/ 
UI-KeyboardLayout-0.7901-try1
River stage zero No dependents

Changes for version 0.7901

  • Put a “Beware” in the docs about: That flag sadly doesn't affect the kernel's Alt-Numpad buffer. – On 0x04 in ToUnicode(). Replace finite automaton to finite state machine in the docs. Fix one more place mentioning 8-bit bitmasks. Add a link to history of the Alt-Numpad input. Document junk-returned-as-state after non-first prefix key in a sequence. (ligatures-resulting in a dead-key too) The documentation of the ultra-simple approach did not mention prefix keys. Document that Ctrl-2 produces U+0000 in console only — due to a bug. Correct grep⸣s misparsed with Perl⸣s before 5.14. In the docs, change “mask” to “bitmap”. In the docs, it completely slipped through the cracks that “KBDALT stripping” does not happen on key-up events. Mention related still-unresolved challenges for writing “a perfect handler”. Add references to standards on tensors. Document one layers’ assignment using lCtrl+AltGr flavors (28 of 27+1 layers for 4 personalities×2, 18 math + super/sub). Start inserting cross-references to “robust_input”.
  • examples/dumper-msg.c renamed to examples/robust_input/robust_input_front.c Attempt of a comment on treating sent messages. Many bug fixes. Add code for checking self-consistency of different approaches to “detection”. A few minutes of crazy typing on the modifiers did not lead to any detected inconsistencies. So now we can detect “real” keypresses of lCtrl, and detect KLLF_ALTGR (almost always on time). Finite state mechine for Alt-NUMPAD entering. Document 16 possible types of key input. Implement Alt-Numpad input. (Extensions: the extra digit — decimal or hex — determines: odd: cp1252, even: unicode symbols for controls+space, 0x2400 etc) Implement detection of extra-effort lCtrl and lCtrl-lAlt with the same-effect-as-AltGr. Start implementing detection of passive and active modifier keys. (Unclear what to do with no-KBDALT-stripping — e.g., when KBDKANA is latinized on Cyr…) Try to catch menu-related messages which MAY indicate that we lost focus without getting KILLFOCUS. A few of 16 cases still unsupported. Our detection of “real” lCtrl is still flaky: https://stackoverflow.com/questions/78224966/detection-of-focus-change-events-when-wm-killfocus-and-wm-setfocus-do-not-ar Allow to preceed a legacy 8-bit character code in Alt-NUMPAD input by a codepage, or a shortcut for a codepage. Implement such shortcuts 6/7/9 for Webding/Wingdings1–3/Symbol encodings (Wingdings2–3 unaccessible yet). Allow the Alt-NUMPAD unicode input made by several UTF-8 and surrogate UTF-16 codepoints: if 2 bytes, follow by 00 Detect maybe-tapAlt. (Not yet correctly acted upon. Usually see the up-of-keytap-terminating-focus-on-menu-or-icon.) Is lCtrl_first a good choice??? GetKeyState() was parsed as a BYTE. Fix all detected discrepancies (e.g., MessageBox was done before processing key-down, so sanity check in key-down was too late.) Clarify many “???”. New case lCtrl_dn_unknown. Cyrillic codepages in 40⸣s. Implement cp872 = cp855+€; cp802 = cp866+€; likewise for 848, 849. Alias 878 for KOI-8r=20866, 1168 for KOI-8u=21866. Implement patching codepages to support Ukr/Bel/Lit flavors. Un-deHex() cp=0x25. More selective handling of deHex of other “real” codepages. Mirror web/ding-bats to prefix=20–23 too (in addition to 6 and 7). Our “reserved” did not include KF_MENUMODE; Use the macro KF_ALTDOWN. F10, Esc, Alt-going-up should be in the Ctrl function!!!! ??? Alt-tap can be detected by SYS on the Alt-keyup message. SYS is disabled on VK_JUNJA. On “interrupt” messages, just set an “interrupted” flag, do not reset. Reset on the following keyboard event. This is similar to how it is done in https://referencesource.microsoft.com/#PresentationCore/Core/CSharp/System/Windows/Interop/HwndKeyboardInputProvider.cs,246 Separate tracking of lCtrl and KLLFA into subroutines. Make the preceding call conditional on a flag f_noLCtrlKLLFA. Move all MessageBox()es to the end of the pump cycle. New subclassing for the duration of the check for KLLF_ALTGR. The branch in old subclassing may be removed now. Do not try to detect KLLF_ALTGR in thread messages. Trigger lCtrl-lAlt actions only if KLLFA is known (it seems to be safer to insert a character…). New flags to control lCtrl-lAlt. Allow other handness than lCtrl-lAlt (any but lCtrl-rAlt). Separate HEX alt-Numpad input into 17th case. Control this case by a separate flag. Add recognition of ctrl-stripping and report of stripped modifiers and the base. Document (in comments) Alt-Numpad6 as an extra case of input (now 18). Reset count of active modifiers on languague change. (Cannot test due to our “custom” VK-code…) Emit reports on all bindable keys. React on WM_EXITSIZEMOVE too (as https://referencesource.microsoft.com/#PresentationCore/Core/CSharp/System/Windows/Interop/HwndKeyboardInputProvider.cs,337). Use Ctrl_state() instead of (wrong) Ctrl_state_withAlt when only the Alt-state “before” is known. Convert ctrl-string to a message ourselves. Two new options pTMf_deliverCtrlCh, pTMf_eatCtrlCh. 3-level detection of ctrl-string: at start, the first seen, the last seen. (E.g., if Alt-Ctrl-Backspace deliver ⌫, then these would be 0, ^? and ^H. Not yet merged: ∅
    • Copy of the executable at http://ilyaz.org/software/tmp/robust_input_front-0_7901.zip
  • examples/Keyboard_API.pm Minor change to error message in WriteConsole().

Documentation

UI::KeyboardLayout::izKeys
example keyboard layout generated with UI::KeyboardLayout

Modules

UI::KeyboardLayout
Module for designing keyboard layouts

Examples

Other files