I noticed a busy period when closing the connui when it is performing
the sleep with progress bar stage; returning from the function when
the channel disconnects seems like the right call.
I also made the loop slightly more efficient by only repainting if
the line has changed.
I don't really like this as a solution, however: I'd rather
spend time transitioning towards async IO for this instead of
troubleshooting further here.
This commit doesn't do anything specific to scrollback though!
It moves the implementation of the TermWizTermTab away from
a directly manipulated Surface and over to using the term::Terminal,
making the renderer look more like the one used by the local tab
and domain implementation.
As a side effect of doing this, we get scrollback management
for free.
refs: https://github.com/wez/wezterm/issues/201
I noticed that we still weren't quite right.
I think the net change is really just that we must force
the use of the CSI representation when using CTRL/SHIFT
modifiers.
refs: https://github.com/wez/wezterm/issues/203
Otherwise, during early startup we'll trigger one before the frontend
is started and show a toast message, and then later on we'll also
show the wezterm error window, which feels a bit redundant.
This way we reserve the toasts for panics and will generally
use the wezterm window.
This is a bit fiddly because the serde library doesn't let us know
eg: which key in a map had an error, so we have to show the information
from the value and hope that there is enough context for the user.
This commit adds our own Debug impl for ValueWrapper as that one
provided by mlua isn't useful for tables; our impl will iterate
the pairs in the table and pretty print them.
The struct error message has been adjusted to include the lua
value that failed to deserialize in addition to the type name.
Because the config structs can get quite large, this commit adjusts
the ordering of the components in the error message; we now print
the outermost error first and leave the most specific portion at
the bottom. This is so that the error window is more likely to
show the most useful information at the bottom.
I've added some similar improvements when processing enums.
It's not perfect but hopefully more useful when figuring out
a config problem.
This invalid config:
```lua
local wezterm = require 'wezterm';
return {
keys = {
{key="{", mods="SHIFT|ALT", action="Search"},
}
}
```
yields:
```
While (re)loading configuration: Error converting lua value returned by script /tmp/issue-201.lua to Config struct: while processing a struct of type `Config` with value:
{
"keys": [
{
"mods": "SHIFT|ALT",
"key": "{",
"action": "Search",
},
],
}
while processing a struct of type `Key` with value:
{
"mods": "SHIFT|ALT",
"key": "{",
"action": "Search",
}
while processing an enum of type `KeyAssignment` and value "Search"
which has allowed variants `ActivateCopyMode`, `ActivateTab`, `ActivateTabRelative`, `ClearScrollback`, `CloseCurrentTab`, `CompleteSelection`, `CompleteSelectionOrOpenLinkAtMouseCursor`, `Copy`, `DecreaseFontSize`, `DisableDefaultAssignment`, `ExtendSelectionToMouseCursor`, `Hide`, `HideApplication`, `IncreaseFontSize`, `MoveTab`, `MoveTabRelative`, `Nop`, `OpenLinkAtMouseCursor`, `Paste`, `PastePrimarySelection`, `QuitApplication`, `ReloadConfiguration`, `ResetFontSize`, `ScrollByPage`, `Search`, `SelectTextAtMouseCursor`, `SendString`, `Show`, `ShowLauncher`, `ShowTabNavigator`, `SpawnCommandInNewTab`, `SpawnCommandInNewWindow`, `SpawnTab`, `SpawnWindow`, `ToggleFullScreen`
Expected a variant with parameters but got a unit variant instead
```
refs: https://github.com/wez/wezterm/issues/201
We were treating DECCKM as the sole thing to enable application
cursor reporting, but looking closely at the docs, that mode only
takes effect when both DECANM (Vt52 emulation mode) AND DECKPAM
(application keypad mode) are both active.
neovim enables DECCKM and DECKPAM but not DECANM.
refs: https://github.com/wez/wezterm/issues/203
The captured variable used to track the last cursor time wasn't
working as intended; rather than reflecting the time of the last
blink update it was always reading as the time at which the window
was created which meant that we would choose to invalidate the
window on every candidate frame (every 35ms) rather than every
cursor blink interval, burning some CPU (~4-5% continuous on my system).
This was hard to spot because it looked like it was doing the right
thing. It didn't show up as a hotspot in the profiler either :-p
This commit moves that state into the TermWindow itself; I see CPU
utilization drop to ~0.7% which is much better.
refs: https://github.com/wez/wezterm/issues/200
