Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- N/A or Added/Fixed/Improved ...
Self-Review Checklist:
- [X] I've reviewed my own diff for quality, security, and reliability
- [N/A] Unsafe blocks (if any) have justifying comments
- [X] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [ ] Tests cover the new/changed behavior
- [X] Performance impact has been considered and is acceptable
Release Notes:
- Removed support for ACP extensions. Installed ACP extensions will be
migrated to use the ACP servers as provided by the ACP registry instead.
---------
Co-authored-by: Ben Brandt <benjamin.j.brandt@gmail.com>
Co-authored-by: Marshall Bowers <git@maxdeviant.com>
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [ ] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- N/A or Added/Fixed/Improved ...
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [ ] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- N/A
Adds a `runtime_diagnostics` section to the dev-only "Show Git Job
Queue" output so that when the queue gets stuck, we can tell from the
dump itself whether a git subprocess is wedged (and on supported
platforms, where it's wedged) without needing the user to run
`ps`/`sample`/`lsof` by hand.
The new section contains:
- **`processes`** — every transitive descendant of the Zed process, with
PID, PPID, name, executable, full argv, sysinfo status
(`Run`/`Sleep`/`Stop`/`Zombie`/etc.), and elapsed runtime.
Cross-platform via `sysinfo`. This is the single most useful field: it
instantly answers "is there a stuck `git` child or not?"
- **`linux_proc`** *(Linux only)* — for each descendant,
`/proc/<pid>/wchan` (the kernel function the thread is sleeping in, e.g.
`futex_wait_queue`, `pipe_read`) and `State:` from `/proc/<pid>/status`.
- **`macos_git_children`** *(macOS only)* — for any descendant whose
name contains `git`, a 2-second `sample` user-space stack and `lsof -p`
output. Each is included only if the corresponding system binary exists;
otherwise it's skipped.
- Windows gets just `processes` (no portable way to grab another
process's stack).
### Safety
- Cross-platform: only `sysinfo` (an existing workspace dep) is on the
always-compiled path. All `sample`/`lsof`/`/proc` code is behind
`#[cfg(target_os = ...)]` gates, so Windows builds never see those
symbols.
- Every fallible step is handled individually: on error it logs a
warning and the corresponding key is omitted from the JSON. The queue
dump is built and shown the same way whether `gather()` returned a
populated object, an empty object, or partial data.
- Diagnostics gather runs under `cx.background_spawn(...)` so the macOS
`sample` 2-second wait can't block the foreground.
- `sample`/`lsof` output is truncated to 64 KB per process at a UTF-8
char boundary.
Release Notes:
- N/A
Currently, the `diagnostics` tool queries the current state of the
workspace. However, if the LSP is slow to update diagnostics, this can
lead to stale diagnostics being given to the agent.
This PR changes the `diagnostics` tool to attempt to pull fresh
diagnostics. If it fails, it informs the agent and falls back to the old
behaviour (not all LSPs support pull-based diagnostics).
Release Notes:
- N/A or Added/Fixed/Improved ...
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- N/A
---------
Co-authored-by: Ben Kunkle <ben@zed.dev>
In the interactive MCP OAuth flow, the MCP client registers itself with
the authorization in one of three ways:
- Client ID Metadata Document aka CIMD (recommended default). This is
already implemented: https://zed.dev/oauth/client-metadata.json.
- Dynamic Client Registration (DCR). This is the traditional method.
Also already implemented in Zed.
- Pre-registration: the client is registered out of band, typically in
the IdP or SaaS provider's UI. You get a client id and maybe a client
secret, that have to be provided by the MCP client when it wants to
exchange an access token. This is what this pull request is about.
This PR has two main parts:
- Allow users to configure a client id and optional client secret for an
MCP server in their configuration, under a new `oauth` key, and take it
into account
- Make the MCP server state and the configuration modal aware of the
intermediate states (client secret missing) and error cases stemming
from client pre-registration.
The client secret can be stored either in the system keychain or in
plain text in the MCP server configuration. The UI tries to steer user
towards the more secure option: the keychain.
<img width="715" height="201" alt="Screenshot 2026-04-10 at 16 48 06"
src="https://github.com/user-attachments/assets/5e64103e-6746-4ef0-8bd9-533d492b6912"
/>
<img width="884" height="544" alt="Screenshot 2026-04-10 at 16 47 07"
src="https://github.com/user-attachments/assets/0e35bb3c-cbc4-4e8c-a713-66323597b2e2"
/>
<img width="785" height="558" alt="Screenshot 2026-04-10 at 16 47 23"
src="https://github.com/user-attachments/assets/03339187-1508-461a-87ae-a7c2647df9a5"
/>
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes
https://github.com/issues/assigned?issue=zed-industries%7Czed%7C52198
**Note for the reviewer: I know how busy the AI team is at the moment so
please treat this as low priority, we don't have signal that this is a
highly desired feature. It's a rather large PR, so I'm happy to pair
review / walk through it.**
Release Notes:
- Added support for OAuth client pre-registration (client id, client
secret) to the built-in MCP client.
