Fixes#1356
As stated in the code, the specification itself doesn't explicitly
specify this behavior (as far as I can tell), but testing in Kitty
results in this working and I believe this is how they do it.
This ensures that the start or end x values do not get adjusted when
they are off the viewport in rectangle select.
This also includes some tests for the toViewport method.
Fixes#1339.
Fixes#1343
If a wide character is found at the right edge of a terminal screen and
can't be printed without wrapping, the wide character is ignored. This
matches xterm behavior.
The one funky behavior is with grapheme clustering enabled and VS16
emoji. For VS16, we act as if VS16 was never received. This is not
specified in any way and no other terminal handles this correctly at the
time of authoring this so we're just making this up because it seems
most sensible.
When calling resize, Ghostty should be respecting the wraparound state.
This behavior matches xterm. Note that this same bug was also found in
kitty.
Fixes: #1343
- Output to stdin instead of a file
- Less nesting
- Utilize ranged for loops instead of while loops
- Eliminate unnecessary state tracking
- Put help in a struct
Doc strings attached to fields of the Config struct and doc strings
attached to the run function of actions will be used to generate Zig
code that makes those doc strings available to be used at runtime.
Based on PR #853 by @Raiden1411
Fixes#1329
Some shells and scripts use spaces and soft-wrapping as a way to move to
the next line instead of using newline (`\n`). Line selection
(triple-click by default) considers a soft-wrapped line as a single
line, so it was selecting the prompt.
This commit makes it so line selection considers semantic prompt state
(prompt vs command output) an additional boundary condition. This
requires shell integration but will make selection behave more
expectedly.
