Cherry-picked Kb/wasm panics (#35319)
Follow-up of https://github.com/zed-industries/zed/pull/34208
Closes https://github.com/zed-industries/zed/issues/35185
Previous code assumed that extensions' language server wrappers may leak
only in static data (e.g. fields that were not cleared on deinit), but
we seem to have a race that breaks this assumption.
1. We do clean `all_lsp_adapters` field after
https://github.com/zed-industries/zed/pull/34334 and it's called for
every extension that is unregistered.
2. `LspStore::maintain_workspace_config` ->
`LspStore::refresh_workspace_configurations` chain is triggered
independently, apparently on `ToolchainStoreEvent::ToolchainActivated`
event which means somewhere behind there's potentially a Python code
that gets executed to activate the toolchian, making
`refresh_workspace_configurations` start timings unpredictable.
3. Seems that toolchain activation overlaps with plugin reload, as
`2025-07-28T12:16:19+03:00 INFO [extension_host] extensions updated.
loading 0, reloading 1, unloading 0` suggests in the issue logs.
The plugin reload seem to happen faster than workspace configuration
refresh in
c65da547c9/crates/project/src/lsp_store.rs (L7426-L7456)
as the language servers are just starting and take extra time to respond
to the notification.
At least one of the `.clone()`d `adapter`s there is the adapter that got
removed during plugin reload and has its channel closed, which causes a
panic later.
----------------------------
A good fix would be to re-architect the workspace refresh approach, same
as other accesses to the language server collections.
One way could be to use `Weak`-based structures instead, as definitely
the extension server data belongs to extension, not the `LspStore`.
This is quite a large undertaking near the extension core though, so is
not done yet.
Currently, to stop the excessive panics, no more `.expect` is done on
the channel result, as indeed, it now can be closed very dynamically.
This will result in more errors (and backtraces, presumably) printed in
the logs and no panics.
More logging and comments are added, and workspace querying is replaced
to the concurrent one: no need to wait until a previous server had
processed the notification to send the same to the next one.
Release Notes:
- Fixed warm-related panic happening during startup
Co-authored-by: Kirill Bulatov <kirill@zed.dev>
Cherry-picked project_panel: Fix autoscroll to treat entries behind
sticky items as out of viewport (#35067)
Closes#34831
Autoscroll centers items only if they’re out of viewport. Before this
PR, entry behind sticky items was not considered out of viewport, and
hence actions like `reveal in project panel` or focusing buffer would
not autoscroll that entry into the view in that case.
This PR fixes that by using recently added `scroll_to_item_with_offset`
in https://github.com/zed-industries/zed/pull/35064.
Release Notes:
- Fixed issue where `pane: reveal in project panel` action was not
working if the entry was behind sticky items.
Co-authored-by: Smit Barmase <heysmitbarmase@gmail.com>
Cherry-picked gpui: Add `scroll_to_item_with_offset` to
`UniformListScrollState` (#35064)
Previously we had `ScrollStrategy::ToPosition(usize)` which lets you
define the offset where you want to scroll that item to. This is the
same as `ScrollStrategy::Top` but imagine some space reserved at the
top.
This PR removes `ScrollStrategy::ToPosition` in favor of
`scroll_to_item_with_offset` which is the method to do the same. The
reason to add this method is that now not just `ScrollStrategy::Top` but
`ScrollStrategy::Center` can also uses this offset to center the item in
the remaining unreserved space.
```rs
// Before
scroll_handle.scroll_to_item(index, ScrollStrategy::ToPosition(offset));
// After
scroll_handle.scroll_to_item_with_offset(index, ScrollStrategy::Top, offset);
// New! Centers item skipping first x items
scroll_handle.scroll_to_item_with_offset(index, ScrollStrategy::Center, offset);
```
This will be useful for follow up PR.
Release Notes:
- N/A
Co-authored-by: Smit Barmase <heysmitbarmase@gmail.com>
Cherry-picked client: Send `User-Agent` header on WebSocket connection
requests (#35280)
This PR makes it so we send the `User-Agent` header on the WebSocket
connection requests when connecting to Collab.
We use the user agent set on the parent HTTP client.
Release Notes:
- N/A
Co-authored-by: Marshall Bowers <git@maxdeviant.com>
This fixes an issue where we were not setting the context server working
directory at all.
Release Notes:
- Context servers will now be spawned in the currently active project
root.
---------
Co-authored-by: Danilo Leal <danilo@zed.dev>
Cherry-picked keymap_ui: Fix bug introduced in #35208 (#35237)
Closes #ISSUE
Fixes a bug that was cherry picked onto stable and preview branches
introduced in #35208 whereby modifier keys would show up and not be
removable when editing a keybind
Release Notes:
- (preview only) Keymap Editor: Fixed an issue introduced in v0.197.2
whereby modifier keys would show up and not be removable while recording
keystrokes in the keybind edit modal
Co-authored-by: Ben Kunkle <ben@zed.dev>
Now the edit tool can access files outside the current project (just
like the terminal tool can), but it's behind a prompt (unlike other edit
tool actions).
