery concurrently with a
bounded parallelism and skip None results.
Closes#50988
Before you mark this PR as ready for review, make sure that you have:
- [x] Added a solid test coverage and/or screenshots from doing manual
testing
- [x] Done a self-review taking into account security and performance
aspects
- [x] Aligned any UI changes with the [UI
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
Release Notes:
- N/A
We were sending the raw tool debug output as JSON to the model rather
than whatever the tool intended as content for the model.
Which meant we were sending unneeded information to the model, which
matters in the edit tool case.
Release Notes:
- N/A
This PR updates the `UserStore` to use the `default_organization_id`
returned from the server to set the default organization rather than
always picking the first one.
Closes CLO-530.
Release Notes:
- N/A
Closes#12628
Adds a `$ZED_LANGUAGE` task variable that resolves to the language name
of the active buffer (e.g., "Rust", "Python", "Shell Script"). This lets
tasks adapt behavior based on language without parsing file extensions.
Use cases from the issue:
- Pass syntax highlighting language to external tools (e.g., `rg --type
$ZED_LANGUAGE`)
- Adjust comment wrapping width per language
- Handle extensionless files and untitled buffers that have a language
assigned
VS Code provides equivalent functionality via
`${command:activeEditorLanguageId}`. Neovim exposes it as `&filetype`.
## Changes
- Added `Language` variant to `VariableName` in
`crates/task/src/task.rs`
- Populated from `buffer.language().name()` in
`BasicContextProvider::build_context`
(`crates/project/src/task_inventory.rs`), following the same pattern as
`File` and `Stem`
- Updated test fixtures in `crates/tasks_ui/src/tasks_ui.rs` to include
the new variable
- Added `ZED_LANGUAGE` to the variable list in `docs/src/tasks.md`
## Testing
Updated the existing `test_task_variables` test in `tasks_ui` to verify
`ZED_LANGUAGE` resolves to "Rust" and "TypeScript" for the respective
test buffers.
This contribution was developed with AI assistance (Claude Code).
Release Notes:
- Added `$ZED_LANGUAGE` task variable that exposes the current buffer's
language name
---------
Co-authored-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com>
Co-authored-by: Kirill Bulatov <mail4score@gmail.com>
Before we kind of had it mixed: you could change thinking but not model
etc. We brute forced a solution by disabling, but it seems better to
just allow all of them to be updated whenever we would build the
request.
Release Notes:
- N/A
Rather than always having a ThreadHistory with an internal option,
changed it to be an Option<ThreadHistory> so we can better distingiuish
between an empty list and unsupported
Release Notes:
- N/A
Closes #ISSUE
Before you mark this PR as ready for review, make sure that you have:
- [ ] Added a solid test coverage and/or screenshots from doing manual
testing
- [ ] Done a self-review taking into account security and performance
aspects
- [ ] Aligned any UI changes with the [UI
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
Release Notes:
- N/A *or* Added/Fixed/Improved ...
Closes#35947
## Summary
When a settings file was moved to a second pane and Zed restarted, the
file appeared in both panes. Closing it in one pane would close it in
the other.
## Root cause
`Editor::deserialize` restored standalone files (like settings.json) by
calling `workspace.open_abs_path()`, which routes through
`open_path_preview` → `pane.open_item()`, adding the editor to the
workspace's default pane. The caller (`SerializedPane::deserialize_to`)
then also adds the item to the target pane, so it ends up in two panes.
This also caused the SQL constraint violations @MrSubidubi noted: the
`items` table has `PRIMARY KEY(item_id, workspace_id)`, so the duplicate
triggers errors on the next serialisation cycle.
## Fix
Replace `workspace.open_abs_path()` with `project.open_local_buffer()`,
which opens the buffer without touching any pane. Pane placement is left
to `deserialize_to`.
https://github.com/user-attachments/assets/68d3c5b4-d002-429f-b907-ec21cb0019ec
## Test plan
- [x] Reproduced the original bug (settings file duplicated across panes
after restart)
- [x] Verified the fix: file restores only to the correct pane
- [x] Added regression test
(`test_deserialize_non_worktree_file_does_not_add_to_pane`)
- [x] Existing `items::tests::test_deserialize` passes (all 6 cases)
- [x] `cargo clippy -p editor` clean
Release Notes:
- Fixed settings file being restored to multiple panes after restart
([#35947](https://github.com/zed-industries/zed/issues/35947)).
