U++ forum: Welcome to the forum

Re: Building & using U++ without TheIDE [message #11605 is a reply to message #11604]

Wed, 19 September 2007 09:34

mirek
Messages: 13980
Registered: November 2005

Ultimate Member

sergei wrote on Tue, 18 September 2007 18:43

Update:

I'm done implementing unicode. I don't exactly like the way I did it, but it works. I prefer some global strings manager like the TSTR I suggested, but since Mirek said he doesn't want TCHARs I've updated all API calls manually. I added PLATFORM_UNICODE define, and replaced many #ifdef PLATFORM_WINCE with it. Conversion mostly implemented through ToSystemCharset and FromSystemCharset.

In the process I've also found and eliminated a "security vulnerability" (that's what memory bugs are called nowadays Razz

) in Log.h:

sprintf(h, "* %s %02d.%02d.%04d %02d:%02d:%02d, user: %s\n",
(const char*)FromSystemCharset(exe),
t.day, t.month, t.year, t.hour, t.minute, t.second, (const char*)FromSystemCharset(user));

I added the (const char*), otherwise I got segmentation fault in debug. Probably without the explicit cast sprintf thought that String is a char array.

I received another segmentation fault earlier (but maybe it was this one...), and that one was solved by replacing all To/FromSysChrSet with To/FromSystemCharset. Not a big deal IMHO, they were all used in Win32-specific code anyway...

Uh oh, that is no good. There definitely should be FromSysCharset, which returns a pointer to static char * array.

The reason is that Log is supposed to work independently from the rest of U++, so that it can be used in situation when the rest failed or is not available. By using vanilla FromSystemCharset, you make it dependent on String and in turn on memory allocator. Therefor it will not work when heap crashes or cannot be used to debug memory allocator... Smile

Quote:

I've tested unicode filenames - it did open a multilingually-named file and read its content successfully. I've also tested registry - successfully wrote that filename to a REG_SZ key, it's fine. Didn't try to create unicode-named keys (don't wanna kill my windows).

Sort of redundant work. The final solution will use dynamic .dll loading to choose between W and A variants.

Quote:

P.S.2 since you know about the MSVC bug, did anyone report to MS? There might be a chance that they fix it...

No.

Quote:

Edit: I think UTFBOM class I posted above, or something else implementing that functionality, should be added to some place in U++. That way unicode support will be complete - unicode filenames + unicode text.

Well, the question is the definition. Thinking about it, I just fail to see what is the exact description. If detection of utf-8 files is the purpose, that can be easily done withou specific code. Also, what is going to happen if there is no UTFBOM?

Should it be available as

bool SkipUTFBOM(Stream& in);
const char *SkipUTFBOM(const char *s);

?

Mirek

Report message to a moderator