SIGHUP correctly states the intention behind sending a signal when a
pane is closed: The controlling terminal is "hung up".
Also, SIGHUP is better suited than SIGTERM since bash ignores the
latter. This led to the zombie processes observed by some users.
Furthermore, SIGHUP has a special meaning in bash's job control, namely
re-sending the signal to all owned jobs before exiting.
* feat(sessions): mirrored sessions
* fix(tests): input units
* style(fmt): make rustfmt happy
* fix(tests): make mirrored sessions e2e test more robust
* refactor(sessions): remove force attach
* style(fmt): rustfmtify
* docs(changelog): update change
* fix(e2e): retry on all errors
* feat(cwd-pane): Add a new trait to get the cwd of a given pid
Co-authored-by: Quentin Rasmont <qrasmont@gmail.com>
* feat(cwd-pane): Allow for setting the cwd when spawning a new terminal
Co-authored-by: Quentin Rasmont <qrasmont@gmail.com>
* feat(cwd-pane): Add an active_pane field to the Pty struct
Co-authored-by: Quentin Rasmont <qrasmont@gmail.com>
* feat(cwd-pane): Update Pty with Tab's active pane id
Co-authored-by: Quentin Rasmont <qrasmont@gmail.com>
* feat(cwd-pane): Refactor spawn_terminal to use cwd by default
Co-authored-by: Quentin Rasmont <qrasmont@gmail.com>
* feat(cwd-pane): Fix tests and lints
Co-authored-by: Quentin Rasmont <qrasmont@gmail.com>
* feat(cwd-pane): Fix formatting
* feat(cwd-pane): Refactor child pid fetching to handle errors better
Instead of panicking when transfering the process id of the forked child
command we just return an empty process id.
* feat(cwd-pane): Add non Linux/MacOS targets for the get_cwd method.
This will allow Zellij to compile on non Linux/MacOS targets without
having an inherited cwd.
* feat(cwd-pane): Refactor spawn_terminal method to use ChildId struct.
The spawn_terminal methods been refactored to use the ChildId struct in
order to clarify what the Pid's returned by it are. The documentation
for the ChildId struct was improved as well.
* feat(cwd-pane): Fix tests/lints
Co-authored-by: Jesse Tuchsen <not@disclosing>
Co-authored-by: Quentin Rasmont <qrasmont@gmail.com>
* refactor(panes): move to parametric pane sizes
* Fixed the simpler errors by casting to usize
* The least I can do is pass the formatting check...
* Move to stable toolchain
* Well, it compiles?
* And now it doesn't! ;)
* Baseline functionality with the new Dimension type
* Working POC for percent-based resizing
* REVERT THIS COMMIT – DELETES TESTS
* Perfected the discrete resize algorithm
* Fixed fixed-size panes
* Basic bidirectional resize
* feat(resize): finalised parametric resize algorithm
* Reduce the logging level a bit
* Fixed nested layouts using percents
* Bug squishing for implicit sizing
* Here is a funky (read: rubbish) rounding approach
* And now it's gone again!
* Improve discretisation algorithm to fix rounding errors
* Fix the last layout bug (maybe?)
* Mixed explicit and implied percents work now
* Let's pretend that didn't happen...
* Make things a bit less crashy
* Crash slightly more for now (to find bugs)
* Manaually splitting of panes works now
* Start moving to percent-based resizes
* Everything but fullscreen seems to be working
* Fix compilatation errors
* Culled a massive amount of border code
* Why not pause to please rustfmt?
* Turns out I was still missing a few tests...
* Bringing back even more tests!
* Fix tests and pane boarders
* Fix the resize system without gaps
* Fix content offset
* Fixed a bug with pane closing
* Add a hack to fix setting of the viewport
* Fix toggling between shared borders and frames
* fix(tests): make e2e properly use PaneGeom
* style(fmt): make rustfmt happy
* Revert unintentional rounding of borders
* Purge some old borderless stuff
* Fix busted tab-bar shrinking
* Update E2E tests
* Finish implementing fullscreen!
* Don't crash anymore?
* Fix (almost) all tests
* Fix a lack of tab-stops
* All tests passing
* I really can't be bothered to debug a CI issue
* Tie up loose ends
* Knock out some lingering FIXMEs
* Continue to clean things up
* Change some naming and address FIXMEs
* Cull more code + FIXMEs
* Refactor of the resize system + polish
* Only draw frames when absolutely necessary
* Fix the tab-bar crash
* Fix rendering of boarders on reattach
* Fix resizing at small pane sizes
* Deduplicate code in the layout system
* Update tab-bar WASM
* Fixed the pinching of panes during resize
* Unexpose needlessly public type
* Add back a lost test
* Re-add tab tests and get them to compile
* All tabs need layouts
* Start fixing tests + bug in main
* Stabilize the resize algorithm rounding
* All tests from main are now passing
* Cull more dead code
* fix unexpected pane closing issue with nushell
nushell doesn't create a new process group when spawnning a process,
so all processes including the ones spwanned by us are in the same
foreground group. So if kernel will send signal to every member of this
group.
This patch fixes this issue by creating a new foreground process group
when spawnning a new terminal pane.
Fix#615
Signed-off-by: Tw <tw19881113@gmail.com>
* add comment about unsafe
Co-authored-by: Aram Drevekenin <aram@poor.dev>
* refactor(os_input_output): use Pid for child process
* fix(debug): change debug_to_file to write &[u8]
This patch changes debug_to_file to write entire slices in one call, to
avoid the overhead of opening and closing files for each byte written.
* refactor(ServerOsApi): remove unnecessary muts from methods
Co-authored-by: KOVACS Tamas <ktamas@fastmail.fm>
