fix(core): handle backspace decomposition by mcdurdin · Pull Request #15488 · keymanapp/keyman

mcdurdin · 2026-01-24T20:29:31Z

Ensure that when a single backspace decomposes the last NFC character in the app context, the remainder of the 'cluster' is preserved, matching the implication of the CLDR keyboard specification.

This addresses the behavior in #15487 where the cached context became out of sync with the app context after deleting an entire NFC cluster such as ê, which caused a loop ending up with the entire context being deleted, at which point the loop exited with a fail-safe.

Note that the LDML keyboard tests (ldml.cpp) do not currently exercise the normalization code; this is a gap that should be addressed to ensure that we are testing final application behavior.

Fixes: #15487
Build-bot: release:windows,linux,mac
.

User Testing

Preparation: install the keyboard bksp_ldml.kmp from the archive below.

bksp_ldml.zip

For these tests, always type the following key sequences:

E-ACUTE: abc e/
D-ACUTE: abc d/
A-UMLAUT: abc '

GROUP_WINDOWS: Test on Windows 11
GROUP_MAC: Test on macOS
GROUP_LINUX: Test on Linux

TEST_COMPLIANT_E_ACUTE_VERIFY_CONTEXT: In a compliant app, type the E-ACUTE sequence. Select the text, copy and paste it into a character viewer and verify that the last character is U+00E9.
TEST_COMPLIANT_E_ACUTE: In a compliant app, type the E-ACUTE sequence. Then press bksp. The result should be abc e.
TEST_NONCOMPLIANT_E_ACUTE_VERIFY_CONTEXT: In a non-compliant app, type the E-ACUTE sequence. Select the text, copy and paste it into a character viewer and verify that the last character is U+00E9.
TEST_NONCOMPLIANT_E_ACUTE: In a non-compliant app, type the E-ACUTE sequence. Then press bksp. The result should be abc e.

TEST_COMPLIANT_D_ACUTE_VERIFY_CONTEXT: In a compliant app, type the D-ACUTE sequence. Select the text, copy and paste it into a character viewer and verify that the last two characters are U+0064 U+0301.
TEST_COMPLIANT_D_ACUTE: In a compliant app, type the D-ACUTE sequence. Then press bksp. The result should be abc d.
TEST_NONCOMPLIANT_D_ACUTE_VERIFY_CONTEXT: In a non-compliant app, type the D-ACUTE sequence. Select the text, copy and paste it into a character viewer and verify that the last two characters are U+0064 U+0301.
TEST_NONCOMPLIANT_D_ACUTE: In a non-compliant app, type the D-ACUTE sequence. Then press bksp. The result should be abc d.

TEST_COMPLIANT_A_UMLAUT_VERIFY_CONTEXT: In a compliant app, type the A-UMLAUT sequence. Select the text, copy and paste it into a character viewer and verify that the last character is U+00E4.
TEST_COMPLIANT_A_UMLAUT: In a compliant app, type the A-UMLAUT sequence. Then press bksp. The result should be abc a.
TEST_NONCOMPLIANT_A_UMLAUT_VERIFY_CONTEXT: In a non-compliant app, type the A-UMLAUT sequence. Select the text, copy and paste it into a character viewer and verify that the last character is U+00E4.
TEST_NONCOMPLIANT_A_UMLAUT: In a non-compliant app, type the A-UMLAUT sequence. Then press bksp. The result should be abc a.

Note on compliant and non-compliant apps

Windows 11: Some compliant apps are Notepad, Word, Firefox. Some non-compliant apps are Text Editor (artifact on this PR), Chrome.
macOS: One compliant app is TextEdit. One non-compliant app is Chrome.
Linux: One compliant app is gEdit. One non-compliant app is Chrome.

Ensure that when a single backspace decomposes the last NFC character in the app context, the remainder of the 'cluster' is preserved, matching the implication of the CLDR keyboard specification. This addresses the behavior in #15487 where the cached context became out of sync with the app context after deleting an entire NFC cluster such as ê, which caused a loop ending up with the entire context being deleted, at which point the loop exited with a fail-safe. Note that the LDML keyboard tests (ldml.cpp) do not currently exercise the normalization code; this is a gap that should be addressed to ensure that we are testing final application behavior. Fixes: #15487