Zed-managed npm installers were resolving a concrete latest version with
`npm info` and then installing `package@version`. That is brittle when
users
configure npm release-age filtering via `before` or `min-release-age`:
npm's
installer applies those rules during resolution, but our pinned install
target
could disagree with it, and therefore fail to install.
This changes managed npm installs to install `package@latest` and let
npm apply
its own resolver and user config. The local latest-version lookup
remains as a
best-effort cache freshness check, not as the exact install target.
Exact extension API installs remain unchanged because extensions
explicitly
request a package and version. If we want to revisit that we can.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes#53611
Release Notes:
- Fixed npm-backed tool installs to better respect npm release-age
filters.
Closes TRA-150
This PR makes the restricted mode more obvious by:
- Immediately opening the restricted mode modal upon opening an
untrusted project
- Disabling dismissing the modal on escape or click away to force
choosing one of the two options (and avoid accidentally staying in
restricted mode by simply dismissing it)
- Showing the LSP button but with communication about language servers
being disabled for untrusted projects
- Showing a banner in the project settings with the same communication
The motivation for this change was that we tried to be minimal with how
we communicate a project is untrusted, but it was so minimal that people
were confused as to why language servers and other settings weren't
working. It was easy to miss the title bar button, for some reason. The
changes in this PR makes it so acting on this decision (trust or not a
project) is mandatory in order to even start to interact with the
project. I appreciate changes here are more aggressive, but I think it's
better to make you think about this decision vs. letting you be confused
as to why you don't see LS completions or formatting.
Release Notes:
- Made restricted mode more obvious, demanding immediate action when
opening an untrusted project.
Registry archive agents install each version into a versioned cache
directory, but older extracted archives were left behind after updates
and
could accumulate disk usage.
After resolving the current archive cache directory, remove other
versioned
cache directories for the same registry agent while preserving the
current
version.
Release Notes:
- Fix certain ACP registry agents not cleaning up old versions
---------
Co-authored-by: Ben Brandt <benjamin.j.brandt@gmail.com>
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
Release Notes:
- Fixed context server availability updates when a new worktree is added
to or removed from a project.
---------
Co-authored-by: Ben Brandt <benjamin.j.brandt@gmail.com>
Closes AI-230
This PR makes skills, added as /-mentions, be rendered in the agent
panel as creases, like anything you'd @-mention. Naturally, clicking on
the crease button opens the corresponding skill file in a buffer.
It turned out to be quite a bit of plumbing to make this work,
particularly as I am also introducing an interface to display dividers
and headers in the completion menu. This was relevant to me to add
because it sets a good foundation to convert many agent panel-related
actions as slash commands.
Release Notes:
- N/A
---------
Co-authored-by: MartinYe1234 <52641447+MartinYe1234@users.noreply.github.com>
Adds a new language model provider that lets users authenticate with
their ChatGPT Plus/Pro subscription and use OpenAI models
(codex-mini-latest, o4-mini, o3) directly in the Zed agent — without
needing a separate API key.
## How it works
1. **OAuth 2.0 + PKCE sign-in**: Uses OpenAI's official Codex CLI client
ID to run an authorization code flow. A local HTTP server on
`127.0.0.1:1455` captures the callback, exchanges the code for tokens,
and stores them in the system keychain.
2. **Token refresh**: Access tokens are automatically refreshed when
they're within 5 minutes of expiry, using the stored refresh token.
3. **Responses API**: Requests go to
`https://chatgpt.com/backend-api/codex/responses` using the existing
`open_ai::responses` client (Responses API format, not Chat Completions
which was deprecated for this endpoint in Feb 2026).
4. **Required headers**: `originator: zed`, `OpenAI-Beta:
responses=experimental`, `ChatGPT-Account-Id` (extracted from JWT),
`store: false` in the body.
