Comments on: Pimpl your C++ code http://anteru.net/2009/03/14/385/ Graphics, programming & software engineering Thu, 22 Mar 2012 06:25:01 +0000 hourly 1 http://wordpress.org/?v=3.3.2 By: Anteru http://anteru.net/2009/03/14/385/comment-page-1/#comment-12564 Anteru Thu, 05 Nov 2009 22:29:18 +0000 http://anteru.net/?p=385#comment-12564 <a href="#comment-12561" rel="nofollow">@John Doe</a> Either you commented at the wrong blog, or at least at the wrong entry, as there is no export template here at all. If you mean the "C++ tricks, #6: Explicit template instantiation" entry, I use <code>extern template</code> there since the beginning; all the code on my blog is actually tested with MSVC at least, and typically with GCC as well. @John Doe
Either you commented at the wrong blog, or at least at the wrong entry, as there is no export template here at all. If you mean the “C++ tricks, #6: Explicit template instantiation” entry, I use extern template there since the beginning; all the code on my blog is actually tested with MSVC at least, and typically with GCC as well.

]]> By: John Doe http://anteru.net/2009/03/14/385/comment-page-1/#comment-12561 John Doe Thu, 05 Nov 2009 21:38:59 +0000 http://anteru.net/?p=385#comment-12561 The second part of this blog post is just plain wrong, there is no such keyword as "export template" and the "export" keyword is only implemented by Comeau and Digital Mars compilers, not by MSVC, GCC, Intel. There is a keyword "extern template" but it just says that the compiler shouldn't generate code for that template instantiation in the current translation unit (expecting it to already have been generated in another one). This helps out compilers which aren't smart enough to do comdat folding at linktime. The second part of this blog post is just plain wrong, there is no such keyword as “export template” and the “export” keyword is only implemented by Comeau and Digital Mars compilers, not by MSVC, GCC, Intel.

There is a keyword “extern template” but it just says that the compiler shouldn’t generate code for that template instantiation in the current translation unit (expecting it to already have been generated in another one). This helps out compilers which aren’t smart enough to do comdat folding at linktime.

]]> By: Anteru http://anteru.net/2009/03/14/385/comment-page-1/#comment-5424 Anteru Sat, 14 Mar 2009 17:04:40 +0000 http://anteru.net/?p=385#comment-5424 Because std::auto_ptr is dangerous ;) If you use std::auto_ptr, and forget to implement the copy-constructor, the first copy using std::auto_ptr will destroy the source. With shared_ptr, you get shared state, which is much less of a problem (read: it won't crash). Because std::auto_ptr is dangerous ;) If you use std::auto_ptr, and forget to implement the copy-constructor, the first copy using std::auto_ptr will destroy the source. With shared_ptr, you get shared state, which is much less of a problem (read: it won’t crash).

]]> By: bubu http://anteru.net/2009/03/14/385/comment-page-1/#comment-5423 bubu Sat, 14 Mar 2009 16:17:37 +0000 http://anteru.net/?p=385#comment-5423 Why do you use there boost::shared_ptr instead of std::auto_ptr ? Why do you use there boost::shared_ptr instead of std::auto_ptr ?

]]>