keymanapp-test-bot · 2026-01-24T20:29:36Z

User Test Results

Test specification and instructions

✅ GROUP_WINDOWS: Test on Windows 11
12 tests PASSED
- ✅ TEST_COMPLIANT_E_ACUTE_VERIFY_CONTEXT (PASSED) (notes)
- ✅ TEST_COMPLIANT_E_ACUTE (PASSED) (notes)
- ✅ TEST_NONCOMPLIANT_E_ACUTE_VERIFY_CONTEXT (PASSED) (notes)
- ✅ TEST_NONCOMPLIANT_E_ACUTE (PASSED) (notes)
- ✅ TEST_COMPLIANT_D_ACUTE_VERIFY_CONTEXT (PASSED) (notes)
- ✅ TEST_COMPLIANT_D_ACUTE (PASSED) (notes)
- ✅ TEST_NONCOMPLIANT_D_ACUTE_VERIFY_CONTEXT (PASSED) (notes)
- ✅ TEST_NONCOMPLIANT_D_ACUTE (PASSED) (notes)
- ✅ TEST_COMPLIANT_A_UMLAUT_VERIFY_CONTEXT (PASSED) (notes)
- ✅ TEST_COMPLIANT_A_UMLAUT (PASSED) (notes)
- ✅ TEST_NONCOMPLIANT_A_UMLAUT_VERIFY_CONTEXT (PASSED) (notes)
- ✅ TEST_NONCOMPLIANT_A_UMLAUT (PASSED) (notes)
🟩 GROUP_MAC: Test on macOS
- ✅ TEST_COMPLIANT_E_ACUTE_VERIFY_CONTEXT (PASSED) (notes)
- ✅ TEST_COMPLIANT_E_ACUTE (PASSED) (notes)
- ✅ TEST_NONCOMPLIANT_E_ACUTE_VERIFY_CONTEXT (PASSED) (notes)
- 🟩 TEST_NONCOMPLIANT_E_ACUTE (SKIPPED): A separate issue was encountered on macOS which blocked testing: bug(mac): backspace does not follow rules #15543
- ✅ TEST_COMPLIANT_D_ACUTE_VERIFY_CONTEXT (PASSED) (notes)
- 🟩 TEST_COMPLIANT_D_ACUTE (SKIPPED): A separate issue was encountered on macOS which blocked testing: bug(mac): backspace does not follow rules #15543
- ✅ TEST_NONCOMPLIANT_D_ACUTE_VERIFY_CONTEXT (PASSED) (notes)
- ✅ TEST_NONCOMPLIANT_D_ACUTE (PASSED) (notes)
- ✅ TEST_COMPLIANT_A_UMLAUT_VERIFY_CONTEXT (PASSED) (notes)
- ✅ TEST_COMPLIANT_A_UMLAUT (PASSED) (notes)
- ✅ TEST_NONCOMPLIANT_A_UMLAUT_VERIFY_CONTEXT (PASSED) (notes)
- 🟩 TEST_NONCOMPLIANT_A_UMLAUT (SKIPPED): A separate issue was encountered on macOS which blocked testing: bug(mac): backspace does not follow rules #15543
✅ GROUP_LINUX: Test on Linux
12 tests PASSED
- ✅ TEST_COMPLIANT_E_ACUTE_VERIFY_CONTEXT (PASSED): verified ok with Text Editor
- ✅ TEST_COMPLIANT_E_ACUTE (PASSED): no problems encountered
- ✅ TEST_NONCOMPLIANT_E_ACUTE_VERIFY_CONTEXT (PASSED): verified ok with Chrome
- ✅ TEST_NONCOMPLIANT_E_ACUTE (PASSED): no problems encountered
- ✅ TEST_COMPLIANT_D_ACUTE_VERIFY_CONTEXT (PASSED): verified ok with Text Editor
- ✅ TEST_COMPLIANT_D_ACUTE (PASSED): no problems encountered
- ✅ TEST_NONCOMPLIANT_D_ACUTE_VERIFY_CONTEXT (PASSED): verified ok with Chrome
- ✅ TEST_NONCOMPLIANT_D_ACUTE (PASSED): no problems encountered
- ✅ TEST_COMPLIANT_A_UMLAUT_VERIFY_CONTEXT (PASSED): verified ok with Text Editor
- ✅ TEST_COMPLIANT_A_UMLAUT (PASSED): no problems encountered
- ✅ TEST_NONCOMPLIANT_A_UMLAUT_VERIFY_CONTEXT (PASSED): verified ok with Chrome
- ✅ TEST_NONCOMPLIANT_A_UMLAUT (PASSED): no problems encountered (notes)

