Difference between revisions of "Compiling Inkscape on Windows 32-bit"
(un spam) |
|||
Line 26: | Line 26: | ||
Although that simplifies things, it's not required. I've done Win32 programs that are 16-bit Unicode and run on Windows 95/98/ME with no problems years before they released that. | Although that simplifies things, it's not required. I've done Win32 programs that are 16-bit Unicode and run on Windows 95/98/ME with no problems years before they released that. | ||
Soo.... | Soo.... _UNICODE can't be used as it makes your binary unusable on Win9x/ME. Then it turns out that TCHAR and _T can't be used either. (Oh, and MS can't really use them, since Office does run on Win9x/ME). So just explicitly use some 16-bit datatype for chars. Microsoft is misleading in their documentation when they state that wchar_t is 16-bit. Most other platforms (including Mac OS, Linux, Solaris, BSD, etc) follow the language standard's recommendation and make it 32-bit. So it is best to avoid wchar_t for any real cross-platform code. | ||
So, some general guidelines are: | So, some general guidelines are: |
Revision as of 10:27, 3 November 2004
See the Win32 Build ticket for some info.
Inkscape under Windows
We have made the building of Inkscape on Win32, or cross-compiling from Linux to Win32, as easy as 1-2-3.
See this page for simple Win32 building instructions
-- Bob Jamison
A Note from Jon about UNICODE on Win32
Actually, Microsoft themselves never use _UNICODE, nor it's 'evil' friends TCHAR and _T. If it is used, it results in a binary that can not be run on Windows 95/98/ME. It is more of a "Windows NT only" define.
Microsoft applications like MS Office have been pure 16-bit Unicode internally for years now. Also remember that BSTRs in Win32 are required to have 16-bit Unicode data. So for any COM access on a Win32 box running Win9x/ME 'multibyte' or '8-bit' COM calls convert all ANSI data to Unicode once you pass things in. To avoid this, Microsoft just keeps data 16-bit Unicode and then translates to local ANSI codepage only when data needs to be passed directly to a Win32 call.
To help developers, Microsoft finally made public their API they had been using. It's "The Microsoft Layer for Unicode".
Although that simplifies things, it's not required. I've done Win32 programs that are 16-bit Unicode and run on Windows 95/98/ME with no problems years before they released that.
Soo.... _UNICODE can't be used as it makes your binary unusable on Win9x/ME. Then it turns out that TCHAR and _T can't be used either. (Oh, and MS can't really use them, since Office does run on Win9x/ME). So just explicitly use some 16-bit datatype for chars. Microsoft is misleading in their documentation when they state that wchar_t is 16-bit. Most other platforms (including Mac OS, Linux, Solaris, BSD, etc) follow the language standard's recommendation and make it 32-bit. So it is best to avoid wchar_t for any real cross-platform code.
So, some general guidelines are:
- Never define _UNICODE
- Don't use _T
- Don't use TCHAR
- Don't use wchar_t
- Use some explicit 16-bit type for characters/strings
- IBM's ICU does this
So then it's a simple matter of converting data from the standard UTF-8 GTK+ data to UTF-16 data before passing to Win32-land.
The result? A single binary that executes properly on both Win9X/ME and WinNT/2K/XP, including full and proper font and filename support.
And, as I've mentioned before, I've actually achieved this for a few different shipping products and projects.
-- [JonCruz Jon C]
CVS and Windows
If you want to do developer CVS access to SF on windows, I'd heartily recomend TortoiseCVS. It has the SSH stuff built into it so you dont have to mess about with putty, and it integrates right into explorer, so CVS becomes just an extension of the file commands. get it from http://tortoisecvs.sourceforge.net/
John