Release Notes:
- Fixed send failure for pasted file#line mention links
Handle `MentionUri::Selection` with a file path in
`confirm_mention_for_uri`
by resolving it to text content via existing line-range loading logic,
instead
of returning "Unsupported mention URI type for paste".
Keep untitled-buffer selections (`abs_path: None`) as an explicit
unsupported
error for now.
Add a regression test for pasting selection mentions to ensure they
resolve to
`Mention::Text` with the expected line range.
Signed-off-by: Xiaobo Liu <cppcoffee@gmail.com>
Update Helix's handling of the following actions in order to ensure that
selections are created when using subword motions:
* `vim::NextSubwordStart`
* `vim::NextSubwordEnd`
* `vim::PreviousSubwordStart`
* `vim::PreviousSubwordEnd`
The handling of these motions was done by
`vim::helix::Vim::helix_move_and_collapse`, which simply moved the
cursor without doing any selection. This commit updates it to correctly
use both `vim::helix::Vim::helix_find_range_forward` and
`vim::helix::Vim::helix_find_range_backward`.
The added tests have been confirmed against Helix's behavior of the following commands:
* `move_next_sub_word_start`
* `move_prev_sub_word_start`
* `move_next_sub_word_end`
* `move_prev_sub_word_end`
Closes#41767
Release Notes:
- Fix subword motions in Helix mode to select traversed text
---------
Co-authored-by: Conrad Irwin <conrad.irwin@gmail.com>
Co-authored-by: dino <dinojoaocosta@gmail.com>
### Split token display for xAI
Extends the split input/output token display (introduced in #46829 for
OpenAI) to all xAI models.
Instead of the combined `48k / 1M` token counter, xAI models now show:
- **↑** input tokens used / input token limit
- **↓** output tokens used / output token limit
#### Before
<img width="513" height="128" alt="Screenshot 2026-02-08 at 11 07 13 AM"
src="https://github.com/user-attachments/assets/14e5cb4a-9b5c-4081-bbfb-407a737bf234"
/>
#### After
<img width="610" height="126" alt="Screenshot 2026-02-08 at 11 05 36 AM"
src="https://github.com/user-attachments/assets/92396dcb-8905-4f87-9b9e-d8b0f63225ba"
/>
#### Changes
- **x_ai.rs** — Override `supports_split_token_display()` to return
`true` on `XAiLanguageModel`. All built-in Grok models already implement
`max_output_tokens()`, so no additional plumbing was needed.
- **cloud.rs** — Add `XAi` to the `matches!` pattern in
`CloudLanguageModel::supports_split_token_display()` so cloud-routed xAI
models also get the split display.
#### Tests
- `test_xai_supports_split_token_display` — Verifies all built-in Grok
model variants return `true` for split token display.
- `test_xai_models_have_max_output_tokens` — Validates all built-in Grok
models report `max_output_tokens` that is `Some`, positive, and less
than `max_token_count` (required for the UI to compute the input token
limit).
- `test_split_token_display_supported_providers` — Confirms the cloud
provider match pattern includes `OpenAi` and `XAi` while excluding
`Anthropic` and `Google`.
Release Notes:
- Changed the display of tokens for xAI models to reflect the
input/output limits.
---------
Co-authored-by: Ben Brandt <benjamin.j.brandt@gmail.com>
Co-authored-by: Smit Barmase <heysmitbarmase@gmail.com>
## Summary
When no stop tokens are provided, Zed was sending an empty array
(`"stop": []`) to Ollama. This caused Ollama to override the model's
default stop tokens (defined in its Modelfile) with nothing, resulting
in models like rnj-1:8b generating infinitely with literal stop tokens
appearing in the output.
