Previously we did not respond to malformed or unhandled LSP requests.
The JSONRPC spec says that all non-notification requests must have
responses:
> When a rpc call is made, the Server MUST reply with a Response,
> except for in the case of Notifications
(Note that Helix is the "Server" in this case. Also from the spec:
"The Server is defined as the origin of Response objects and the
handler of Request objects.")
So this change sends error replies for requests which can't be parsed
or handled. Request IDs are also now added to the log messages for
unhandled requests.
This moves the `Application::claim_term` and
`helix-term::application::restore_term` functions into the helix-tui
crate. How the terminal should be claimed and restored is a TUI concern
and is implemented differently through different TUI backends.
This cleans out a lot of crossterm and TUI code in Application and
makes it easier to modify claim/restore based on information we query
from the terminal host. The child commit will take advantage of this
to cache the check for whether the host terminal supports the keyboard
enhancement protocol. Without this change, caching that information
takes much more code which is not easily reusable for anything else.
The code to restore the terminal is somewhat duplicated by this patch:
we want to restore the terminal in cases of panics. Panic handler hooks
must live for `'static` and the Application's terminal does not.
This refactors the snippet logic to be largely unaware of the rest of
the document. The completion application logic is moved into
generate_transaction_from_snippet which is extended to support
dynamically computing replacement text.
When accepting a snippet completion we automatically delete the
placeholders for now as doing so manual is quite cumbersome. In the
future we should keep these as a mark + virtual text that is
automatically removed once the cursor moves there.
Add a restart debug session command, which would issue a
[Restart Request][1], if the debugger supports it and a session is
running. It uses the same arguments and requests used to start the
initial session, when recreating it.
It builds upon #5532, making use of the changes to the termination
workflow of a session.
[1]: https://microsoft.github.io/debug-adapter-protocol/specification#Requests_RestartCloses: #5594
Signed-off-by: Filip Dutescu <filip.dutescu@gmail.com>
* Fix lack of space for popup crash
* Fix saturating -> wrapping
* Fix wrapping -> saturating (I am an idiot)
* Remove useless "mut" in helix-tui/src/buffer.rs
Co-authored-by: Michael Davis <mcarsondavis@gmail.com>
* Remove redundant bound-check
* Return bound-check back
* Add bound-check for set_style
* Remove set_style bound-check
* Revert bound-check
---------
Co-authored-by: Michael Davis <mcarsondavis@gmail.com>
This is a workaround for a freeze when suspending Helix with C-z on
non-Windows systems. The check for the keyboard enhancement protocol
locks up crossterm's internal event reading/polling system by trying to
set up multiple concurrent readers. `input_stream.next()` sets up one
reader looking for regular crossterm events while the
`supports_keyboard_enhancement` query sets up another looking for
internal events. The latter hangs for two seconds or until the former
yields an event. By handling signals first we don't lock up the mutex
by trying to read keyboard events.
Since crossterm 0.26.x, we receive press/release keyboard events on
Windows always. We can ignore the release events though to emulate
the behavior of keyboard input on Windows on crossterm 0.25.x.
When the Kitty Keyboard Protocol is enabled, S-backspace is
distinguished from backspace with no modifiers. This is awkward when
typing because it's very easy to accidentally hold shift and press
backspace temporarily when typing capital letters.
Kakoune (which is also a Kitty Keyboard Protocol application) treats
S-backspace as backspace too:
3150e9b3cd/src/input_handler.cc (L1275)
Check if the stack frames contain the thread id and the frame before
trying to get the frame id. If case any of the two fails to be
found, provide the user with messages to inform them of the issue and
gracefully return.
Closes: #5625
Signed-off-by: Filip Dutescu <filip.dutescu@gmail.com>
Send a `Disconnect` DAP request if the `Terminated` event is received.
According to the specification, if the debugging session was started by
as `launch`, the debuggee should be terminated alongside the session. If
instead the session was started as `attach`, it should not be disposed of.
This default behaviour can be overriden if the `supportTerminateDebuggee`
capability is supported by the adapter, through the `Disconnect` request
`terminateDebuggee` argument, as described in
[the specification][discon-spec].
This also implies saving the starting command for a debug sessions, in
order to decide which behaviour should be used, as well as validating the
capabilities of the adapter, in order to decide what the disconnect should
do.
An additional change made is handling of the `Exited` event, showing a
message if the exit code is different than `0`, for the user to be aware
off the termination failure.
[discon-spec]: https://microsoft.github.io/debug-adapter-protocol/specification#Requests_DisconnectCloses: #4674
Signed-off-by: Filip Dutescu <filip.dutescu@gmail.com>
The completion component has a separate branch for handling the
Escape key but it can use the `ignore_escape_key` helper added for
signature-help instead.
This should not cause a behavior change - it's just cleaning up the
completion component.
* feat(ui): deprecated completions
Mark deprecated completions using strike-through
(CROSSED_OUT modifier). The deprection information
is taken either from the `deprecated` field of the
completion item or from the completion tags.
The field seems to be the older way of passing
the deprecated information and it was already
marked as deprecated for Symbol. In completion
item the field is still valid but it seems that
the LSP is moving in the general direction of using
tags for this kind of information and as such
relying on tags as well seems reasonable and
future-proof.
So far LSP always required that `PositionEncoding.characters` is an
UTF-16 offset. Now that LSP 3.17 is available in `lsp-types` request
the server to send char offsets (UTF-32) or byte offsets (UTF-8)
instead. For compatability with old servers, UTF-16 remains as the
fallback as required by the standard.
* Make `m` textobject look for pairs enclosing selections
Right now, this textobject only looks for pairs that surround the
cursor. This ensures that the pair found encloses each selection, which
is likely to be intuitively what is expected of this textobject.
* Simplification of match code
Co-authored-by: Michael Davis <mcarsondavis@gmail.com>
* Adjust logic for ensuring surround range encloses selection
Prior, it was missing the case where the start of the selection came
before the opening brace. We also had an off-by-one error where if the
end of the selection was on the closing brace it would not work.
* Refactor to search for the open pair specifically to avoid edge cases
* Adjust wording of autoinfo to reflect new functionality
* Implement tests for surround functionality in new integration style
* Fix handling of skip values
* Fix out of bounds error
* Add `ma` version of tests
* Fix formatting of tests
* Reduce indentation levels for readability, and update comments
* Preserve each selection's direction with enclosing pair surround
* Add test case for multiple cursors resulting in overlap
* Mark known failures as TODO
* Make tests multi-threaded or they fail
* Cargo fmt
* Fix typos in integration test comments
---------
Co-authored-by: Michael Davis <mcarsondavis@gmail.com>
Example:
```
test
testitem
```
Select line 2 with x, then type Alt-C; Helix will go into an infinite
loop. The saturating_sub keeps the head_row and anchor_row pinned at 0,
and a selection is never made since the first line is too short.
This matches the behavior from 42ad1a9e04
but for the first and last change. The selection rules are the same
as for goto_next/prev_change: additions and modifications select the
added and modified range while deletions are represented with a point.
This will allow testing more of the code base, as well as enable UI-
specific testing.
Debug mode builds are prohibitively slow for the tests, mostly
because of the concurrency write tests. So there is now a profile for
integration tests that sets the optimization level to 2 for a few helix
crates, and lowers the number of rounds of concurrent writes to 1000.