Release Notes:
- The edit tool can now access files outside the current project, but
only if the user grants it permission to.
Cherry-picked keymap_ui: Additional keystroke input polish (#35208)
Closes #ISSUE
Fixed various issues and improved UX around the keystroke input
primarily when used for keystroke search.
Release Notes:
- Keymap Editor: FIxed an issue where the modifiers used to activate
keystroke search would appear in the keystroke search
- Keymap Editor: Made it possible to search for repeat modifiers, such
as a binding with `cmd-shift cmd`
- Keymap Editor: Made keystroke search matches match based on ordered
(not necessarily contiguous) runs. For example, searching for `cmd
shift-j` will match `cmd-k cmd-shift-j alt-q` and `cmd-i g shift-j` but
not `alt-k shift-j` or `cmd-k alt-j`
- Keymap Editor: Fixed the clear keystrokes binding (`delete` by
default) not working in the keystroke input
Co-authored-by: Ben Kunkle <ben@zed.dev>
Cherry-picked git: Touch up amend UX (#35114)
Follow-up to #26114
- Ensure that the previous commit message is filled in when toggling on
amend mode from the context menu
- Fix keybinding flicker in context menu
Release Notes:
- N/A
Co-authored-by: Cole Miller <cole@zed.dev>
Discovered in this issue: #34513
Previously, we were propagating deserialization errors to users when
using LMStudio, instead of the actual error message sent from the
LMStudio server. This change will help users understand why their
request failed while streaming responses.
Release Notes:
- lmsudio: Display specific backend error messaging on failure rather
than generic ones
---------
Signed-off-by: Umesh Yadav <git@umesh.dev>
Co-authored-by: Peter Tripp <peter@zed.dev>
In project panel, `rename` and `duplicate` action further needs user
input for editing, so if panel is closed we should open it.
Release Notes:
- Fixed project panel not opening when `project panel: rename` and
`project panel: duplicate` actions are triggered from workspace.
Someone encountered this in production, which should not happen:
<img width="1266" height="623" alt="Screenshot 2025-07-24 at 10 38
40 AM"
src="https://github.com/user-attachments/assets/40f3f977-5110-4808-a456-7e708d953b3b"
/>
This moves certain errors into the category of "never retry" and reduces
the number of retries for some others. Also it adds some diagnostic
logging for retry policy.
It's not a complete fix for the above, because the underlying issues is
that the server is sending a HTTP 403 response and although we were
already treating 403s as "do not retry" it was deciding to retry with 2
attempts anyway. So further debugging is needed to figure out why it
wasn't going down the 403 branch by the time the request got here.
Release Notes:
- N/A
There were two different underlying reasons for the issues with
ctrl-number and ctrl-punctuation:
1. Some keys in the ctrl-0..9 range send codes in the `\1b`..`\1f`
range. For example, `ctrl-2` sends keycode for `ctrl-[` (0x1b), but we
want to map it to `2`, not to `[`.
2. `ctrl-[` and four other ctrl-punctuation were incorrectly mapped,
since the expected conversion is by adding 0x40
Closes#35012
Release Notes:
- N/A
Mistral released their new DevstralMedium model to be used via API:
https://mistral.ai/news/devstral-2507
Release Notes:
- Add support for Mistral Devstral Medium
This PR splits the resize logic into separate left/right propagation
methods and improve code organization around column width adjustments.
It also allows resize to work for both the left and right sides as well,
instead of only checking the right side for room
Release Notes:
- N/A *or* Added/Fixed/Improved ...
---------
Co-authored-by: Ben Kunkle <ben@zed.dev>
This change dims rows in the keymap editor for which the corresponding
keybind is overridden by other keybinds coming from higher priority
sources.
Release Notes:
- N/A
This PR adds resizable columns to the keymap editor and the ability to
double-click on a resizable column to set a column back to its default
size.
The table uses a column's width to calculate what position it should be
laid out at. So `column[i]` x position is calculated by the summation of
`column[..i]`. When resizing `column[i]`, `column[i+1]`’s size is
adjusted to keep all columns’ relative positions the same. If
`column[i+1]` is at its minimum size, we keep seeking to the right to
find a column with space left to take.
An improvement to resizing behavior and double-clicking could be made by
checking both column ranges `0..i-1` and `i+1..COLS`, since only one
range of columns is checked for resize capacity.
Release Notes:
- N/A
---------
Co-authored-by: Anthony <anthony@zed.dev>
Co-authored-by: Remco Smits <djsmits12@gmail.com>
As suggested in https://github.com/zed-industries/zed/pull/34418, this
proposes various changes to language configs to make block comments and
doc-block-style comments more similar. In doing so, it introduces some
breaking changes into the extension schema.
