Screen readers expect closing a modal dialog to reannounce the previous focus.
This makes the exit confirm dialog more modal in this sense: it will unfocus
the layout and then focus it back when closed, giving the desired behavior.
* Add per-axis scroll speed config for input devices.
Accepts negative values to inverse scroll direction.
Properly complements/overrides global `scroll-direction` setting.
Includes docs and tests.
* Update per-axis scroll factor implementation after testing
- Refined configuration structure in niri-config
- Updated input handling to use per-axis scroll factors
- Added comprehensive test snapshots
- Updated documentation with per-axis examples
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Simplify per-axis scroll factor implementation per review feedback
- Make documentation concise and clear
- Remove unnecessary comments and test helper functions
- Use inline snapshots for tests
- Rename get_factors() to h_v_factors() for clarity
- Remove unnecessary .clone() calls (ScrollFactor is Copy)
- Reduce test count to essential cases only
- Fix comment about window factor override behavior
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Remove unnecessary ScrollFactor::new() helper function
The maintainer prefers minimal code, so removing this helper and
constructing ScrollFactor directly in tests.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Fix scroll factor behavior - all settings now multiply with window factor
Per maintainer feedback, both combined and per-axis scroll settings
should multiply with the window-specific scroll factor, not override it.
This ensures consistent behavior regardless of configuration method.
Also removed the now-unused has_per_axis_override() method.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Final cleanup: remove redundant comments and unused snapshot files
- Removed unused snapshot files (now using inline snapshots)
- Removed redundant inline comments in tests
- Simplified test descriptions to be more concise
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Convert scroll factor parsing tests to use assert_debug_snapshot
Updates parse_scroll_factor_combined, parse_scroll_factor_split, and
parse_scroll_factor_partial tests to use assert_debug_snapshot instead
of manual assert_eq comparisons, as requested in PR review.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Convert to inline snapshots as requested
- Convert all scroll factor parsing tests to use inline snapshots instead of external files
- Remove external snapshot files to keep test directory clean
- All tests still pass with inline snapshot assertions
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Fix missed assert_eq in parse_scroll_factor_mixed test
Converts the remaining assert_eq calls to assert_debug_snapshot
with inline snapshots in the mixed syntax test function.
Also fixes raw string delimiters from ### to #.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* Convert scroll_factor_h_v_factors test to use assert_debug_snapshot
Makes all scroll factor tests consistent by using snapshots instead
of assert_eq for better maintainability.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* fixes
---------
Co-authored-by: Bernardo Kuri <github@bkuri.com>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Ivan Molodetskikh <yalterz@gmail.com>
* feat: config reload ipc event
* cleanups
* Rename and move the new config option
* rename to ConfigLoaded and emit at connection
---------
Co-authored-by: Ivan Molodetskikh <yalterz@gmail.com>
* Add window sizes and positions to the IPC
* basic fixes
* report window_loc instead of window pos
* clean ups
* make scrolling indices 1-based
* add printing to niri msg windows
* don't include render offset in floating tile pos
---------
Co-authored-by: Ivan Molodetskikh <yalterz@gmail.com>
To be used for animation tests that need animation snapshots.
The renderer is optional to avoid creating it thousands of times in tests when
it's not needed, plus it can deadlock in mesa apparently.
* niri-config: add disable-set-bpc option
setting bpc to 8 bricks some OLED displays driven by amdgpu
* change to keep-max-bpc-unchanged and add to wiki
* fmt
* Update wiki/Configuration:-Debug-Options.md
---------
Co-authored-by: Ivan Molodetskikh <yalterz@gmail.com>
* refactor config load logic, and properly watch the system config path
* move config creation to niri-config, and make the errors a bit nicer
notably, "error creating config" is now a cause for "error loading
config", instead of it being one error and then "error loading config:
no such file or directory". also, failure to load a config is now
printed as an error level diagnostic (because it is indeed an error, not
just a warning you can shrug off)
* refactor watcher tests; add some new ones
now they check for the file contents too! and i added some tests for
ConfigPath::Regular, including a messy one with many symlink swaps
* fixes
---------
Co-authored-by: Ivan Molodetskikh <yalterz@gmail.com>
Without explicit sync, we have no way to signal the PipeWire consumer when the
rendering is done. So, wait until it's done before giving it the frame.
This should fix flickering screencasts on NVIDIA.
We're gonna need to store these in the future, and the lifetime on the pw-rs
Buffer prevents us from easily doing that. Besides, we'll need access to
metadata which pw-rs doesn't expose yet.
Ivan Molodetskikh