Re: The trend is to move away from Delphi...

David Dean <ozchzhq02@xxxxxxxxxxxxxx> wrote in news:ozchzhq02-
FC0DDC.12253515082006@xxxxxxxxxxxxx:


There are weaknesses, but C++ conformance is slightly higher, and
there are a lot of runtime bugs that got fixed.

Um...I'd disagree. I'd say that the new compiler is simply "differently
broken" than it was before. Many things were fixed, but many things that
used to work don't work now (or don't work as well). (The poppack.h thing
is a good example, and the non-debug optimizers or pre-compiled headers
problem is yet another. Plus the makefile generator is still buggier than
the old verison. Plus...well, I could go on.) And due to the fact that
important libraries like Boost don't track Borland's compiler changes very
quickly, so getting Boost to build is still a headache. (And Boost still
lags on Borland's compiler as compared to Visual C++, as Microsoft's
compiler is more standards-conformant at the present time.)

The C++ community has
been pushing for things and the last year has been a real boon to us.
That being said, we still have quite a hill to climb.


Borland's C++ compiler is pretty long in the tooth -- it's basically a
patched version of the same compiler they've been using for the last six or
seven years. Nothing wrong with that, really, but it hasn't kept pace with
the times: standards-conformance, optimizations, and 64-bit support are
important areas where Borland's compiler lags. But the toolchain is also
out of date in terms of not having current Win32/ATL headers, and a
standard library that has changed substantially with every version release:
verison 5 used Rogue Wave, version 6 used STLPort (ick), and BDS 2006 uses
DinkumWare. Even the linker could use some love (although it did get a big
update when BDS 2006 came out so it doesn't barf on big projects).

I'm happy to see Borland's re-commitment to their C++ toolchain, but they
have a *lot* of work ahead of them if they want to be competitive in this
space. I've said before that I think they'd do better to just fork a
version of GCC 4.x and hack it to compile VCL sources (easier said than
done, I know). The GCC compiler is in other respects more standards-
compliant and far better supported than Borland's own compiler.

The problem for a lot of C++ people is that we don't see Borland (or DevCo)
as "serious" about C++ -- Delphi seems to get the lion's share of attention
and mindshare.

mr_organic
.

Relevant Pages

  • preprocessor implementation GURU question
    ... I'm trying to resolve a disagreement between friends --Digital Mars ... The problem arose when I was trying to compile a boost example program ... compiler shouldn't be adding a space. ... involve concatenation, as it does not cause tokens to merge. ...
    (comp.lang.cpp)
  • Re: Copy constructors
    ... One of these days I must explore the "boost" library, ... I expect the boost::noncopyable class does the same thing, as the compiler ... can't generate the defaults in derived classes if the base class versions ... the empty base class optimization. ...
    (microsoft.public.vc.language)
  • Re: generic way to access C++ libs?
    ... The fact that Windows is the only widespread OS where getting a compiler ... Boost does refuse to make any assumptions, ... > assumptions could be built into Pyste. ...
    (comp.lang.python)
  • Re: VC7.1/8 Explicit templated constructor invoked instead of copy ctor
    ... You are one of the boost guys. ... old compiler) and Comeau online compiler, ... ctor in 'derived' (using static_cast to force the use of correct ... Fortunately the workaround is viable in our real code base. ...
    (microsoft.public.vc.language)
  • Re: swap the array items
    ... |> And as I already pointed out, installing and using boost is *easy*. ... Some boost libraries ... So you just want a wierd compiler error-message? ... | out-of-the-box (with binary libraries). ...
    (alt.comp.lang.learn.c-cpp)