⭐ Football Update: microsoft/terminal/pull/

Leonard Hecker (lhecker) · 2026-04-02T11:02:25Z

NOTE:
This commit changes our input model to treat all
Ctrl-Alt combinations as an AltGr modifier by default.
Ctrl+Alt+/ will not produce "\x1b\x1f" anymore.

This fixes the following issues:

Base-key report of non-ASCII keys was broken
AltGr on international keyboards would spam
Ctrl/Alt up/down events repeatedly forever
KKP doesn't report modifier key repeats
AltGr+Key combinations were reported as Ctrl+Alt+key
but should've been just plain Key-But-Modified

Closes #19977

Validation Steps Performed

This cannot be easily tested in unit tests, due to
the dependence on international keyboard layouts.
It appears to work correctly from what I can tell.

Leonard Hecker (lhecker) · 2026-04-02T11:06:54Z

    }

    // Keep track of key repeats.
-    key.keyRepeat = _lastVirtualKeyCode == key.virtualKey;


Placing the detection outside a if (key.keyDown) block meant that a single key up/down would already be detected as a key repeat.

Leonard Hecker (lhecker) · 2026-04-02T11:07:09Z

+        if (const auto flags = _controlKeyStateFromVirtualKey(key.virtualKey, key.controlKeyState))
+        {
+            key.keyRepeat = (_previousControlKeyState & flags) != 0;
+        }


These 3 lines fix the AltGr key repeat.

Leonard Hecker (lhecker) · 2026-04-02T11:10:25Z

+    key.altGrPressed = anyAltPressed && anyCtrlPressed;
+    key.ctrlPressed = key.altGrPressed ? (bothCtrlPressed || (rightAltPressed && leftCtrlIsReallyPressed)) : anyCtrlPressed;
+    key.altPressed = key.altGrPressed ? bothAltPressed : anyAltPressed;
+    key.shiftPressed = WI_IsFlagSet(key.controlKeyState, SHIFT_PRESSED);


To make reliable AltGr detection possible throughout the code, I've made a bit of a more controversial change and removed the entire key.controlKeyState reliance throughout the code.

This crucially means that even the legacy key encoding will now treat AltGr distinct from other modifiers.

Leonard Hecker (lhecker) · 2026-04-02T11:11:42Z

    // NOTE: MSDN states that `lpSrcStr == lpDestStr` is valid for LCMAP_LOWERCASE.
-    len = LCMapStringW(LOCALE_INVARIANT, LCMAP_LOWERCASE, &buf[0], len, &buf[0], ARRAYSIZE(buf));
-    // NOTE: LCMapStringW returns the length including the null terminator.
-    len -= 1;


len -= 1 was wrong and only applies to encoding null-terminated strings. I have switched to a length-based buffer a long time ago and missed fixing this. 🙈

Leonard Hecker (lhecker) · 2026-04-02T11:12:23Z

+    // In the context of KKP, AltGr acts more like a keyboard "layer" toggle.
+    // It's not a modifier that's ever transmitted as-is and instead modifies the actual base key code.
+    if (key.altGrPressed)
+    {
+        controlKeyState |= LEFT_CTRL_PRESSED | RIGHT_ALT_PRESSED;
+    }


This new if condition ensures we treat AltGr like a layer shift that changes the base encoding of the key.

Leonard Hecker (lhecker) · 2026-04-03T16:19:25Z

-        { L"D Shift+Alt+a", L"\x1b[97;4u", D, true, 'A', 0x1E, L'A', Shift | Alt },
-        { L"D Shift+a", L"A", D, true, 'A', 0x1E, L'A', Shift },
+        { L"D Ctrl+a", L"\x1b[97;5u", D, true, 'A', 0x10, 0, Ctrl },
+        { L"D Ctrl+Alt+a", L"æ", D, true, 'A', 0x10, L'æ', Ctrl | Alt },


This has been rewritten to use French AZERTY events.

all of them ??? is that safe? ... are we guaranteed that it exists on OneCore?

should we have tests in US-104 for the most part and allow a TestCase to specify a keyboard layout to set?

all of them ??? is that safe?

Yes and yes.

are we guaranteed that it exists on OneCore?

No. But all of this code relies on APIs that are categorically unavailable in OneCore, so it probably doesn't matter.

should we have tests in US-104 for the most part and allow a TestCase to specify a keyboard layout to set?

That's how it works now, via SetTerminalInputKeyboardLayout.

Dustin L. Howett (DHowett) · 2026-04-03T17:18:55Z

    Log::Comment(L"Sending every possible VKEY at the input stream for interception during key DOWN.");

-    BYTE vkey = LOBYTE(OneCoreSafeVkKeyScanW(0));
+    BYTE vkey = LOBYTE(VkKeyScanExW(0, layout));


We still need to use OneCoreSaveVkKeyScanW; this test will fail to run (or, maybe the entire fixture will fail to load?) if we link VkKeyScanW.