## Files changed
- `crates/open_ai/src/responses.rs`: Add `store: Option<bool>` field to
`Request`; add `extra_headers` param to `stream_response` for
per-provider header injection
- `crates/language_models/src/provider/openai_subscribed.rs`: New
provider (sign-in UI, OAuth flow, token storage/refresh, model list)
- `crates/language_models/src/provider/open_ai.rs`,
`open_ai_compatible.rs`, `opencode.rs`: Pass `vec![]` for new
`extra_headers` param
- `crates/language_models/src/language_models.rs`: Register the new
provider
- `crates/language_models/Cargo.toml`: Add `rand` and `sha2` deps for
PKCE
## Open questions / known gaps
- [ ] **Terms of service**: Usage appears to be within OpenAI's ToS
(interactive use via their official CLI client ID), but needs legal
sign-off before shipping
- [ ] **Redirect URI**: Currently `http://localhost:1455/auth/callback`
— may need to match exactly what OpenAI's Codex CLI uses
- [ ] **UI polish**: The sign-in card is functional but minimal; needs
design review
- [ ] **Error messages**: OAuth error responses from the callback URL
aren't surfaced to the user yet
- [ ] **`o3` availability**: o3 may require a higher subscription tier;
consider gating it
## Testing
Sign-in flow was designed to match the Copilot Chat provider pattern.
Manual testing against the live OAuth endpoint is needed.
Release Notes:
- Added ChatGPT subscription provider, allowing users to use their
ChatGPT Plus/Pro subscription with the Zed agent
---------
Co-authored-by: Zed Zippy <234243425+zed-zippy[bot]@users.noreply.github.com>
Co-authored-by: Richard Feldman <richard@zed.dev>
Co-authored-by: Richard Feldman <oss@rtfeldman.com>
Co-authored-by: Agus Zubiaga <agus@zed.dev>
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
fixvuejs/language-tools#5956
Align with
3fb3329dfb/extensions/vscode/src/extension.ts (L223).
Release Notes:
- Always pass null when body is non-existent in Vue language server
request
---------
Co-authored-by: Smit Barmase <heysmitbarmase@gmail.com>
With long lines or a small viewport, the stage/restore buttons often
cover code.
Adds a setting to hide the buttons altogether. Defaults to showing them.
Release Notes:
- Added: Setting to hide Git Stage/Restore buttons
The tool definition is very clearly contradicting the previous behavior.
Performance impact is unclear to me, we increase the work in a
potentially expensive loop, but it seems necessary to have both the
specified behavior from the tool definition, as well as the
heuristic/fallback for misbehaving models that seems to be intended.
Self-Review Checklist:
- [X] I've reviewed my own diff for quality, security, and reliability
- [X] Unsafe blocks (if any) have justifying comments
- [X] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [X] Tests cover the new/changed behavior
- [X] Performance impact has been considered and is acceptable
Closes#56225
Release Notes:
- Fixed tool paths preferring files in subdirectories named after the
project root
### Summary
Added a branch diff base branch picker that shows the selected diff base
and prioritizes branches on the active branch’s remote. This also
generalizes the branch picker’s checkout vs. select only behavior so its
UI/search/rendering can be reused by a future git graph picker, and
makes branch diff drop stale in progress base update tasks when the base
or repository changes for faster response.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #54050
Release Notes:
- git: Allow choosing branch diff base in branch diff view
---------
Co-authored-by: Remco Smits <djsmits12@gmail.com>
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- Fixed an issue where re-running the `git: branch diff` after changing
the active project would not refresh the branch diff to show the branch
diff of the active project
This PR adds support to create custom git task that are triggered from
the git graph context menu. Both Sublime Merge and Fork have custom
command support, and I use custom commands I've built out frequently to
speed up some of my tasks at Zed. Thus, I'd like to have support in
Zed's git graph so I can use Zed even more!
It offers initial support for the following env variables:
- `ZED_GIT_SHA`
- `ZED_GIT_SHA_SHORT`
- `ZED_GIT_REPOSITORY_NAME`
- `ZED_GIT_REPOSITORY_PATH`
These are only populated in the git graph context.
This PR also introduces a the `git-command` tag, which is needed so the
git graph can filter down to these custom git commands, and so other
tasks aren't polluting the context menu.
An example would be (in the global `tasks.json`):
```json
{
"label": "Branches containing commit: $ZED_GIT_SHA_SHORT",
"command": "git",
"args": ["branch", "-a", "--contains", "$ZED_GIT_SHA"],
"tags": ["git-command"],
},
```
And then in the context menu:
<img width="646" height="296" alt="SCR-20260511-mnfa"
src="https://github.com/user-attachments/assets/0e7b811b-f47d-4a2f-9270-99e392c38663"
/>
And the output in the terminal:
<img width="585" height="184" alt="SCR-20260511-mnks"
src="https://github.com/user-attachments/assets/54d7d205-6212-4eff-8dbb-c8e908996747"
/>
The awesome thing about using tasks is we get all the task
infrastructure for free, such as history!
