"Laurentiu Pancescu" <user@nowhere.near> wrote in message news:adfopf$1ast$1@digitaldaemon.com...
> It's not only for my own template metaprogramming... <g>.  Important and high-performance libraries, like Blitz++, Boost, or Loki, cannot be used with DMC.  Even using STLport has serious limitations.  The headers are
not
> compliant (like in <math.h> vs. <cmath>), so that anything that uses the legal "#include <iostream>/using namespace std;" won't compile.  It's
pretty
> much impossible to use DMC (otherwise, an excellent compiler) for C++,
since
> most code is modern, and ISO-C++ compliant.  And when I write code, I
write
> it ISO-C++ compliant as much as possible (Win32 extensions can't be
avoided
> in some files), for a good reason: it must be usable by anybody who has a ISO-C++ compliant compiler.  Almost anybody, that is...

Oh, I'm painfully aware of that. I am trying to get the compiler up to date. -Walter

"Walter" <walter@digitalmars.com> schrieb im Newsbeitrag news:adh53r$2qmg$1@digitaldaemon.com...

> Oh, I'm painfully aware of that. I am trying to get the compiler up to date. -Walter

That's good to hear! I only can agree with what the others have posted about standard conformance. My biggest problem with DMC++ is the STL stuff. Hopefully we will see a release in the near future where some of the biggest gaps are closed.


--
Robert M. Münch
IT & Management Freelancer
Mobile: +49 (0)177 2452 802
Fax   : +49 (0)721 8408 9112
Web   : http://www.robertmuench.de

I'm really glad to hear this!  I was afraid that D is taking all of your time (or most of it ;).  After all, writing a new frontend for gcc isn't exactly trivial...

If you manage to get the template support to be fully compliant with the ISO-C++ Standard (maybe except for the "export" keyword - currently only Comeau has it in beta), it'll be already a great improvement, since we could use STLport out-of-the-box, and it has its own iostream implementation, fully compliant (but dependent on full template support).  After that we can write files like cmath by ourselves:  [LOL]

// cmath
namespace std {
#include <math.h>
}

Regards, and good luck,
  Laurentiu

"Walter" <walter@digitalmars.com> wrote in message news:adh53r$2qmg$1@digitaldaemon.com...
>
> "Laurentiu Pancescu" <user@nowhere.near> wrote in message news:adfopf$1ast$1@digitaldaemon.com...
> > It's not only for my own template metaprogramming... <g>.  Important and high-performance libraries, like Blitz++, Boost, or Loki, cannot be used with DMC.  Even using STLport has serious limitations.  The headers are
> not
> > compliant (like in <math.h> vs. <cmath>), so that anything that uses the legal "#include <iostream>/using namespace std;" won't compile.  It's
> pretty
> > much impossible to use DMC (otherwise, an excellent compiler) for C++,
> since
> > most code is modern, and ISO-C++ compliant.  And when I write code, I
> write
> > it ISO-C++ compliant as much as possible (Win32 extensions can't be
> avoided
> > in some files), for a good reason: it must be usable by anybody who has
a
> > ISO-C++ compliant compiler.  Almost anybody, that is...
>
> Oh, I'm painfully aware of that. I am trying to get the compiler up to date. -Walter
>
>

I would pay a sizable whack (more than $200 !) for that, also.

I currently have a lot of issues with DMC and templates, but am persevering (see http://stlsoft.org, http://winstl.org, http://comstl.org), as I like the compiler, and anything that reduces M$'s stranglehold on C++ developments on Intel platforms

"Laurentiu Pancescu" <user@nowhere.near> wrote in message news:adfopf$1ast$1@digitaldaemon.com...
>
> "Jan Knepper" <jan@smartsoft.cc> wrote in message news:3CFB60DF.86B51DC5@smartsoft.cc...
> > Laurentiu Pancescu wrote:
> >
> > > - December 1987: g++ 1.13, first C++ release of gcc
> >
> > Where did you get that?
> > I never heard g++ was 'native' that early.
>
> I first read about it in a document (Postscript) downloaded from www.oonumerics.org.  But please check at http://gcc.gnu.org , in their "Releases" page: "December 18th 1987 - 1.15.3 (g++)".  I assume it was native, since it's just a frontend for the gcc code generator, just as it
is
> now, I think...
>
> > > And they are both behind the ISO-C++ standard, which really makes me feel sorry.  Those compilers generate too good code, it's a pity that
> you
> > > must think whether you should use them or not, based on portability
and
> > > compliance considerations (C++ only, C standard is implemented
> correctly).
> > > :-(
> >
> > Oh, I choose to not use the latest ISO/ANSI C++ standard and compile my
> code
> > usually with more than one compiler... Have not done that for some time
> though
> > as I have been rather busy lately and really want to spend time with my
> wife as
> > well...
>
> It's not only for my own template metaprogramming... <g>.  Important and high-performance libraries, like Blitz++, Boost, or Loki, cannot be used with DMC.  Even using STLport has serious limitations.  The headers are
not
> compliant (like in <math.h> vs. <cmath>), so that anything that uses the legal "#include <iostream>/using namespace std;" won't compile.  It's
pretty
> much impossible to use DMC (otherwise, an excellent compiler) for C++,
since
> most code is modern, and ISO-C++ compliant.  And when I write code, I
write
> it ISO-C++ compliant as much as possible (Win32 extensions can't be
avoided
> in some files), for a good reason: it must be usable by anybody who has a ISO-C++ compliant compiler.  Almost anybody, that is...
>
> And this is extremely important: by writing old-style code, you risk that your code can only be compiled with old compilers, since there's no guarantee that new compilers will still support obsolete language features for an unlimited period of time.  Even Microsoft and Borland seem to have compliance as an important target...
>
> But think how good it could get for DMC: we could have a cheap compiler, fully ISO-C++ compliant, lightning-fast when compiling, generating
extremely
> tight and fast code for multiple targets, also with a MFC license.  I'd
pay
> not $25, not $50, but more than $200 for that!  Only if we could convince Walter of this...
>
> Laurentiu
>
>