## Problem
Models with custom stop tokens in their Modelfile (like `<|eot_id|>` for
rnj-1:8b) would generate forever because:
1. Agent thread creates request with `stop: Vec::new()` (empty)
2. Ollama provider converts this to `stop: Some(vec![])`
3. Serializes as `"stop": []` in JSON
4. Ollama interprets this as "override default stop tokens with nothing"
5. Model generates forever, outputting stop tokens as literal text
## Solution
1. In `crates/language_models/src/provider/ollama.rs`:
- Only send `stop` when explicitly provided
- When empty, use `None` so the field is omitted from JSON
2. In `crates/ollama/src/ollama.rs`:
- Add `#[serde(skip_serializing_if = "Option::is_none")]` to all
`ChatOptions` fields
- Ensures `None` values are omitted, not serialized as `null`
## Testing
Added 4 new tests in `crates/ollama/src/ollama.rs`:
- `test_chat_options_serialization`: Verifies None fields are omitted
- `test_chat_request_with_stop_tokens`: Verifies stop tokens are
serialized when provided
- `test_chat_request_without_stop_tokens_omits_field`: Verifies empty
stop is omitted
All 11 ollama tests pass, plus 1 language_models ollama test.
Fixes#47798
Release Notes:
- Fixed Ollama models with custom stop tokens generating infinitely by
not overriding model defaults when no stop tokens are specified.
## Summary
This PR enables token usage reporting in streaming responses for
OpenAI-compatible APIs (OpenAI, xAI/Grok, OpenRouter, etc).
## Problem
Currently, the token counter UI in the Agent Panel doesn't display usage
for some OpenAI-compatible providers because they don't return usage
data during streaming by default. According to OpenAI's API
documentation, the `stream_options.include_usage` parameter must be set
to `true` to receive usage statistics in streaming responses.
## Solution
- Added StreamOptions struct with `include_usage` field to the open_ai
crate
- Added `stream_options` field to the Request struct
- Automatically set `stream_options: { include_usage: true }` when
`stream: true`
- Updated edit_prediction requests with `stream_options: None`
(non-streaming)
## Testing
Tested with xAI Grok models - token counter now correctly shows usage
after sending a message.
## References
- [OpenAI Chat Completions API -
stream_options](https://platform.openai.com/docs/api-reference/chat/create#chat-create-stream_options)
- [xAI API Documentation](https://docs.x.ai/api)
Closes#46078
When we unloaded sessions, we weren't necesarily making sure we had
saved the current state. Now we save it whenever we remove it from the
session map.
Release Notes:
- Agent: Fix cases where thread state wasn't saved before closing the
thread.
Closes#37836
This behavior was already fixed for Supermaven in #37047, but is still
present in Copilot. What's actually happening:
- Receive a multi-line edit prediction
- Dismiss it with escape
- Clicking anywhere in the editor below the cursor calls
`Editor::select` which starts out by calling `Editor::hide_context_menu`
-> `Editor::update_visible_edit_prediction`, bringing back the
prediction that was just dismissed and updating the editor's display map
- The subsequent selection logic in `Editor::select` now operates using
a display map that is inconsistent with what the user saw when clicking
- If the click was anywhere where the prediction inlay used to be,
`Editor::select` thinks the user clicked on the inlay and does nothing,
and the inlay reappears
- If the click was below where the prediction inlay used to be, the
inlay is immediately removed again but the cursor is moved to the wrong
position because the inlay temporarily added a vertical offset to all
lines after it in the buffer
Ultimately, `Editor::select` should be handling the user input using the
same display map that the user saw when making the input. This can
obviously be solved in multiple ways, I chose to clear the current
completions in `CopilotCompletionProvider::discard` such that any
subsequent calls to `Editor::update_visible_edit_prediction` doesn't
immediately nullify the dismissal of the edit prediction by the user.
Note that this also changes the behavior that occurs after dismissing an
edit prediction, moving to a different position in the buffer, and then
returning: currently, this resurfaces the dismissed edit prediction
(which the `test_copilot` test exercises), and after this change it
doesn’t. This current behavior didn't seem desirable to me because it
doesn't happen when using Zeta or Supermaven, but if we want to keep it,
then we could fix the incorrect selection behavior some other way.
