The updates all the apis so that the basekeyboardid or klid
can be passed in as an argument. This in needed so that
elevated process required to compile the keyboard has
the call users keyboard base id.
The updates all the apis so that the basekeyboardid or klid
can be passed in as an argument. This in needed so that
elevated process required to compile the keyboard has
the call users keyboard base id.
* Remove references to `koSwitchLanguageForAllApplications` and
`koAutoSwitchOSKPages` options, translated strings for them, and
related `GlobalKeyboardChangeManager` -- no longer required since
Win8.
* Remove Windows Vista-, 7-, and 8-specific code paths - except for the
one user-facing setting for installing keyboards on downlevel
operating systems.
* Remove commented code
Fixes: #16433
Test-bot: skip
Cleanup of unused `wm_keyman_control` and `wm_keyman_refresh` values,
and some other unused messages, which were making the implementation
harder to understand.
Relates-to: #15333
Follows: #16435
Clean up the keyboard refresh process. Keyman COM API will post a
`wm_keyman_refresh`, `KR_REQUEST_REFRESH` message when changes have been
made to the list of installed keyboards (or some other keyboard-related
settings change).
Keyman Engine (any arch) will handle this message, in any process, and
post a message to the master controller (keyman.exe, TfrmKeyman7Main),
`wm_keyman_refresh`, `KR_PRE_REFRESH`.
Keyman Engine (keyman32) will handle _this_ message in the master
controller thread context, and increment the global refresh tag. It also
asks the other architecture host process(es) to update their global
refresh counter to the same value (by posting `wm_keyman_refresh`,
`KM_PRE_REFRESH` to those process(es).)
This refresh tag is then compared on-demand in each thread to the
thread's current refresh tag value, when a keyboard activation is
received, on set focus, or on a keystroke, and the keyboard list is
refreshed at that point.
This means that keyboard installation no longer triggers a large amount
of activity reloading all the keyboards on all windowed threads, but
rather they will gradually reload as the user interacts with them. It
also fixes#15333, which related to having threads with only child
windows -- that is, that were parented by top-level windows from other
threads -- which thus never received the global broadcast in the earlier
design. For example, VSCode, Notepad in Win11, Chrome.
Fixes: #15333
The Caps Lock and Num Lock flags in Globals::ShiftState() are updated
only as key events pass through the engine, so they go stale when the
toggle is changed while a non-Keyman layout is active. The existing
resync in GetCapsAndNumlockState() runs on window focus change, which a
keyboard switch does not trigger, so the core processor was told Caps
Lock was off until the user toggled it twice.
Extract the toggle resync as RefreshToggleState() and call it from
TIPActivateKeyboard(), which is the profile activation path for both
Win+Space and the Windows language selector.
Fixes#16422
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011GdFpFi5eqJP6C1YjuQytY
What we was needed was the ability to recompile
installed keyboards against a basekeyboard id.
MCompileForBaseKeyboard which will take the
base keyboard id. This allows the call to be done
elevated and seperates out the actuall selecting
the base keyboard.
kmc-convert provides tooling to convert between various common keyboard
description file formats. Each conversion will be implemented in its
single, separate module (and each module will done in its own PR)
This PR will address the conversion **keylayout → kmn**
Co-authored-by: Marc Durdin <marc@durdin.net>
# Keyman Conventional Commit suggestions:
#
# - Link to a Sentry issue with git trailer:
# Fixes: _MODULE_-_ID_
# - Give credit to co-authors:
# Co-authored-by: _Name_ <_email_>
# - Use imperative, present tense ('attach' not 'attaches', 'attached' etc)
# - Don't include a period at the end of the title
# - Always include a blank line before trailers
# - More: https://github.com/keymanapp/keyman/wiki/Pull-Request-and-Commit-workflow-notes
Copy the handle of the icon from kmcomapi rather than assigning, so that
we do not end up calling DestroyIcon twice on the same handle (once by
FTrayIcon destruction, and once by kmcomapi).
Co-authored-by: Ross Cruickshank <rc-swag@users.noreply.github.com>
Quietly handle invalid font files while installing, and cleanly abort
install of broken packages. Report the errors to Sentry but do not
crash.
Fixes: #12812
Fixes: KEYMAN-WINDOWS-45K
Fixes: KEYMAN-WINDOWS-6S8
Fixes: KEYMAN-WINDOWS-430
Fixes: KEYMAN-WINDOWS-7T3
Fixes: KEYMAN-WINDOWS-7SY
Fixes: KEYMAN-WINDOWS-42Z
When the user clicks either Shift key on the OSK, that emits a VK_LSHIFT
down key event. If the user subsequently presses and releases the Right
Shift key on their hardware keyboard, that does not reset the isDown
state for the Left Shift key, so the key is not released on the OSK, and
the state becomes inconsistent for apps also.
This change emits a key up event, if when Right Shift is released,
Keyman detects that Left Shift is still "down". Note that if the user
presses and holds Left Shift, and presses and releases Right Shift, the
left shift state will be reset before the users releases Left Shift. But
that's a pathological state!
Fixes: #12611