But it depends on LoadKeyboardLayout and ToUnicodeEx now. Won't all of these fail to work anyway? Do they have to work in the first? The TerminalInput logic is platform-independent with the exception of a single ToUnicodeEx call.

Okay, so here's what I have come to understand.

We refer to these APIs using an extension APIset.

ext-ms-win-ntuser-keyboard-l1-1-0 GetKeyState GetKeyboardLayout MapVirtualKeyW ToUnicodeEx VkKeyScanW

OneCoreSafeEtc runs those through coniosrv or equivalent rather than ntuser.
In theory, we could have done some effort to compose onecore with a new extension apiset that fills the same contract. Instead we used runtime dynamic dispatch. Eh.

As long as the implementation calls OneCoreSafeEtc, we're fine.

We will need to prevent the kitty tests from running on OneCore, so that we do not experience test gate failures. :)

I have added exception guards around SetTerminalInputKeyboardLayout. However, I have not checked if it results in linking failures in OneCore. I'm somewhat confused how this works in general given that we have never defined a OneCoreSafe wrapper for ToUnicodeEx.

Leonard Hecker (lhecker) · 2026-04-03T18:36:11Z

It is rather annoying that the /alternatename linker thing doesn't seem to work for x86.
And some of the KeyPressTests don't seem to work in CI which is extraordinarily annoying. This may mean that we have to either disable the KKP tests or remove them.

Dustin L. Howett (DHowett) · 2026-04-03T19:41:37Z

It is rather annoying that the /alternatename linker thing doesn't seem to work for x86. And some of the KeyPressTests don't seem to work in CI which is extraordinarily annoying. This may mean that we have to either disable the KKP tests or remove them.

Unfortunately, the deal with TestHook is that the no-op implementation must be in a separate .obj in the LIB-under-test. It cannot be an alternatename or some other weak definition kind of deal.

This type of test hook relies on the classic model of linking where objects are only pulled in even from libs if they are referred to for an unresolved symbol. TerminalInput.lib!TestHook.obj must be able to be skipped wholesale because it does not contribute an unresolved symbol to the test DLL.

Dustin L. Howett (DHowett) · 2026-04-03T19:42:10Z

(Whereas you have TestHook.obj in the UT project; you rather need TestHook.obj in TerminalInput which contains the no-op implementation; then you can implement the hook in the UT in whatever object file you want, including the test file that relies on the overridden hook)

…-altgr

Dustin L. Howett (DHowett)

OK, I think I understand this now. Thanks for being patient. Let's get it in Canary and see what happens.

NOTE: This commit changes our input model to treat all Ctrl-Alt combinations as an AltGr modifier by default. Ctrl+Alt+/ will not produce `"\x1b\x1f"` anymore. This fixes the following issues: * Base-key report of non-ASCII keys was broken * AltGr on international keyboards would spam Ctrl/Alt up/down events repeatedly forever * KKP doesn't report modifier key repeats * AltGr+Key combinations were reported as Ctrl+Alt+key but should've been just plain Key-But-Modified Closes #19977 ## Validation Steps Performed This cannot be easily tested in unit tests, due to the dependence on international keyboard layouts. It appears to work correctly from what I can tell. (cherry picked from commit fbda436) Service-Card-Id: PVTI_lADOAF3p4s4BQX0-zg6CmdI Service-Version: 1.25

Windows Terminal >= 1.24 aliases every Ctrl+Alt combo to AltGr (microsoft/terminal#20052), so the legacy "ESC + control byte" pair that matchesKey() looks for (ctrl+alt+r = \x1b\x12) is never emitted there and the keystroke dies inside the terminal. WSL panes inherit that input translation, so they need the same treatment; pi's own useWindowsKeybindings() draws the line in the same place (it is not exported from the package root, so the check is mirrored). - lib/shortcuts.ts: one place for the three shortcuts (t/n/r). Windows-like platforms register ctrl+alt+<letter> and alt+<letter>; everyone else keeps ctrl+alt only (macOS: plain Option composes characters). The cheatsheet label is derived from the key id that is actually registered, so a tab can no longer advertise a key the platform never binds; on macOS it renders "Ctrl+option+r" like pi's own formatter does. - extensions/section-headers.ts registers every variant through pi.registerShortcut; extensions/footer.ts matches through the same object. Both build it once at module scope (matches() runs per keystroke). - /reload still intercepts handleInput on purpose: ExtensionContext.reload() exists in the typings but the context handed to shortcut handlers does not implement it. - README: platform notes (Windows/WSL, macOS, Linux GNOME owning ctrl+alt+t) and PI_TUI_ESC_TIMEOUT for the intermittent "Escape + stray letter" glitch — pi flushes a lone ESC after ~10 ms, which splits ESC-prefixed shortcuts on slow or multiplexed PTYs (verified: 20 ms gap -> ["\x1b","r"], Escape fired; 100 ms -> ["\x1br"], shortcut fired). Verified: npm test 10/10 (incl. new lib/shortcuts.ts self-check), platform matrix forced for win32 / WSL / linux / darwin, live registration probe ["ctrl+alt+t","alt+t"] and cheatsheet [Alt+t]/[Alt+n]/[Alt+r] on win32.

