Re: stable STL implementation

[Nathan Myers <ncm-nospam at cantrip dot org>]

On Sat, Nov 30, 2002 at 10:54:25AM +0530, Vijay Patil wrote:
>            which is the most stable implementation of STL?? , is it Dinkum,
> SGI, STLport, Roguewave, or somethnig else
> 
> i am working on platform independent product, we have used STL every where, 
> currently we support Win2000(VC++) , Linux(gcc2.95.3), Solaris(gcc2.95.3) ,
> but some times i get very strange behavior of STL on Win2000, and same 
> thing works fine on Linux and Solaris

Clearly the most stable implementation, by far, is the one in gcc-2.95:
it hasn't changed at all in well over five years.  It's also the least 
standard-conformant, and quite buggy besides.  (It's the most stable
because it has been entirely unmaintained.)  From context, I suspect 
you meant to ask not about stability, but about bugginess, but it's 
possible you aren't thinking clearly enough to have actually meant 
anything at all.

What does it matter which is more stable, or which is less buggy?  You 
will use one that works with the compiler you end up forced to use, for 
whatever reason, and you will just have to work around the bugs you find.  
If you use one compiler on all platforms, you will have only one set of
bugs to work around.  Few compilers are released for all your platforms, 
but one is gcc.  

This is the wrong place to ask questions like yours.  This is a forum
for discussion of a particular implementation, and (in particular) not
the one in gcc-2.95, and not the one in MSVC, and not STLport, and not
Rogue Wave's, and not Dinkumware's.  The implementation maintained 
here is the one that comes with gcc-3.x.  It works pretty well, albeit 
with very slow iostreams.  It has been ported to all the platforms you
mention.  Give it a try and decide for yourself.  Be aware that code that 
compiles against gcc-2.95's library is almost certainly not standard-
conformant, so you will need to bring it up to date to work with a
standard-conformant library.

Nathan Myers
ncm-nospam@cantrip.org