<img width="591" height="292" alt="SCR-20260511-mnud"
src="https://github.com/user-attachments/assets/6315be8f-dd33-470f-bfcd-aa56d7fbfdce"
/>
<img width="602" height="173" alt="SCR-20260511-moch"
src="https://github.com/user-attachments/assets/b528422c-efcc-4a7d-9783-73d945e9665b"
/>
And we have all the task configuration options too out of the box.
---
Right now, this only works with global tasks. It isn't clear how to shoe
in support for worktree-specific tasks (`.zed/tasks.json`) in a clear
way, as not all invocations of the git graph are in an area that has a
clear worktree id, and so we sort of have to guess or fallback to
something else. That can be a followup once it's more clear how we
should cover that. Or maybe someone else has a better solution.
I chose to only ship with these 4 git-specific variables for now. The
git graph also currently ONLY resolve those variables. We can adjust /
add in more in follow up PRs.
Self-Review Checklist:
- [X] I've reviewed my own diff for quality, security, and reliability
- [X] Unsafe blocks (if any) have justifying comments
- [X] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [X] Tests cover the new/changed behavior
- [X] Performance impact has been considered and is acceptable
Release Notes:
- Added support for running global custom Git command tasks from the Git
Graph commit context menu.
Helps #51567
Refs #54531
Summary:
- Add total timeouts for ACP Registry JSON fetches and icon fetches,
including response body reads.
- Download registry icons concurrently and keep icon failures non-fatal,
so a blocked icon CDN does not delay registry availability by one
timeout per agent.
- Surface the stored registry fetch error in the ACP Registry empty
state and add a retry action.
This addresses cases where the registry request, or one of the icon
requests, never finishes. It does not make blocked networks succeed, but
it prevents the UI from sitting on `Loading registry...` indefinitely
and gives the user something actionable instead.
Test plan:
- `git diff --check HEAD~1..HEAD`
- `cargo fmt --check --package project --package agent_ui`
- `cargo check -p project`
- `cargo check -p agent_ui`
- `cargo test -p project --features test-support registry_refresh_`
Release Notes:
- Fixed the ACP Registry getting stuck on loading when registry or icon
requests hang.
---------
Co-authored-by: Ben Brandt <benjamin.j.brandt@gmail.com>
Co-authored-by: Smit Barmase <heysmitbarmase@gmail.com>
The function is unsound due to the classic fact that one can leak tasks,
sidestepping the blocking drop behavior resulting in a use after free.
Release Notes:
- N/A or Added/Fixed/Improved ...
Closes#48032
When restoring a diff hunk, we first unstage it unconditionally. That
unstaging operation is a no-op in terms of the index text if the hunk
was already not staged, but previously we would still always do
`spawn_set_index_text_job` and bump the
`hunk_staging_operation_count_as_of_write`. Bumping that count in turn
causes us to skip a diff recalculation in response to the change in the
buffer's text. That works out fine in the local case, because when the
worktree picks up the write to `.git/index` we kick off another diff
recalculation which is not skipped. But in the remote case, we don't get
an `UpdateDiffBases` proto message if the index text didn't actually
change, so there is no subsequent diff calculation to do the cleanup,
and we end up with a stale no-op hunk.
This PR fixes the issue by skipping the write to the index and the
`hunk_staging_operation_count_as_of_write` bump if the new and old index
texts are the same.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- Fixed a bug where restoring diff hunks in remote projects would leave
stale no-op hunks in the UI.
Adds a command to help debugging stuck git job queues
Release Notes:
- N/A or Added/Fixed/Improved ...
---------
Co-authored-by: Anthony Eid <hello@anthonyeid.me>
Experiment with allowing users to manage terminal sessions along with
threads in the sidebar.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- N/A
---------
Co-authored-by: Bennet Bo Fenner <bennetbo@gmx.de>
Git's `-d` flag deletes a branch only if it's fully merged into its
upstream or HEAD - this is what we were using before, which caused the
"not fully merged" error. The `-D` flag force deletes a branch even with
unmerged changes (equivalent to `--delete --force`).
### Before
Deleting an unmerged branch failed with a "not fully merged" error
toast.
### After
- Deleting an unmerged branch prompts for confirmation to force delete
- Delete button tooltip shows "Hold alt to force delete" hint
- Holding **alt** turns the delete icon red and tooltip changes to
"Force Delete Branch"
- Force delete keybinding: `cmd-alt-shift-backspace`
Release Notes:
- Added confirmation prompt when deleting unmerged git branches, with
option to force delete.
