[gotoh]

What does the pull request do?

This PR fixes Win32 IME preedit caret handling in Imm32InputMethod.

When IMM32 reports GCS_CURSORPOS, Avalonia now forwards that caret position to TextInputMethodClient.SetPreeditText, so the visual caret in the preedit string stays in sync with the actual IME composition cursor.

The change is limited to the Win32 IMM32 path and does not modify the higher-level text input model or platform-independent preedit behavior.

What is the current behavior?

On Windows, Avalonia updates the preedit text content, but it does not track the caret position inside the active composition string.

As a result, when editing an in-progress IME composition, the logical caret can move inside the composition string while the visual preedit caret remains stale, typically staying at the end.

What is the updated/expected behavior with this PR?

Avalonia now updates the preedit caret position together with the preedit text on the Win32 IMM32 path.

This makes the caret follow IME cursor movement correctly when editing an active composition string, including cases where the user returns from conversion to composition editing and moves the caret within the preedit text.

Manual verification was done with Japanese IME on Windows 10/11 using samples/ControlCatalog.Desktop (TextBox page), including both Google IME and Microsoft IME.

Before	After

How was the solution implemented (if it's not obvious)?

• cached the current Win32 composition cursor position in Imm32InputMethod
• propagated GCS_CURSORPOS updates to TextInputMethodClient.SetPreeditText
• cleared cached cursor state on reset, composition start/end, result commit, and client changes
• avoided reapplying stale cursor state when GCS_RESULTSTR and GCS_CURSORPOS are reported together without a new GCS_COMPSTR

Checklist

• Added unit tests (if possible)?
• Added XML documentation to any related classes?
• Consider submitting a PR to https://github.com/AvaloniaUI/avalonia-docs with user documentation

Breaking changes

None.

Obsoletions / Deprecations

None.

Fixed issues

Fixes #10933

Likely fixes #21618 as well, since the reported symptom matches this Win32 IMM32 caret synchronization issue. However, my manual verification for this PR was performed with Japanese IMEs (Google IME and Microsoft IME), not with a Chinese IME specifically.

[avaloniaui-bot]

You can test this PR using the following package version. 12.1.999-cibuild0066768-alpha. (feed url: https://nuget-feed-all.avaloniaui.net/v3/index.json) [PRBUILDID]

[cla-avalonia]

• All contributors have signed the CLA.

[gotoh]

@cla-avalonia agree

[avaloniaui-bot]

You can test this PR using the following package version. 12.1.999-cibuild0066770-alpha. (feed url: https://nuget-feed-all.avaloniaui.net/v3/index.json) [PRBUILDID]

[avaloniaui-bot]

You can test this PR using the following package version. 12.1.999-cibuild0066796-alpha. (feed url: https://nuget-feed-all.avaloniaui.net/v3/index.json) [PRBUILDID]

[avaloniaui-bot]

You can test this PR using the following package version. 12.1.999-cibuild0066838-alpha. (feed url: https://nuget-feed-all.avaloniaui.net/v3/index.json) [PRBUILDID]

[MrJul]

Tested with both Japanese and Chinese IME (MS Pinyin). Moving the caret works as advertised.
LGTM!

[love-linger]

Can this PR backport to 11.3.x branch?

[MrJul]

Can this PR backport to 11.3.x branch?

No, 11.3 won't receive any new updates (unless it's a security fix). If you have an enterprise support agreement, you can contact the team to request additional fixes.

[Warm-winter]

@gotoh Thanks, using the test version can resolve the issue.