Test Artifacts

Developer
Keyboards
- Test Keyboards
macOS
- Keyman for macOS
- Keyman for macOS (old PRs)
Web
- KeymanWeb Test Home
Windows

mcdurdin · 2026-01-25T20:17:27Z

core/src/actions_normalize.cpp

Fix is entirely in this file; other changes are additional unit tests and enabling correct logging of keystrokes in the unit test log files.

mcdurdin · 2026-01-25T20:19:30Z

@rc-swag @sgschantz @ermshiperete given this change is a significant fix to Keyman Core, I would like to request review from all of you. @srl295, feel free to ignore; just as you have capacity to take a look.

mcdurdin · 2026-01-25T21:11:08Z

core/tests/unit/ldml/keyboards/k_213_backspace_decomposition.xml

+@@expected:  abce
+Comment:     #15487: bksp will delete just acute (even though it will have combined to NFC, internally we do NFD bksp)
+
+TODO: this seems to be working NFD on app context side also, do we need to set a flag? this may break other tests so care is needed


TODO: this seems to be working NFD on app context side also, do we need to set a flag? this may break other tests so care is needed.

Ensure that when a single backspace decomposes the last NFC character in the app context, the remainder of the 'cluster' is preserved, matching the implication of the CLDR keyboard specification. This addresses the behavior in #15487 where the cached context became out of sync with the app context after deleting an entire NFC cluster such as ê, which caused a loop ending up with the entire context being deleted, at which point the loop exited with a fail-safe. Note that the LDML keyboard tests (ldml.cpp) do not currently exercise the normalization code; this is a gap that should be addressed to ensure that we are testing final application behavior. This cherry-pick only adds the direct normalization patch and unit test and skips the additional changes for logging. Fixes: #15487 Cherry-pick-of: #15488

srl295 · 2026-01-26T18:34:36Z

core/src/actions_normalize.cpp

+    if(app_context_nfd == cached_context_string.substr(0, app_context_nfd.length())) {
+      context_final = cached_context_string.substr(app_context_nfd.length());


maybe want some more comments here.
What we're trying to say is, if the app context matches the beginning of the cache,
break but remember if there was any content that was added after the cache.

So context_final will now have any 'suffix' that is NOT reflected in the app context.

srl295 · 2026-01-26T18:36:10Z

core/src/actions_normalize.cpp

  */

-  auto output_nfc = output;
+  auto output_nfc = context_final + output;


Here, any of that 'final' suffix that was NOT already in the app context, needs to be prepended to the output.

so the full picture before the insertion point is:

<app_context><context_final><output>|

srl295 · 2026-01-26T18:38:11Z

looks good! i think it could be good to add some comments and even put the case into the comments (the example case) to document what context_final is doing. but that could be a later step.

Co-authored-by: srl295@gmail.com

sgschantz · 2026-01-27T20:58:29Z

@mcdurdin For the E-ACUTE test cases, shouldn't the final unicode value be U+00E9?

mcdurdin · 2026-01-27T22:05:22Z

@mcdurdin For the E-ACUTE test cases, shouldn't the final unicode value be U+00E9?

Thank you, good catch -- I had started with A-ACUTE... but then wanted to differentiate.

When normalizing, we need to stop processing on an NFC boundary, not an NFD boundary, to support normalizations such as in Bengali, where appending `U+09D7` to a context of `U+0995 U+09C7` should result in `U+0995 U+09CC`. The specification is unclear on this; see https://unicode-org.atlassian.net/browse/CLDR-19218 This also updates the ldml keyboard unit test suite to support running in full NFC mode (used in all Engine implementations) as well retaining the NFD mode (now only used by the debugger). Side note: the Bengali normalization failure case was picked up by the improvements to the unit test suite, proving once again that good tests are so valuable. Fixes: #15491 Fixes: #15505 Follows: #15488 Relates-to: CLDR-19218

core/tests/unit/ldml/ldml.cpp

darcywong00 · 2026-01-28T03:59:10Z

core/tests/unit/kmnkbd/actions_normalize.tests.cpp

+  test_actions_normalize(
+    "One backspace to delete last NFD character (#15487)",
+    /* app context pre transform: */     u"abcê", // NFC
+    /* cached context post transform: */ u"abce",