Fix various KKP encoding issues

446f0df

github-advanced-security AI found potential problems Apr 2, 2026

View reviewed changes

Comment thread src/terminal/input/terminalInput.cpp Fixed

Comment thread src/terminal/input/terminalInput.cpp Fixed

Comment thread src/terminal/input/terminalInput.cpp Fixed

Comment thread src/terminal/input/terminalInput.cpp Fixed

This comment has been minimized.

Sign in to view

Spel

e27f76d

Leonard Hecker (lhecker) commented Apr 2, 2026

View reviewed changes

Leonard Hecker (lhecker) added 3 commits April 3, 2026 18:16

Rethought how AltGr works

f0e3432

Fixed TerminalInputModifierKeyTests

fcda6e9

Added test hooks, Use French AZERTY for tests

0e4d6b8

Leonard Hecker (lhecker) commented Apr 3, 2026

View reviewed changes

github-advanced-security AI found potential problems Apr 3, 2026

View reviewed changes

This comment has been minimized.

Sign in to view

SPELL CHECK

552a79f

Dustin L. Howett (DHowett) reviewed Apr 3, 2026

View reviewed changes

Leonard Hecker (lhecker) added 2 commits April 17, 2026 00:30

Merge remote-tracking branch 'origin/main' into dev/lhecker/19977-kkp…

5141679

…-altgr

Fix compilation failures?

6e3bb0e

This comment has been minimized.

Sign in to view

Leonard Hecker (lhecker) added 2 commits April 17, 2026 20:32

Fixed SetTerminalInputKeyboardLayout

63b6bba

Merge remote-tracking branch 'origin/main' into dev/lhecker/19977-kkp…

9354b5c

…-altgr

Dustin L. Howett (DHowett) mentioned this pull request Apr 29, 2026

[TQ: re-check] Strange behavior on the PT keyboard in Windows 11 #20146

Closed

Can Celik (ogulcancelik) mentioned this pull request May 21, 2026

Swedish layout: å/ä/ö (and likely other Level 1/3 keys) duplicated when typed inside herdr panes under Windows Terminal Preview herdrdev/herdr#239

Closed

Dustin L. Howett (DHowett) approved these changes Jul 31, 2026

View reviewed changes

Dustin L. Howett (DHowett) merged commit fbda436 into main Jul 31, 2026
20 checks passed

Dustin L. Howett (DHowett) deleted the dev/lhecker/19977-kkp-altgr branch July 31, 2026 23:14

Dustin L. Howett (DHowett) mentioned this pull request Jul 31, 2026

Fix AltGr-composed characters dropped under the Kitty keyboard protocol #20361

Closed

1 of 4 tasks

Dustin L. Howett (DHowett) added this to 1.25 Servicing Pipeline Sep 8, 2026

github-project-automation Bot moved this to To Cherry Pick in 1.25 Servicing Pipeline Sep 8, 2026

Dustin L. Howett (DHowett) moved this from To Cherry Pick to Cherry Picked in 1.25 Servicing Pipeline Sep 8, 2026

HvC (H4vC) mentioned this pull request Sep 23, 2026

fix(tui): recovered AltGr text dropped by Windows console kitty encoding can1357/oh-my-pi#12984

Merged

3 tasks done

JJ Liebig (JJLiebig) mentioned this pull request Sep 23, 2026

Windows Terminal Preview (Kitty keyboard protocol): AltGr characters on Finnish layout are delivered as ESC+key to every pane herdrdev/herdr#4543

Closed

2 tasks done

Conversation

Leonard Hecker (lhecker) commented Apr 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Validation Steps Performed

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

This comment has been minimized.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Leonard Hecker (lhecker) Apr 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

This comment has been minimized.

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Leonard Hecker (lhecker) commented Apr 3, 2026

Uh oh!

Dustin L. Howett (DHowett) commented Apr 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Dustin L. Howett (DHowett) commented Apr 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment has been minimized.

Dustin L. Howett (DHowett) left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Leonard Hecker (lhecker) commented Apr 2, 2026 •

edited

Loading

Leonard Hecker (lhecker) Apr 17, 2026 •

edited

Loading

Dustin L. Howett (DHowett) commented Apr 3, 2026 •

edited

Loading

Dustin L. Howett (DHowett) commented Apr 3, 2026 •

edited

Loading