Takuro Ashie
ashie****@homa*****
2004年 8月 30日 (月) 17:15:08 JST
足永です． tmlからMatthias氏，0wen氏，私宛に以下の議論が振られています (なぜMLでやらない...)． DMですが，公開しても問題ない内容のはずなので転送します． 後半のwide charの話は単純にwin95を無視する方向で進んでいて，まあそれでい いんじゃないかと私も思っています． 前半のサポート言語の話ですが，これはja, ko, zhだけで十分でしょうか? 問題がある場合は教えて下さい． Begin forwarded message: Date: Mon, 30 Aug 2004 05:53:39 +0000 From: Tor Lillqvist <tml****@iki*****> To: Matthias Clasen <mclas****@redha*****> Cc: otayl****@redha*****, ashie****@homa***** Subject: Re: IME input method integration With minimal changes (already committed to CVS), it builds and installs fine. (Well, as "fine" as the input modules can be said to install on mingw; one always have to do some manual editing of the generated gtk.immodules file.) There are some minor issues with the code: imime.c sets itself as default for all languages ("*" in the GtkIMContextInfo struct). It probably should be default for "ja:ko:zh"? gtkimcontextime.c uses #ifdef UNICODE to switch between the wide-char and multibyte ("ANSI") API at compile-time. No GTK+ code is built with UNICODE defined. Instead it could use G_WIN32_HAVE_WIDECHAR_API() (in HEAD GLib) to decide at run-time which one to use. (We also should add more stuff like that in gdk/win32, to for instance set window titles using the wide-char API if available.) But then, the G_WIN32_HAVE_WIDECHAR_API() just tells the truth for the "base" Win32 API, which is available in wide-char versions just on the NT-based Windowses. Some (all?) of the Imm* API is available in wide-char versions also Win98/Me, but not Win95. Should we just ignore Win95 (and then just use the wide-char Imm* API all the time), or check the g_win32_get_windows_version() return value do differentiate between 95 and 98/Me? If the latter, the DLLs on 95 probably don't even have a dummy entry point present, so if we want to support 95, we need to do run-time lookup of the function names. --tml -- Takuro Ashie <ashie****@good-*****>