"Matthew Wilson" <mwilson@nextgengaming.com> wrote in message news:adotjl$1kdo$1@digitaldaemon.com...
> I would pay a sizable whack (more than $200 !) for that, also.
>
> I currently have a lot of issues with DMC and templates, but am
persevering
> (see http://stlsoft.org, http://winstl.org, http://comstl.org), as I like the compiler, and anything that reduces M$'s stranglehold on C++ developments on Intel platforms

STLsoft's web site says they support Digital Mars! Though I will try to get all the template support standard compliant. -Walter

Sorry, was unclear. I meant that my perseverance has paid off partially, in that most of the STLSoft's libraries (and most of the WinSTL's also) work correctly with DMC. There are still some issues which I am yet to iron out.

Most of these issues crop up when one tries to involve the standard library, although some (eg. DMC appears to strongly dislike template methods defined outside the class definition) are independent of this

BTW, I was serious about the $DMC++.


"Walter" <walter@digitalmars.com> wrote in message news:adp4pl$1rdi$1@digitaldaemon.com...
>
> "Matthew Wilson" <mwilson@nextgengaming.com> wrote in message news:adotjl$1kdo$1@digitaldaemon.com...
> > I would pay a sizable whack (more than $200 !) for that, also.
> >
> > I currently have a lot of issues with DMC and templates, but am
> persevering
> > (see http://stlsoft.org, http://winstl.org, http://comstl.org), as I
like
> > the compiler, and anything that reduces M$'s stranglehold on C++ developments on Intel platforms
>
> STLsoft's web site says they support Digital Mars! Though I will try to
get
> all the template support standard compliant. -Walter
>
>

"Matthew Wilson" <mwilson@nextgengaming.com> wrote in message news:adpafe$2cti$1@digitaldaemon.com...
> Sorry, was unclear. I meant that my perseverance has paid off partially,
in
> that most of the STLSoft's libraries (and most of the WinSTL's also) work correctly with DMC. There are still some issues which I am yet to iron
out.
>
> Most of these issues crop up when one tries to involve the standard
library,
> although some (eg. DMC appears to strongly dislike template methods
defined
> outside the class definition) are independent of this
>
> BTW, I was serious about the $DMC++.

I didn't realize you were doing the support work there. Thanks! I can help - can you prioritize the compiler problems that cause the most grief? Also, boiling them down to 10 lines or less for each problem helps a lot. (I've found about 90% of the time dealing with STL compilation problems is whittling away all the dense brush obfuscating what really went wrong.)

It's all me, baby! :)

STLSoft is simply an open-source imprint of my company (http://synesis.com.au), in order that I can move a lot of useful software into the public-domain: partly to use with clients (since it simplifies the separation, at least in the minds of their legal departments!); partly so that I can write (articles, etc.) about the techniques without having the encumberance of having my company badged all over it (not to mention its overweening headers dependency tree); partly out of the good-ole' goodness of my public-domain open-source heart.

Seriously though, Walter, am in the midst of some serious time-consuming activities (as per our off-line conversation), but will be glad to take you up on these issues when I get back to STLSoft-ing in a couple of weeks.

Thanks for the offer. Much appreciated. (We can probably do each other some good, you get some tough STL nuts to crack, and I get more comprehensive and powerful support for my libraries)



"Walter" <walter@digitalmars.com> wrote in message news:adpbb6$2dpv$1@digitaldaemon.com...
>
> "Matthew Wilson" <mwilson@nextgengaming.com> wrote in message news:adpafe$2cti$1@digitaldaemon.com...
> > Sorry, was unclear. I meant that my perseverance has paid off partially,
> in
> > that most of the STLSoft's libraries (and most of the WinSTL's also)
work
> > correctly with DMC. There are still some issues which I am yet to iron
> out.
> >
> > Most of these issues crop up when one tries to involve the standard
> library,
> > although some (eg. DMC appears to strongly dislike template methods
> defined
> > outside the class definition) are independent of this
> >
> > BTW, I was serious about the $DMC++.
>
> I didn't realize you were doing the support work there. Thanks! I can
help -
> can you prioritize the compiler problems that cause the most grief? Also, boiling them down to 10 lines or less for each problem helps a lot. (I've found about 90% of the time dealing with STL compilation problems is whittling away all the dense brush obfuscating what really went wrong.)
>
>
>