If the app context pre transform is the same as l.327 (both have NFC \u00EA),
shouldn't cached context post transform match l.328? (u"abce\u0323\u0302")

No, because the action line is different, for l.319, we have no output (U""), but for l.330, we have U"\u0323\u0302"

Meng-Heng · 2026-01-29T06:03:24Z

Prerequisite

Uninstall active Keyman
Install Keyman 19.0.193-alpha-test-15488
Install the bksp_ldml keyboard
Character view: https://unicode.scarfboy.com/
Sequences:
- E-ACUTE: abc e/
- D-ACUTE: abc d/
- A-UMLAUT: abc '

GROUP_MAC:

Test Specs

MacBook Intel macOS Sequoia
TextEdit v1.20
Chrome v144.0.7559.110

TEST_COMPLIANT_E_ACUTE_VERIFY_CONTEXT (PASSED):

In TextEdit
Type the E-ACUTE sequence
Paste the last character into the character viewer
Verified: The last character is “é” (U+00E9).

TEST_COMPLIANT_E_ACUTE (PASSED):

In TextEdit
Type the E-ACUTE sequence
Press backspace
Verified: The result is abc e.

TEST_NONCOMPLIANT_E_ACUTE_VERIFY_CONTEXT (PASSED):

In Chrome, Google docs
Type the E-ACUTE sequence
Paste the last character into the character viewer
Verified: The last character is “é” (U+00E9)

TEST_NONCOMPLIANT_E_ACUTE (FAILED):

In Chrome, Google docs
Type the E-ACUTE sequence
Press backspace
Failed: The result is abce.

TEST_COMPLIANT_D_ACUTE_VERIFY_CONTEXT (PASSED):

In TextEdit
Type the D-ACUTE sequence
Paste d́ into the character viewer
Verified: The last two characters' unicode points are U+0064 U+0301.

TEST_COMPLIANT_D_ACUTE (FAILED):

In TextEdit
Type the D-ACUTE sequence
Press backspace
Failed: the result is abc .

TEST_NONCOMPLIANT_D_ACUTE_VERIFY_CONTEXT (PASSED):

In Chrome, Google docs
Type the D-ACUTE sequence
Paste d́ into the character viewer
Verified: The last two characters' unicode points are U+0064 U+0301.

TEST_NONCOMPLIANT_D_ACUTE (PASSED):

In Chrome, Google docs
Type the D-ACUTE sequence
Press backspace
Verified: the result is abc d.

TEST_COMPLIANT_A_UMLAUT_VERIFY_CONTEXT (PASSED):

In TextEdit
Type the A-UMLAUT sequence
Paste ä into the character viewer
Verified: The last character’s unicode point is U+00E4.

TEST_COMPLIANT_A_UMLAUT (PASSED):

In TextEdit
Type the A-UMLAUT sequence
Press backspace
Verified: The result is abc a.

TEST_NONCOMPLIANT_A_UMLAUT_VERIFY_CONTEXT (PASSED):

In Chrome, Google docs
Type the A-UMLAUT sequence
Paste ä into the character viewer
Verified: The last character’s unicode point is U+00E4.

TEST_NONCOMPLIANT_A_UMLAUT (FAILED):

In Chrome, Google docs
Type the A-UMLAUT sequence
Press backspace
Failed: The result is abca.

GROUP_WINDOWS:

Test Specs

macOS Virtualbox
Windows 11 AMD64
Notepad v11.2510
Chrome v144.0.7559.110

TEST_COMPLIANT_E_ACUTE_VERIFY_CONTEXT (PASSED):

In Notepad
Type the E-ACUTE sequence
Paste the last character into the character viewer
Verified: The last character is “é” (U+00E9)

TEST_COMPLIANT_E_ACUTE (PASSED):

In Notepad
Type the E-ACUTE sequence
Press backspace
Verified: The result is abc e.

TEST_NONCOMPLIANT_E_ACUTE_VERIFY_CONTEXT (PASSED):

In Chrome, Google docs
Type the E-ACUTE sequence
Paste the last character into the character viewer
Verified: The last character is “é” (U+00E9)

TEST_NONCOMPLIANT_E_ACUTE (PASSED):

In Chrome, Google docs
Type the E-ACUTE sequence
Press backspace
Verified: The result is abc e.

