This zero-width char will force the shaper to treat detected RTL
lines as LTR.
RTL text would still be rendered correctly. But this fixes the wrong
behavior of it being displayed aligned to the right.
Signed-off-by: Mohammad AlSaleh <CE.Mohammad.AlSaleh@gmail.com>
This causes a functional change that may not be obvious.
When running `cosmic-term` from a shell prompt as a detached
process, `stdout`/`stderr` descriptors get deleted, and a `println!()`
message would take the `cosmic-term` process down.
No such an issue with log messages.
This can be tested by running `htop` then exiting the tool, which
would trigger a "TODO: CursorBlinkingChange" message.
Signed-off-by: Mohammad AlSaleh <CE.Mohammad.AlSaleh@gmail.com>
This can happen if configured font name is uninstalled for example.
Pick the first font name available if that's the case.
Signed-off-by: Mohammad AlSaleh <CE.Mohammad.AlSaleh@gmail.com>
* Store a font name => font faces map
* Only list font names that have `NORMAL` and `BOLD` faces with
`Normal` stretch. This will give us defaults and fall-backs that are
guaranteed to always exist.
* Filter by stretch first, with `Normal` chosen as the always existing
default.
* Then only list font weights supported by the font name stretch
selected, for the normal and bold cases.
* When changing the font name selected, the stretch/weight options
will stay the same if supported by the new font name, or revert to
the always existing fall-backs.
Signed-off-by: Mohammad AlSaleh <CE.Mohammad.AlSaleh@gmail.com>
In `Application::subscription()`, `subscription::channel()` would
create a dead terminal tab if it was called before
`config_subscription()`.
Then `subscription::channel()` would be triggered a second time
creating a second terminal tab that works.
This commit fixes this issue by only creating the first terminal
tab after a `Message::Config{}` was handled, i.e. after
`config_subscription()` was run.
Signed-off-by: Mohammad AlSaleh <CE.Mohammad.AlSaleh@gmail.com>
Fixes#13, shell history, and many more.
All this code is probably ripe for some refactors and higher
abstractions. But this is good enough for now.
Signed-off-by: Mohammad AlSaleh <CE.Mohammad.AlSaleh@gmail.com>