This change is needed to support the changes I'm working on in #34418,
to be able to support `rewrap` in block comments like `/* really long
comment ... */`. As is, we can do this in C-style doc-block comments (eg
`/** ... */`) because of the config in `documentation`, but we can't do
this in regular block comments because we lack the info about what the
line prefix and indentation should be.
And while I was here, I did various other clean-ups, many of which feel
nice but are optional.
I would love special attention on the changes to the schema, version and
related changes; I'm totally unfamiliar with that part of Zed.
**Summary of changes**
- break: changes type of `block_comment` to same type as
`documentation_comment` (**this is the important change**)
- break: rename `documentation` to `documentation_comment` (optional,
but improves consistency w/ `line_comments` and `block_comment`)
- break/refactor?: removes some whitespace in the declaration of
`block_comment` delimiters (optional, may break things, need input; some
langs had no spaces, others did)
- refactor: change `tab_size` from `NonZeroU32` to just a `u32` (some
block comments don't seem to need/want indent past the initial
delimiter, so we need this be 0 sometimes)
- refactor: moves the `documentation_comment` declarations to appear
next to `block_comment`, rearranges the order of the fields in the TOML
for `documentation_comment`, rename backing `struct` (all optional)
**Future scope**
I believe that this will also allow us to extend regular block comments
on newline – as we do doc-block comments – but I haven't looked into
this yet. (eg, in JS try pressing enter in both of these: `/* */` and
`/** */`; the latter should extend w/ a `*` prefixed line, while the
former does not.)
Release Notes:
- BREAKING CHANGE: update extension schema version from 1 to 2, change
format of `block_comment` and rename `documentation_comment`
/cc @smitbarmase
Change channel reorganization (move up/down) from `cmd-up/down` (mac) /
`ctrl-up/down` (linux) to `alt-up/down` (both) to match moving lines in
the editor.
Also fix an issue where if you selected channels using down/up in the
filter field, the movement shortcuts would not work (`editing` vs
`not_editing`).
Release Notes:
- N/A
Quick follow up to #34896 which ensures that the Agent Panel cannot be
caught by actions like `workspace: toggle left dock` when `disable_ai`
is set to true.
Also removes a method that was introduced but unused in the workspace
because `first_enabled_panel_idx` already covers all cases this method
could be useful for.
Release Notes:
- N/A
This commit adds an example with deep children hierarchy.
The depth of a tree can be tweaked with GPUI_TREE_DEPTH env variable.
With depth=100
<img width="301" height="330" alt="image"
src="https://github.com/user-attachments/assets/844cd285-c5f3-4410-a74e-981bf093ba2e"
/>
With this example, I can trigger a stack overflow at depth=633 (and
higher).
Release Notes:
- N/A
Collect the iterator instead of manually looping over it to utilize
possible size hints. Zed usually passes in owned `Vec`'s, meaning we get
to reuse memory as well.
Release Notes:
- N/A
we were join(",") and split(",") to serialize the patterns.
This doesn't work when pattern includes a ","
example: *.{ts,tsx} (very common pattern used by agent)
help needed:
how will this work on version mismatch?
Release Notes:
- Fixed search filter patterns on remote projects.
https://github.com/user-attachments/assets/d472fbdd-7736-4bd7-8a90-8cca356b2815
This PR adds `editor: diff clipboard with selection` - good for spotting
the differences in eerily-similar code, which is when refactoring code,
as you need to see what needs to be passed in in order to maintain
previous behavior of both snippets.
1. Copy some text from anywhere
2. Highlight some text in Zed
3. Run `editor: diff clipboard with selection`
Like JetBrains' IDEs and VS Code with the `PartialDiff` package, if the
selection is empty, we take the entire buffer as the selection.
Caveats:
- We do not know the language of the text in the clipboard. I went ahead
and just assumed that in most cases, it will be the same language as the
selected text, which does mean we will highlight the old text
incorrectly if they are copying from a different language, but I think
in most cases, it will be the same, and the alternative of always having
no syntax highlighting is worse. PyCharm seems to do the same thing.
Release Notes:
- Added an `editor: diff clipboard with selection` action
---------
Co-authored-by: Junkui Zhang <364772080@qq.com>
Co-authored-by: Ben Kunkle <ben@zed.dev>
Closes#7759
While opening a new SSH project if we are reusing an existing window,
i.e. drop the existing workspace and create a new one, then before
dropping it, we should remove `session_id` from that workspace's
serialized entry. That way:
1. Upon closing (cmd-w) this remote workspace (which also clears
`session_id` for this), and then quitting. No workspace with that
`session_id` is found, and it starts fresh.
2. Upon directly quitting (cmd-q) this remote workspace, only this
workspace exists in db (among two of them) with that `session_id`, and
it restores correctly.
Release Notes:
- Fixed an issue while closing remote workspace restores last local
workspace on startup.