TEST_COMPLIANT_D_ACUTE_VERIFY_CONTEXT (PASSED):

In Notepad
Type the D-ACUTE sequence
Paste d́ into the character viewer
Verified: The last two characters' unicode points are U+0064 U+0301.

TEST_COMPLIANT_D_ACUTE (PASSED):

In Notepad
Type the D-ACUTE sequence
Press backspace
Verified: the result is abc d.

TEST_NONCOMPLIANT_D_ACUTE_VERIFY_CONTEXT (PASSED):

In Chrome, Google docs
Type the D-ACUTE sequence
Paste d́ into the character viewer
Verified: The last two characters' unicode points are U+0064 U+0301.

TEST_NONCOMPLIANT_D_ACUTE (PASSED):

In Chrome, Google docs
Type the D-ACUTE sequence
Press backspace
Verified: the result is abc d.

TEST_COMPLIANT_A_UMLAUT_VERIFY_CONTEXT (PASSED):

In Notepad
Type the A-UMLAUT sequence
Paste ä into the character viewer
Verified: The last character’s unicode point is U+00E4.

TEST_COMPLIANT_A_UMLAUT (PASSED):

In Notepad
Type the A-UMLAUT sequence
Press backspace
Verified: The result is abc a.

TEST_NONCOMPLIANT_A_UMLAUT_VERIFY_CONTEXT (PASSED):

In Chrome, Google docs
Type the A-UMLAUT sequence
Paste ä into the character viewer
Verified: The last character’s unicode point is U+00E4.

TEST_NONCOMPLIANT_A_UMLAUT (PASSED):

In Chrome, Google docs
Type the A-UMLAUT sequence
Press backspace
Verified: The result is abc a.

mcdurdin · 2026-01-30T05:55:46Z

@Meng-Heng could you run all the tests for mac with the .kmn based keyboard attached below, and report back on results (not as a user test, just as a comment):
bksp_kmn.zip

I would like to try and isolate the cause of the failure on macOS -- whether it is something in the LDML core implementation or a separate compatibility issue. Thanks!

Meng-Heng · 2026-01-30T07:29:26Z

@mcdurdin, hopefully this is easy to follow:

Test Name	Expected	Previous test result	KMN test result
TEST_COMPLIANT_E_ACUTE_VERIFY_CONTEXT	`abc é`	✅ `abc é`	✅ `abc é`
TEST_COMPLIANT_E_ACUTE	`abc e`	✅ `abc e`	❌ `abc`
TEST_NONCOMPLIANT_E_ACUTE_VERIFY_CONTEXT	`abc é`	✅ `abc é`	✅ `abc é`
TEST_NONCOMPLIANT_E_ACUTE	`abc e`	❌ `abca`	❌ `abc`
TEST_COMPLIANT_D_ACUTE_VERIFY_CONTEXT	`abc d́`	✅ `abc d́`	✅ `abc d́`
TEST_COMPLIANT_D_ACUTE	`abc d`	❌ `abc`	❌ `abc`
TEST_NONCOMPLIANT_D_ACUTE_VERIFY_CONTEXT	`abc d́`	✅ `abc d́`	✅ `abc d́`
TEST_NONCOMPLIANT_D_ACUTE	`abc d`	✅ `abc d`	✅ `abc d`
TEST_COMPLIANT_A_UMLAUT_VERIFY_CONTEXT	`abc ä`	✅ `abc ä`	✅ `abc ä`
TEST_COMPLIANT_A_UMLAUT	`abc a`	✅ `abc a`	❌ `abc`
TEST_NONCOMPLIANT_A_UMLAUT_VERIFY_CONTEXT	`abc ä`	✅ `abc ä`	✅ `abc ä`
TEST_NONCOMPLIANT_A_UMLAUT	`abc a`	❌ `abca`	❌ `abc`

Keyman 19.0193-alpha-test-15488

mcdurdin · 2026-01-31T04:56:27Z

@mcdurdin, hopefully this is easy to follow:

@Meng-Heng thanks, this is great -- although it's sad that there are failures. I will follow up with @sgschantz; looks like it's probably a mac-specific issue rather than anything relating to this PR.

mcdurdin · 2026-02-03T18:56:22Z

Test Results

GROUP_LINUX: Test on Linux

Tests run on Ubuntu 24.04 under X11.

