X-Authentication-Warning:	delorie.com: mail set sender to djgpp-bounces using -f
From:	"Carey Evans" <carey DOT evans AT gmail DOT com>
Newsgroups:	comp.os.msdos.djgpp
Subject:	ctype<char> custom table lookup in C++
Date:	19 May 2005 16:41:32 -0700
Organization:	http://groups.google.com
Lines:	29
Message-ID:	<1116546092.867392.97940@g43g2000cwa.googlegroups.com>
NNTP-Posting-Host:	210.54.0.162
Mime-Version:	1.0
X-Trace:	posting.google.com 1116546098 24614 127.0.0.1 (19 May 2005 23:41:38 GMT)
X-Complaints-To:	groups-abuse AT google DOT com
NNTP-Posting-Date:	Thu, 19 May 2005 23:41:38 +0000 (UTC)
User-Agent:	G2/0.2
Complaints-To:	groups-abuse AT google DOT com
Injection-Info:	g43g2000cwa.googlegroups.com; posting-host=210.54.0.162;
	posting-account=a2Dz9g0AAAAQI7DN4N6e935Thq5XWCRh
To:	djgpp AT delorie DOT com
DJ-Gateway:	from newsgroup comp.os.msdos.djgpp
Reply-To:	djgpp AT delorie DOT com

I've been trying to write a custom C++ ctype<char> facet to support the
full range of characters in codepage 850, not just ASCII.  However,
I've run across a problem with the ctype<char>::is(mask,char) lookup on
the custom table I pass to the ctype<char> constructor.

The ISO C++ standard says of is(mask m, char c) that:

> The first form returns table()[(unsigned char)c] & m; ...

but with GCC 3.4.3 in DJGPP, it returns (from ctype_inline.h):

> _M_table[static_cast<unsigned char>(__c + 1)] & __m;

which means that all the character classifications in my custom class
are off by one.

On the other hand, the second form of is(const char*,const char*,mask*)
works with my custom table, but not with the ctype<char> from the
standard locale.

The correct solution appears to me to be to remove the "+ 1" from the
is() function, and add it to the ctype<char> constructors in
ctype_noninline.h to do something like:

> _M_table(__table ? __table : __dj_ctype_flags + 1)

-- 
Carey Evans

- Raw text -