- Added alt+click on delete button to force delete a branch immediately.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
#55770 changed the npm package version spec to `package@<=1.2.3`. On
Windows this fails with `The system cannot find the file specified.`
because:
- `npm` resolves to `npm.cmd`, a batch file. Windows runs `.cmd` files
via cmd.exe, which parses the invocation and treats unquoted `<` as
input redirection.
- The single quotes our shell builder emits around args are PowerShell
string-literal syntax that PS strips during parsing. PS only re-adds
CRT-style transport quotes around native command args containing
whitespace, so `package@<=0.25.3` reaches `npm.cmd` bare and cmd.exe
fails before the batch body even runs.
Switch to npm's hyphen-range syntax (`0.0.0 - <version>`, equivalent to
`<=<version>`), which has no `<`.
Closes#55921
Release Notes:
- Fixed ACP agents failing to launch on Windows with "The system cannot
find the file specified"
Two bugs caused MCP server child processes (e.g. `npm`/`node` for
`mcp-remote`) to accumulate as zombie processes that were never cleaned
up:
**Bug 1: `stop_server()` only called `stop()` for `Running` servers**
If a server completed initialization but was still in `Starting` state
when `stop_server()` was called (a race between the init task completing
and `maintain_servers` restarting), the client/transport/process were
never released. The `Arc<ContextServer>` was moved into a `Stopped`
state with its inner client still holding the transport and child
process handle.
Fix: call `stop()` unconditionally in `stop_server()`. It is a safe
no-op when the client has not been initialized (`None`).
**Bug 2: `kill_on_drop` only killed the direct child, not the process
tree**
`StdioTransport` used a raw `smol::process::Child` with
`kill_on_drop(true)`, which sends SIGKILL only to the direct child
process (the shell/`npm` wrapper). The actual MCP server (e.g. `node
mcp-remote`) runs as a grandchild and survives the kill, getting
reparented to launchd.
Fix: use `util::process::Child`, which already exists in the codebase
for exactly this purpose. It calls `setsid()` via `pre_exec` to make the
child a process group leader, and uses `killpg()` to terminate the
entire process tree on kill. This requires passing a
`std::process::Command` (via `build_std_command`) instead of a
`smol::process::Command` (via `build_smol_command`), because that is
what `util::process::Child::spawn` accepts — it needs to call `pre_exec`
on the `std::process::Command` before internally converting it to
`smol::process::Command` for async I/O.
Release Notes:
- Fixed zombie MCP server processes accumulating over time
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes#54641
Release Notes:
- Fixed creating git worktrees in WSL remote projects from Windows.
## Summary
- Preserve the repository `PathStyle` when constructing git worktree
paths.
- Avoid using local OS path separators when creating worktrees for
remote Posix projects such as WSL.
- Keep worktree archive path checks aligned with the same
path-style-aware worktree directory calculation.
## Root Cause
Remote repository paths are stored as `PathBuf`s, but `PathBuf::join`
and related path operations use the client OS separator. On Windows
clients connected to WSL, this could turn a remote Linux path into a
mixed path like `/home/<user>/\home\<user>\dev\worktrees\...`, causing
worktree creation and opening to fail.
## Validation
- `cargo fmt --all --check`
- `git diff --check`
- `cargo test -p project
test_new_worktree_path_uses_posix_style_for_remote_paths`
- `cargo test -p project test_worktree_directory_uses_remote_path_style`
- `cargo test -p project test_join_path_for_style_uses_remote_separator`
---------
Co-authored-by: Max Brunsfeld <maxbrunsfeld@gmail.com>
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes#55022
## Summary
When a task explicitly uses `cmd.exe`, Zed prepares it as a shell
invocation like `cmd.exe /S /C "<command>"`.
If Python virtual environment activation is enabled,
`Project::create_terminal_task` wraps that prepared task in an outer
activation shell so the venv can be activated before the task runs.
Before this change, the activation formatter treated the prepared `/C`
command string as a normal shell argument and quoted it again. On
Windows, that can make cmd receive escaped quotes literally, producing
errors like `'\\"echo Hi there\\"' is not recognized...`.
This preserves the prepared cmd `/C` command string while building the
activation command, and keeps the existing quoting path for ordinary
task arguments.
## Verification
- `cargo test -p project formats_prepared_cmd_task`
- `cargo test -p project formats_non_cmd_task_for_activation`
- `cargo check -p project`
- `cargo fmt --check --package project`
- Manually verified on Windows with a selected Python `.venv` that a
`cmd.exe` task prints `Hi there` and finishes successfully.
Release Notes:
- Fixed Windows `cmd.exe` tasks failing when run with a selected Python
virtual environment.