Two issues noted on Wayland which do not pertain to this PR:

bug(linux): ldml keyboard desynchronizes and can crash when bksp is used in non-compliant app with decomposed cluster on Wayland #15550
Enter key can become 'stuck' (and sometimes other keys as well), which seems like keyup not being handled correctly

For compliant app, used Gnome Text Editor 46.3. For non-compliant app, used Chrome 144.0.7559.109.

TEST_COMPLIANT_E_ACUTE_VERIFY_CONTEXT (PASS): verified ok with Text Editor
TEST_COMPLIANT_E_ACUTE (PASS): no problems encountered
TEST_NONCOMPLIANT_E_ACUTE_VERIFY_CONTEXT (PASS): verified ok with Chrome
TEST_NONCOMPLIANT_E_ACUTE (PASS): no problems encountered
TEST_COMPLIANT_D_ACUTE_VERIFY_CONTEXT (PASS): verified ok with Text Editor
TEST_COMPLIANT_D_ACUTE (PASS): no problems encountered
TEST_NONCOMPLIANT_D_ACUTE_VERIFY_CONTEXT (PASS): verified ok with Chrome
TEST_NONCOMPLIANT_D_ACUTE (PASS): no problems encountered
TEST_COMPLIANT_A_UMLAUT_VERIFY_CONTEXT (PASS): verified ok with Text Editor
TEST_COMPLIANT_A_UMLAUT (PASS): no problems encountered
TEST_NONCOMPLIANT_A_UMLAUT_VERIFY_CONTEXT (PASS): verified ok with Chrome
TEST_NONCOMPLIANT_A_UMLAUT (PASS): no problems encountered

Performance in non-compliant apps was very slow under X11, but this may have been a configuration issue on my test machine? (@ermshiperete FYI)

mcdurdin · 2026-02-03T18:56:43Z

Test-bot: retest GROUP_MAC TEST_NONCOMPLIANT_E_ACUTE TEST_COMPLIANT_D_ACUTE TEST_NONCOMPLIANT_A_UMLAUT

mcdurdin · 2026-02-03T18:57:40Z

Test Results

GROUP_MAC: Test on macOS

TEST_NONCOMPLIANT_E_ACUTE (SKIP): A separate issue was encountered on macOS which blocked testing: bug(mac): backspace does not follow rules #15543
TEST_COMPLIANT_D_ACUTE (SKIP): A separate issue was encountered on macOS which blocked testing: bug(mac): backspace does not follow rules #15543
TEST_NONCOMPLIANT_A_UMLAUT (SKIP): A separate issue was encountered on macOS which blocked testing: bug(mac): backspace does not follow rules #15543

When normalizing, we need to stop processing on an NFC boundary, not an NFD boundary, to support normalizations such as in Bengali, where appending `U+09D7` to a context of `U+0995 U+09C7` should result in `U+0995 U+09CC`. The specification is unclear on this; see https://unicode-org.atlassian.net/browse/CLDR-19218 This also updates the ldml keyboard unit test suite to support running in full NFC mode (used in all Engine implementations) as well retaining the NFD mode (now only used by the debugger). Side note: the Bengali normalization failure case was picked up by the improvements to the unit test suite, proving once again that good tests are so valuable. Fixes: #15491 Fixes: #15505 Follows: #15488 Cherry-pick-of: #15506 Relates-to: CLDR-19218

keyman-server · 2026-02-04T18:01:32Z

Changes in this pull request will be available for download in Keyman version 19.0.198-alpha

github-project-automation bot added this to Keyman Jan 24, 2026

github-project-automation bot moved this to Todo in Keyman Jan 24, 2026

keymanapp-test-bot bot added the user-test-missing User tests have not yet been defined for the PR label Jan 24, 2026

keymanapp-test-bot bot added this to the A19S21 milestone Jan 24, 2026

github-actions bot added core/ Keyman Core fix labels Jan 24, 2026

keymanapp-test-bot bot added has-user-test user-test-required User tests have not been completed and removed user-test-missing User tests have not yet been defined for the PR labels Jan 24, 2026

mcdurdin commented Jan 25, 2026

View reviewed changes

mcdurdin marked this pull request as ready for review January 25, 2026 20:17

mcdurdin requested review from ermshiperete, rc-swag, sgschantz and srl295 January 25, 2026 20:18

mcdurdin mentioned this pull request Jan 25, 2026

bug(windows): Ctrl+Bksp with LDML keyboard only deletes one character prior to insertion point, not word #15489