Release Notes:
- Fixed bug that resurfaced dismissed Copilot edit predictions when
moving the cursor around
Co-authored-by: Piotr Osiewicz <24362066+osiewicz@users.noreply.github.com>
Co-authored-by: Ben Brandt <benjamin.j.brandt@gmail.com>
Add `gg`/`G` (vim), `cmd-up`/`cmd-down` (macOS), and
`ctrl-home`/`ctrl-end` (Linux/Windows) keybindings to scroll to the top
and bottom of the markdown preview.
The markdown preview already has page scroll (`ctrl-d`/`ctrl-u`), line
scroll (`ctrl-e`/`ctrl-y`), and item scroll (`alt-up`/`alt-down`) but
was missing top/bottom navigation. This adds two new actions —
`ScrollToTop` and `ScrollToBottom` — using the existing
`ListState::scroll_to()` infrastructure, following the same pattern as
the other scroll actions.
- [x] Done a self-review taking into account security and performance
aspects
Release Notes:
- Added scroll-to-top and scroll-to-bottom keybindings for markdown
preview (`gg`/`G` in vim mode, `cmd-up`/`cmd-down` on macOS,
`ctrl-home`/`ctrl-end` on Linux/Windows)
When a file is deleted, the parent folder isn't colored which makes it
hard to know unless we look at the Git Panel.
This PR adds `Color::Deleted` to fix that.
<img width="366" height="428" alt="Screenshot 2026-03-14 at 7 56 04 PM"
src="https://github.com/user-attachments/assets/6ed3a3c7-e7e7-4b48-b89c-705a7c1ff634"
/>
Release Notes:
- Improved project panel to color folder with deleted file
When setting up on macOS 26, `xcodebuild -downloadComponent
MetalToolchain` can fail if Xcode hasn't been initialized yet. Running
`xcodebuild -runFirstLaunch` first resolves this.
Release Notes:
- N/A
What
---
The keybinding hint in the Project Search status bar button tooltip
disappears shortly after the tooltip appears, while the label "Project
Search" remains. This is most reproducible with no file open.
Why
---
`SearchButton` used `Tooltip::for_action` with no stored focus handle,
so `KeyBinding::render` called `window.focused(cx)` at render time to
resolve the `pane::DeploySearch` binding. When the `Tooltip` entity
re-rendered (it subscribes to theme settings changes) and the focused
element was no longer in a `Pane` key context, the keybinding lookup
failed silently and the shortcut hint vanished.
Fix
---
Store the active pane item's `FocusHandle` in `SearchButton` via
`set_active_pane_item` and use `Tooltip::for_action_in` so the
keybinding is always resolved against a stable `Pane` context. Falls
back to `Tooltip::for_action` when no pane item is active.
Closes#51562
Test Plan
---
- [x] `cargo fmt --check`
- [ ] `cargo test --workspace`
- currently blocked by an unrelated compile error test
- tests::test_terminal_kill_allows_wait_for_exit_to_complete ... FAILED
- [x] Manual verification of:
- Try to reproduce the issue and the keyboard shortcut visible for as
long as the cursor is hovering over the button.
Release Notes
- Fixed Project Search status bar button tooltip keybinding hint
disappearing while hovering
Screenshots
---
Current:
<img width="835" height="496" alt="image"
src="https://github.com/user-attachments/assets/cb1ffdf2-7733-47f6-9030-041459c2734c"
/>
Expected:
<img width="835" height="496" alt="image"
src="https://github.com/user-attachments/assets/84cb54b9-290c-4210-b86b-7ae9f8bf9ac0"
/>
Release Notes:
- N/A
Before you mark this PR as ready for review, make sure that you have:
- [x] Added a solid test coverage and/or screenshots from doing manual
testing
- [x] Done a self-review taking into account security and performance
aspects
- [ ] Aligned any UI changes with the [UI
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
Release Notes:
- N/A
Related issue #48600, following #48904
Setting `"formatter": null` does not actually disable the formatter, as
it is functionally equivalent to leaving the setting undefined. In this
case, Zed falls back to the default behavior—either "auto" or a specific
language server, depending on the language.