Follow up: https://github.com/zed-industries/zed/pull/55167,
https://github.com/zed-industries/zed/pull/54468
This is the final PR for adding remote support on the git graph. It uses
the client stream request support added in #55167 to add support for the
initial graph data request. I also fixed a bug where
`GitGraph::FullyLoaded` repository event was never emitted.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [ ] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes ##53040
Co-authored-by: Remco Smits \<djsmits12@gmail.com\>
Release Notes:
- git_graph: Add remote support
---------
Co-authored-by: Ben Kunkle <ben@zed.dev>
Lots of people are using `min-release-age` in their .npmrc files these
days.
I saw two options:
1. Force min-release-age=0 so we can always install the latest
2. Be more lenient in what we allow
I opted for 2, which means we convert `package@0.1.2` to
`package@<=0.1.2`. This means npm can find the latest version we can
that meets the user's requirements.
The downside is, the registry args/env may or may not work with the
resolved version, but that should at least surface better thanks to
#55757
There is also the issue that npm will cache package metadata and an
older version it has cached would still resolve. However, once the
metadata is updated, npm does use the newer tarball at least, so it will
update eventually.
It's a tradeoff, but I'd rather start with this until we have a better
solution on the ACP registry, rather than have users be upset becaue we
installed packages in a way they didn't want.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes
https://github.com/agentclientprotocol/claude-agent-acp/issues/516
Release Notes:
- acp: Better support min-release-age settings for npx-based agents from
the registry
* Perform grouping even for repositories that have no main worktree
* Enable grouping for remote projects
* Delete entire project groups when deleting via the recent project
picker
Release Notes:
- Fixed a bug where each linked worktree appeared as its own entry in
recent projects for repositories without main worktrees
- Fixed a bug where deleting projects from the recent projects sometimes
appeared to have no effect.
Use `future::join` when loading index and committed text for buffer diff
bases, while keeping skipped loads as ready None futures.
Release Notes:
- N/A
Signed-off-by: Xiaobo Liu <cppcoffee@gmail.com>
### Motivation
This is the second of three PRs to add remote/collab support for the git
graph and is a follow-up to #54468. I'm adding remote support for the
search because it's not user accessible without the initial graph fetch
having remote support, so it allows us to merge this without having to
add full remote support. Collab guest support will be added in a
follow-up PR.
#### Summary
For large repos, searching can take a while to fully stream in all
matched results. For example, running a basic search on the Linux repo
took over 10s for me. Because of that, we want to stream search results
in chunks to downstream users to keep the time-to-first-match low. After
this change, the first chunk gets sent back after ~50ms on the Linux
repo from receiving the request.
In order to accomplish that, I added a new proto client API that allows
for a request to map to n responses. e.g.
```/dev/null/example.rs#L1-1
client.add_entity_stream_request_handler(Self::handle_search_commits);
```
Note: The proto API isn't supported over collab yet, that will be
another PR
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- N/A
---------
Co-authored-by: cameron <cameron.studdstreet@gmail.com>
## Context
Previously the Run Debugger gutter arrow would fail silently when the
Cargo.toml had garbage lines such as “asdfasdf”. This fix makes it so
that the error is detected and bubbles up to the editor, which will
notify the user with a toast diagnostic.
Closes#46716
## Fix
https://github.com/user-attachments/assets/2e9ac7e9-1306-4607-a762-457131473572
## How to Review
Small PR - focused on four different files:
In - `crates/languages/src/rust.rs`:
- `target_info_from_abs_path()` - The function signature was changed
from `Option<(Option<TargetInfo>, Arc<Path>)>` to
`Result<Option<(Option<TargetInfo>, Arc<Path>)>>`. A condition was added
to ensure that if the Cargo metadata command is unsuccessful, the
function returns an error instead of causing an EOF error while
deserializing the stdout of the command.
- `build_context()` - Added a `?` in `target_info_from_abs_path(path,
project_env.as_ref()).await` in order to return the error.
In - `crates/project/src/task_store.rs`:
- `local_task_context_for_location()` and
`remote_task_context_for_location()` - The functions signatures were
changed from `Task<Option<TaskContext>>` to
`Task<anyhow::Result<Option<TaskContext>>>` for the purpose of
propagating the error.
In - `crates/editor/src/editor_tests.rs`:
- `build_tasks_context()` - The function signature was changed from
`Task<Option<TaskContext>>` to
`Task<anyhow::Result<Option<TaskContext>>>` .
- `toggle_code_actions()` - In case `build_tasks_context()` fails, the
functions notifies the error to the user as a Toast notification.