Open

mcdurdin commented Jan 25, 2026

View reviewed changes

mcdurdin mentioned this pull request Jan 26, 2026

fix(core): handle backspace decomposition 🍒 🏠 #15494

Merged

Meng-Heng self-assigned this Jan 26, 2026

ermshiperete approved these changes Jan 26, 2026

View reviewed changes

srl295 approved these changes Jan 26, 2026

View reviewed changes

docs(core): add comments post review

033e3ee

Co-authored-by: srl295@gmail.com

srl295 approved these changes Jan 27, 2026

View reviewed changes

ermshiperete approved these changes Jan 27, 2026

View reviewed changes

mcdurdin mentioned this pull request Jan 28, 2026

fix(core): normalization segment should end on NFC boundary, not NFD #15506

Merged

darcywong00 reviewed Jan 28, 2026

View reviewed changes

core/tests/unit/ldml/ldml.cpp Show resolved Hide resolved

darcywong00 reviewed Jan 28, 2026

View reviewed changes

keymanapp-test-bot bot added the user-test-failed label Jan 29, 2026

mcdurdin mentioned this pull request Jan 30, 2026

maint(resources): test-bot does not pick up artifacts for Linux? #15518

Open

keyman-server modified the milestones: A19S21, A19S22 Jan 31, 2026

This was referenced Feb 1, 2026

bug(mac): backspace does not follow rules #15543

Open

bug(linux): ldml keyboard desynchronizes and can crash when bksp is used in non-compliant app with decomposed cluster on Wayland #15550

Open

keymanapp-test-bot bot removed the user-test-required User tests have not been completed label Feb 3, 2026

keymanapp-test-bot bot added user-test-required User tests have not been completed and removed user-test-failed labels Feb 3, 2026

keymanapp-test-bot bot removed the user-test-required User tests have not been completed label Feb 3, 2026

mcdurdin mentioned this pull request Feb 3, 2026

fix(core): normalization segment should end on NFC boundary, not NFD 🍒 🏠 #15551

Merged

mcdurdin merged commit 30f3d35 into master Feb 4, 2026
23 checks passed

mcdurdin deleted the fix/core/15487-bksp branch February 4, 2026 02:53

github-project-automation bot moved this from Todo to Done in Keyman Feb 4, 2026

		if(app_context_nfd == cached_context_string.substr(0, app_context_nfd.length())) {
		context_final = cached_context_string.substr(app_context_nfd.length());

Uh oh!

Conversation

mcdurdin commented Jan 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

User Testing

Note on compliant and non-compliant apps

Uh oh!

keymanapp-test-bot bot commented Jan 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

User Test Results

Test Artifacts

Uh oh!

mcdurdin Jan 25, 2026

Choose a reason for hiding this comment

Uh oh!

mcdurdin commented Jan 25, 2026

Uh oh!

mcdurdin Jan 25, 2026

Choose a reason for hiding this comment

Uh oh!

mcdurdin Jan 25, 2026

Choose a reason for hiding this comment

Uh oh!

srl295 Jan 26, 2026

Choose a reason for hiding this comment

Uh oh!

srl295 Jan 26, 2026

Choose a reason for hiding this comment

Uh oh!

srl295 commented Jan 26, 2026

Uh oh!

sgschantz commented Jan 27, 2026

Uh oh!

mcdurdin commented Jan 27, 2026

Uh oh!

Uh oh!

darcywong00 Jan 28, 2026

Choose a reason for hiding this comment

Uh oh!

mcdurdin Jan 31, 2026

Choose a reason for hiding this comment

Uh oh!

Meng-Heng commented Jan 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Prerequisite

GROUP_MAC:

GROUP_WINDOWS:

Uh oh!

mcdurdin commented Jan 30, 2026

Uh oh!

Meng-Heng commented Jan 30, 2026

Uh oh!

mcdurdin commented Jan 31, 2026

Uh oh!

mcdurdin commented Feb 3, 2026

Test Results

GROUP_LINUX: Test on Linux

Uh oh!

mcdurdin commented Feb 3, 2026

Uh oh!

mcdurdin commented Feb 3, 2026

Test Results

GROUP_MAC: Test on macOS

Uh oh!

Uh oh!

keyman-server commented Feb 4, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

mcdurdin commented Jan 24, 2026 •

edited

Loading

keymanapp-test-bot bot commented Jan 24, 2026 •

edited

Loading

Meng-Heng commented Jan 29, 2026 •

edited

Loading