Hi,
It has been a while since the last release. Here are some highlights
on this release. This release has some important bug fixes.
fcitx5 5.1.10
New candidate action API, planned to be used for mac/android port.
There was an issue between fcitx 5.1.9 and the wayland compositor that
uses zwp_input_method_v2.
Hyprland & river already has upstream fix for it, while I got no reply
from sway swaywm/sway#8143 .
While I prefer to keep the change, I don't want to leave users in a
state that can't use the input method at all, so the relevant change
was reverted in 5.1.10.
There is also a crash due to a bug on fcitx's wayland frontend. The
crash may be triggered when the system is slow.
There is a design flaw in StandardPath::multiOpen API that may open
too many files and exhaust the fd that it can open. Alternative API is
implemented and engines are adapted to the new API in this release.
fcitx5-bamboo 1.0.6
Icon name is renamed to avoid XDG icon path fallback on breeze theme.
fcitx5-chewing 5.1.4
Added Colemak layout support provided by libchewing
Workaround some issue with rust libchewing
fcitx5-chinese-addons 5.1.6
Fix compose key that need AltGr
Add support to use spell-check candidate into the sentence
hide cangjie by default.
fcitx5-configtool 5.1.6
directly launch external tool if ExternalOption is the only option
fcitx5-hangul 5.1.4
Fix typing when caps lock is enabled.
fcitx5-kkc 5.1.4
Rename icons to avoid XDG icon name fallback with breeze icon theme
fcitx5-rime 5.1.7
Add compose support
Fix a bug with combo pinyin with certain preedit mode
fcitx5-skk 5.1.4
Rename icons to avoid XDG icon name fallback with breeze icon theme
fcitx5-table-extra 5.1.6
Rename icons to avoid XDG icon name fallback with breeze icon theme
fcitx5-table-other 5.1.3
Rename icons to avoid XDG icon name fallback with breeze icon theme
libime 1.1.8
Add pinyin like jun -> jvn to fuzzy support
Fix a bug that may produce duplicate pinyin prediction
Fix pinyin parsing when a certain fuzzy flag is not enabled.
Change the word learning mechanism, now all length<=4 characters input
will be learned as a user phrase.