In - `crates/editor/src/runnables.rs`:
- Since `build_tasks_context()` and
`task_store.task_context_for_location()` now return a Result, the
callers` spawn_nearest_task() `and `task_context()` were modified. The
resulting Result types are transformed to match the expected return
types of `TaskContext` and `Task<Option<TaskContext>>`
Two new tests were added. The first, `target_info_from_abs_path_failed`
in `crates/languages/src/rust.rs`, checks if the system properly catches
the error. The second,
`test_toggle_code_actions_build_tasks_context_error_notifies` in
`crates/editor/src/editor_tests.rs`, confirms that the editor triggers
the expected error notification.
## Self-Review Checklist
- [X] I've reviewed my own diff for quality, security, and reliability
- [ ] Unsafe blocks (if any) have justifying comments
- [X] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [X] Tests cover the new/changed behavior
- [X] Performance impact has been considered and is acceptable
Release Notes:
- Fixed the error: Run Debugger failing silently due to invalid
Cargo.toml content
---------
Co-authored-by: Lukas Wirth <lukas@zed.dev>
To support branch create with base branch, added extra optional field in
GitCreateBranch proto.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes#43985
Release Notes:
- git: Fixed remote branch creation based on default branch
Fixes https://github.com/zed-industries/zed/issues/54830
This fixes a bugs where
* when there's no main worktree, we treated the first linked worktree as
main
* the titlebar and sidebar showed two different things when opening a
linked wortree directly
When there's no main worktree, our "project group key" will be the bare
repo path. For displaying this to the user, we try to present something
meaningful:
* If the bare repo is `foo.git`, we'll say "foo"
* If the bare repo is "bar/.bare", we'll "bar"
Release Notes:
- Fixed bugs in Zed's sidebar and titlebar when editing in git worktrees
created from bare repositories.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- N/A
Self-Review Checklist:
- [X] I've reviewed my own diff for quality, security, and reliability
- [ ] Unsafe blocks (if any) have justifying comments
- [ ] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [ ] Tests cover the new/changed behavior
- [X] Performance impact has been considered and is acceptable
Updated `CommitDataReader::read` from single SHA to batched
input/output.
`run_commit_data_reader` is designed to process requests in batches, but
`CommitDataReader::read` previously only sent one SHA per call. So in
practice, commit data was effectively processed one-by-one from the
caller path.
The temporary debug logs show that batching was not performed
previously.
before:
<img width="607" height="154" alt="image"
src="https://github.com/user-attachments/assets/4f9d9041-6209-426f-be44-f28449504c80"
/>
after:
<img width="632" height="121" alt="image"
src="https://github.com/user-attachments/assets/43b349f3-7685-4786-af1e-2a640cdf8ccd"
/>
Release Notes:
- N/A
---------
Co-authored-by: Anthony Eid <56899983+Anthony-Eid@users.noreply.github.com>
Co-authored-by: Anthony Eid <anthony@zed.dev>
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes#54093 (at least the os error 123 in the comments of that
bug)/#54901 (although the diagnosis in that bug is totally wrong)
Used process monitor to work out what that issue was:
<img width="1573" height="94" alt="image"
src="https://github.com/user-attachments/assets/2f82ad1a-8532-465d-9dcd-ba0bd092e9e7"
/>
There's actually a '\n' after node_modules there so it's an invalid
directory. Add trim() to fix.
After adding that change locally, eslint loaded fine
Release Notes:
- Fixed bug where eslint didn't start on Windows
The DAP TCP transport layer was hardcoded to `Ipv4Addr`, so IPv6
addresses like `fd00::a` in a debug config's `connect.host` always
failed with `hostname must be IPv4: invalid IPv4 address syntax`.
Replaced `Ipv4Addr` with `IpAddr` and `SocketAddrV4` with `SocketAddr`
across the `task`, `dap`, `dap_adapters`, and `project` crates. The WASM
extension API still uses `u32` for the host field to avoid a breaking
WIT interface change; IPv4 round-trips through extensions as before.
Fixes#52237
Release Notes:
- Fixed DAP TCP transport rejecting IPv6 addresses when connecting to
remote debug adapters.
---------
Co-authored-by: moktamd <moktamd@users.noreply.github.com>
Found this bug while investigating why configuring nextest based on the
instructions at
https://github.com/rust-lang/rust-analyzer/issues/21137#issuecomment-4254611341
wasn't working within Zed.
Previously, we'd use `serde(untagged)`, preferring cargo over shell
commands. The problem is that every instance of a shell command is a
valid instance of a cargo command. For example, the shell command:
```json
{
"label": "test my_test",
"kind": "shell",
"args": {
"environment": {"RUSTC_TOOLCHAIN": "/path/to/toolchain"},
"cwd": "/project",
"program": "cargo",
"args": ["nextest", "run", "--package", "my-crate", "--lib", "--", "my_test", "--exact", "--include-ignored"]
}
}
```
would end up getting deserialized as a Cargo command, silently dropping
`program` and `args`.
