Put the caret on the right side of the composition string

ui/base/ime/win/imm32_manager.cc

 // Copyright 2013 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "ui/base/ime/win/imm32_manager.h"
 
 #include <stdint.h>
 
 #include <memory>
 
@@ skipping 289 matching lines @@
   int length = static_cast<int>(composition->text.length());
 
   // Find out the range selected by the user.
   int target_start = length;
   int target_end = length;
   if (lparam & GCS_COMPATTR)
     GetCompositionTargetRange(imm_context, &target_start, &target_end);
 
   // Retrieve the selection range information. If CS_NOMOVECARET is specified,
   // that means the cursor should not be moved, then we just place the caret at
   // the beginning of the composition string. Otherwise we should honour the

[yukawa, 2017/04/17 03:01:05]

"we just place the caret at the beginning of the composition string"

Are you sure that CS_NOMOVECARET is specified in this case?  With you change,
now the comment becomes inconsistent with the actual behavior.

   // GCS_CURSORPOS value if it's available.
   // TODO(suzhe): due to a bug of webkit, we currently can't use selection range
   // with composition string. See: https://bugs.webkit.org/show_bug.cgi?id=40805
   if (!(lparam & CS_NOMOVECARET) && (lparam & GCS_CURSORPOS)) {

[yukawa, 2017/04/17 04:11:59]

As far as I've tested with the stock MS-IME Korean on Windows 7 SP1, the
following flags are specified in |lparam|.
 - GCS_COMPATTR
 - GCS_COMPSTR
 - CS_NOMOVECARET
 - CS_INSERTCHAR

https://msdn.microsoft.com/en-us/library/windows/desktop/dd374133.aspx

Here is the important part:
  CS_NOMOVECARET:
    Do not move the caret position as a result of processing the
    message. For example, if an IME specifies a combination of
    CS_INSERTCHAR and CS_NOMOVECARET, the application should insert
    the specified character at the current caret position but should
    not move the caret to the next position. A subsequent
    WM_IME_COMPOSITION message with GCS_RESULTSTR will replace this
    character.

So as far as the API spec is concerned, I think the current Chromium's code
actually makes sense because those flags are intentionally specified by MS-IME
Korean (or Cicero Unaware Application Support).

I don't say that we should give up to fix crbug.com/653160, but if we would
decide to violate the spec requested by the API intentionally, we should try to
minimize the impact so as not to break other IMEs that expect Chromium to
support these flags correctly.

     // IMM32 does not support non-zero-width selection in a composition. So
     // always use the caret position as selection range.
     int cursor = ::ImmGetCompositionString(imm_context, GCS_CURSORPOS, NULL, 0);
     composition->selection = gfx::Range(cursor);
   } else {
-    composition->selection = gfx::Range(0);
+    composition->selection = gfx::Range(length);
   }
 
   // Retrieve the clause segmentations and convert them to underlines.
   if (lparam & GCS_COMPCLAUSE) {
     GetCompositionUnderlines(imm_context, target_start, target_end,
                              &composition->underlines);
   }
 
   // Set default underlines in case there is no clause information.
   if (!composition->underlines.empty())
@@ skipping 271 matching lines @@
                              | IME_CMODE_KATAKANA
                              | IME_CMODE_FULLSHAPE);
       break;
     default:
       *open = FALSE;
       break;
   }
 }
 
 }  // namespace ui