This PR adds a "none" option to the formatter setting, allowing users to
explicitly disable formatting while keeping `code_actions_on_format`
enabled separately.
```json
"formatter": "none"
```
I still have some doubts about the necessity of this setting itself, but
if it is a desired addition, I will update the documentation
accordingly. If not, please feel free to close this PR.
- [x] Tests or screenshots needed?
- [x] Code Reviewed
- [x] Manual QA
Release Notes:
- Added `"formatter": "none"` in settings to explicitly disable the
formatter.
Closes#50151
Before you mark this PR as ready for review, make sure that you have:
- [ ] Added a solid test coverage and/or screenshots from doing manual
testing
- [x] Done a self-review taking into account security and performance
aspects
- [ ] Aligned any UI changes with the [UI
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
Release Notes:
- Fixed a deadlock that froze the entire app when opening a new window
from a PopUp window or switching focus between multiple windows on
macOS.
- check `content_size==0`
- entry is removed from `downloading_files`
- empty file is writte via `smol::fs::write(..., &[])`
this means the receiver handles the empty files completely as soon as
the State message arrives, without ever needing a Chunk that will never
come
Closes#51551
Before you mark this PR as ready for review, make sure that you have:
- [x] Added a solid test coverage and/or screenshots from doing manual
testing
- [x] Done a self-review taking into account security and performance
aspects
- [ ] No UI CHANGES
Release Notes:
- Fixed empty files not downloading from a remote project
Closes#51620
gopls requires more specific initialization for semantic tokens that
other language-servers for reasons that aren't entirely clear to me, it
is however a small fix to make it work. still trying to track down other
malfunctioning language servers in the semantic token realm, so if you
have suggestions let me know.
original behavior (on nightly v0.229):
<img width="1784" height="1123" alt="nightly_behavior"
src="https://github.com/user-attachments/assets/5d976b34-2720-425c-8617-800291862909"
/>
updated behavior:
<img width="1784" height="1124" alt="fix_applied"
src="https://github.com/user-attachments/assets/17c528b9-fea8-4309-9306-44673afc9fc3"
/>
Before you mark this PR as ready for review, make sure that you have:
- [x] Added a solid test coverage and/or screenshots from doing manual
testing
- [x] Done a self-review taking into account security and performance
aspects
- [x] Aligned any UI changes with the [UI
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
Release Notes:
- gopls: fixed semantic token support with gopls.
Currently the web backend for gpui doesn't send any events to the
`InputHandler`'s like `EntityInputHandler` which are needed for the
input example and the editor crate, among others. This PR makes it pass
those events, in addition to also dealing with composition events so
that IME works. It adds an invisible input element to listen for
composition events, since canvases don't receive them.
Release Notes:
- N/A
This PR fixes that when you are below a header and want to select a
previous item, we would select a non-visible item because we did a -1 of
the newly selected index, which could end up inside a collapsed
directory.
**Note**: This is only an issue for the git tree view mode.
**Before** (See selection is not visible anymore after select previous,
because it's inside the collapsed directory)
https://github.com/user-attachments/assets/bd032768-b947-45f8-b5a1-6185010202cf
**After** (See selection now selects the first visible item in the list
which is the collapsed directory)
https://github.com/user-attachments/assets/47f67237-be7c-4ef5-8b99-6a3a2b016086
Before you mark this PR as ready for review, make sure that you have:
- [x] Added a solid test coverage and/or screenshots from doing manual
testing
- [x] Done a self-review taking into account security and performance
aspects
- [x] Aligned any UI changes with the [UI
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
Release Notes:
- Git UI: Fix selecting previous item would select a non-visible item if
it's inside a collapsed directory
## Summary
Fixes#48725
When using the keymap editor to delete or re-bind default keybindings,
two issues caused stale entries to persist:
- **Deleting a default binding** created a `NoAction` suppression entry
in `keymap.json` (correct for GPUI-level suppression), but the keymap
editor displayed it as a phantom `<null>` row alongside the greyed-out
default. Searching by keystroke showed both entries, and clash detection
counted the phantom as a real binding.