With this fix, we now use the provided `kind` as a tag. We do have to
introduce a `#[serde(flatten)]` unfortunately, which has a few side
effects due to internal buffering, but `#[serde(untagged)]` also does
internal buffering so this doesn't make things worse.
I've manually tested this by configuring:
```json
{
"lsp": {
"rust-analyzer": {
"initialization_options": {
"runnables": {
"test": {
"overrideCommand": [
"cargo",
"nextest",
"run",
"--package",
"${package}",
"${target_arg}",
"${target}",
"--",
"${test_name}",
"${exact}",
"${include_ignored}"
]
}
}
}
}
}
}
```
and ensuring that nextest is correctly invoked.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- Fixed deserialization of rust-analyzer shell runnables.
---------
Co-authored-by: Kirill Bulatov <kirill@zed.dev>
These changes update Zed's Git Panel to be able to detect unsafe git
repositories, where the user running Zed doesn't own the `.git`
directory, and show a dedicated empty view in the Git Panel that not
only explains the situation but also allows the user to choose whether
to trust this directory, which will end up running `git config --global
--add safe.directory <path>`.
While testing those changes, it was noted that attempting to add or
reset files after trusting the directory would fail, as expected, but
the UI wouldn't react to the fact that those operations failed. What
this means is, if the user tried to add a file, the UI would show a
checkmark for that file and, after the operation failed, the checkmark
would remain. We now revert that, for both `git add` and `git reset`.
Some more technical notes on this change:
- Introduce `project::git_store::GitAccess` enum to express whether we
actually have access to the repository, exposed via
`git::Repository::access`, which probes the backend with a `git status`
command and classifies the result.
- On unsafe repositories,
`project::git_store::LocalRepositoryState::new` fails to spawn the git
worker, which now cleanly signals `GitAccess::No` rather than leaving
the panel in a broken state.
- Updated `git_ui::git_panel::GitPanel::render_empty_state` with a third
alternative, when `GitAccess::No`, that basically renders a view
explaining why the repository is considered unsafe, together with
buttons for git's documentation on safe directories and a button to add
the repository's folder as a safe directory
- Updated `project::git_store::GitStore` to now watch `~/.gitconfig` and
`$XDG_CONFIG_HOME/git/config`. Watching this files allows the Git Panel
to react when the git configuration is updated, which will be the case
if the user decides to trust the repository. When either changes, a
`GitStoreEvent::GlobalConfigurationUpdated` event is emitted and the
panel refetches repository state.
- Added `project::git_store::Repository::refetch_repo_state` field,
which stores a closure to allow recreating the
`Repository::repository_state` and the job sender after the repository's
directory is trusted, without requiring a project reload.
- Added a `fs::Fs::git_config` trait method, wrapping a real `git
config` invocation. In order to be able to call this, both
`git_store::GitStore::git_config` and `project::Project::git_config`
wrappers have also been introduced. Worth mentioning that this isn't yet
supported for remote projects or collab guests.
- Updated `fs::Fs::git_clone` argument order to match
`fs::Fs::git_config` and `fs::Fs::git_init`.
- Added a new
`project::git_store::pending_op::PendingOps::last_op_errored` method
that allows determining whether the last pending operation failed. This
now allows us to filter out failed operations when determining whether
`git add` or `git reset` failed, so that we can fall through to the real
git status.
- Updated `git_ui::git_panel::GitPanel::change_file_stage` to now call
`update_counts` in its error branch so the cached staged counters stay
consistent with the reverted per-entry state, seeing as we now handle
reverting the UI state if an operation fails.
- Fixed `fs::RealFs::git_init` to fall back to the provided branch name
when `git config --global --get init.defaultBranch` fails, for example,
when the user hsan't configured one.
Co-authored-by: cameron <cameron.studdstreet@gmail.com>
Closes#42286
Release Notes:
- Added a dedicated empty state in the Git Panel for unsafe
repositories, with a "Trust Directory" button that adds the repository
to `safe.directory`
- Fixed stage and unstage checkboxes in the Git Panel not reverting when
a `git add` or `git reset` command failed
---------
Co-authored-by: cameron <cameron.studdstreet@gmail.com>
Co-authored-by: Piotr Osiewicz <24362066+osiewicz@users.noreply.github.com>
This change fixes a small bug where we were showing "Loading project..."
even when in fact we had already started the search.
It also refactors three booleans in the `SearchState` enum, so that it's
harder to make similar mistakes in the future.
Release Notes:
- N/A
