www.digitalmars.com         C & C++   DMDScript  

c++.announce - DM C/C++ 8.46

reply "Walter Bright" <newshound digitalmars.com> writes:
D users will be interested in the updated import libraries and uuid.lib.

http://www.digitalmars.com/changelog.html
Jan 14 2006
next sibling parent reply David Medlock <ashleymedlock no.spam.yahoo.com> writes:
Walter Bright wrote:
 D users will be interested in the updated import libraries and uuid.lib.
 
 http://www.digitalmars.com/changelog.html
 

The link for the CD upgrade is missing: http://www.digitalmars.com/download/freecompiler.html -DavidM
Jan 15 2006
parent reply tyro[a.c.edwards] <tyro[a.c.edwards]_member pathlink.com> writes:
In article <dqdi4s$n10$1 digitaldaemon.com>, David Medlock says...
Walter Bright wrote:
 D users will be interested in the updated import libraries and uuid.lib.
 
 http://www.digitalmars.com/changelog.html
 

The link for the CD upgrade is missing: http://www.digitalmars.com/download/freecompiler.html -DavidM

This version of the. compiler is missing from the ftp site. Lastest version available is still 8.45.
Jan 15 2006
parent reply "Walter Bright" <newshound digitalmars.com> writes:
Fixed. 
Jan 15 2006
parent Jerry van Dijk <jerry jvdsys.demon.nl> writes:
 Fixed. 

Actually, the CD update (cd846.zip) is still missing. -- -- Jerry van Dijk -- Leiden, Holland -- -- In case of headache, follow the instructions on the aspirin bottle: -- "Take two tablets" and "Keep away from children".
Jan 16 2006
prev sibling parent Don Clugston <dac nospam.com.au> writes:
Walter Bright wrote:
 D users will be interested in the updated import libraries and uuid.lib.
 
 http://www.digitalmars.com/changelog.html

Thanks, Walter. Unfortunately, the tgamma() function is still not completely right, in both DMC and DMD. It's only really accurate to double precision. Interestingly, lgamma() is correct. With the version of tgamma I've included in dsource.org in mathextra/etcgamma, the following test passes: const real SQRT_PI = 1.77245385090551602729816748334114518279754945612238L; assert(feqrel(etcgamma.tgamma(0.5L), SQRT_PI) == real.mant_dig); (ie, it is corect to the full 64 bits at this point!) but assert(feqrel(std.math.tgamma(0.5L), SQRT_PI) == real.mant_dig); fails -- it is correct to "only" 56 bits. I wonder if there's an inadvertant conversion to 'double' in there somewhere? There's a certain amount of black humour in this -- it is *much* easier to get this stuff right in D, than in C. I think you should consider implementing the C library in D, rather than vice-versa <g>. Seriously, with the known problems with NaNs in most C++ compilers, and the lack of accuracy guarantees in the C math functions, in the longer term, D is likely to better off minimising its dependence on C. If pow() is often implemented incorrectly, there's no chance with anything more complicated.
Jan 16 2006