- **Changing a default binding's keystroke** (e.g. `⌥⌘T` → `⇧⌃⌘T`) added
the new binding but never suppressed the original default under the old
keystroke. The old default continued appearing in keystroke searches and
clash warnings.
## Fix
1. **Filter `NoAction` suppression bindings from the keymap editor
display.** These remain in the internal binding list for conflict
detection (so overridden defaults are still correctly greyed out), but
are excluded from the visible match results.
2. **Append a `null` suppression for the old keystroke when replacing a
non-user binding with a different keystroke.** When `update_keybinding`
converts a `Replace` to `Add` for a non-user binding and the keystroke
changes, it now also writes `{"old-keystroke": null}` (with the original
context) to suppress the stale default.
## Reproduction steps (verified fixed)
1. Open the keymap editor
2. Click Search by Keystroke and press `⌥⌘T`
3. You should see "agent: new thread" and "pane: close other items"
(both Default)
4. Right-click "agent: new thread" and choose Delete
5. Double-click "pane: close other items", change keystroke to `⇧⌃⌘T`,
click Save
6. Verify no `<null>` phantom row appears, and no stale defaults remain
under `⌥⌘T`
7. Close and reopen the keymap editor, search `⌥⌘T` — no results
8. Search for "editor: wrap selections in tag" and assign `⌥⌘T` — no
clash warnings
## Test plan
- [x] Existing `keymap_update` and `test_keymap_remove` tests pass
- [x] Added test: replacing a non-user binding **without** changing the
keystroke produces no suppression
- [x] Added test: replacing a non-user binding **with** context and
keystroke change produces a suppression that preserves the context
- [x] Manual verification of all reproduction steps above
Release Notes:
- Fixed an issue with the keymap editor where where the defaults would not be completely removed when deleting or overriding default keybindings
---------
Co-authored-by: Ben Kunkle <ben@zed.dev>
Harden the terminal tool's permission system to reject commands
containing shell substitutions and interpolations (`$VAR`, `${VAR}`,
`$(…)`, backticks, `$((…))`, `<(…)`, `>(…)`) before they reach terminal
creation.
## Changes
### Shell command parser (`shell_command_parser`)
- Added structured terminal command-prefix extraction with env-var
prefix support
- Added parser-backed validation that classifies commands as
Safe/Unsafe/Unknown
- Extended normalized command extraction to include scalar env-var
assignments in order
- Preserved quoted assignment values when they contain whitespace or
special characters
### Pattern extraction (`agent/pattern_extraction`)
- Updated terminal pattern extraction to use structured parser output
- Included env-var prefixes in generated allow patterns
- Normalized regex token boundaries to `\s+` while preserving display
whitespace
### Tool permissions (`agent/tool_permissions`)
- Added invalid-terminal-command rejection for forbidden
substitutions/interpolations
- Added unconditional allow-all bypass (global default Allow, or
terminal-specific Allow with empty patterns)
- Preserved hardcoded denial precedence over allow-all
### Terminal tool (`agent/tools/terminal_tool`)
- Updated tool description and input schema to explicitly prohibit shell
substitutions
- Added comprehensive SEC-264 regression test suite (20 new tests)
covering:
- All forbidden constructs (`${HOME}`, `$1`, `$?`, `$$`, `$@`,
`$(whoami)`, backticks, `$((1+1))`, `<(ls)`, `>(cat)`, env-prefix
variants, multiline, nested)
- Allow-all exception paths (global and terminal-specific)
- Hardcoded-denial precedence
- Env-prefix permission flow (matching, value mismatch rejection,
multiple assignments, quoted whitespace)
Closes SEC-264
Release Notes:
- Terminal tool permissions regexes can now match environment variables
(e.g. `FOO=bar cmd arg1 arg2`)
- If terminal tool permissions have active permissions regexes running
on them, then bare interpolations like `$FOO` are disallowed for
security, since regexes wouldn't be able to match on them.
Cameron Mcloughlin