diff options
Diffstat (limited to 'libstdc++-v3/docs/html/19_diagnostics/howto.html')
| -rw-r--r-- | libstdc++-v3/docs/html/19_diagnostics/howto.html | 89 |
1 files changed, 37 insertions, 52 deletions
diff --git a/libstdc++-v3/docs/html/19_diagnostics/howto.html b/libstdc++-v3/docs/html/19_diagnostics/howto.html index 7b072ea83b4..26ba4cde758 100644 --- a/libstdc++-v3/docs/html/19_diagnostics/howto.html +++ b/libstdc++-v3/docs/html/19_diagnostics/howto.html @@ -8,7 +8,7 @@ <META NAME="GENERATOR" CONTENT="vi and eight fingers"> <TITLE>libstdc++-v3 HOWTO: Chapter 19</TITLE> <LINK REL=StyleSheet HREF="../lib3styles.css"> -<!-- $Id: howto.html,v 1.1 2000/12/10 04:04:54 pme Exp $ --> +<!-- $Id: howto.html,v 1.2 2001/03/25 00:01:56 pme Exp $ --> </HEAD> <BODY> @@ -26,7 +26,7 @@ <UL> <LI><A HREF="#1">Adding data to exceptions</A> <LI><A HREF="#2">Exception class hierarchy diagram</A> - <LI><A HREF="#3">Concept checkers</A> + <LI><A HREF="#3">Concept checkers -- <STRONG>new and improved!</STRONG></A> </UL> <HR> @@ -67,58 +67,44 @@ </P> <HR> -<H2><A NAME="3">Concept checkers</A></H2> - <P>First the good news, then the bad news.</P> - <P><STRONG>Good news:</STRONG> As part of their 3.3 release, SGI - added some nifty macros which - perform assertions on type properties. For example, the Standard - requires that types passed as template parameters to <TT>vector</TT> - be "Assignable" (which means what you think it means). +<H2><A NAME="3">Concept checkers -- <STRONG>new and improved!</STRONG></A></H2> + <P>Better taste! Less fat! Literally!</P> + <P>In 1999, SGI added <EM>concept checkers</EM> to their implementation + of the STL: code which checked the template parameters of + instantiated pieces of the STL, in order to insure that the parameters + being used met the requirements of the standard. For example, + the Standard requires that types passed as template parameters to + <TT>vector</TT> be "Assignable" (which means what you think + it means). The checking was done during compilation, and none of + the code was executed at runtime. </P> - <P>The concept checkers allow the source code for <TT>vector</TT> to - declare - <PRE> - __STL_CLASS_REQUIRES(_Tp, _Assignable); - </PRE>inside the template. <TT>_Tp</TT> is the element type of the - vector, and <TT>_Assignable</TT> is the concept to be checked (it is - defined in some back-end header files). When you instantiate - <TT>vector<MyType></TT>, compile-time checking can be done on - whether MyType meets the requirements for vectors. + <P>Unfortunately, the size of the compiler files grew significantly + as a result. The checking code itself was cumbersome. And bugs + were found in it on more than one occasion. </P> - <P>Most (all?) of the containers and sequences are capable of performing - concept checking during compilation, not just vector. + <P>The primary author of the checking code, Jeremy Siek, had already + started work on a replcement implementation. The new code has been + formally reviewed and accepted into + <A HREF="http://www.boost.org/libs/concept_check/concept_check.htm">the + Boost libraries</A>, and we are pleased to incorporate it into the + GNU C++ library. </P> - <P>If a concept is violated (thus causing a compilation failure), the - error messages printed by the compiler will be of the form - <PRE> - _STL_ERROR::__<EM>foo</EM>_violation - </PRE> where <EM>foo</EM> is a description of the precise violation. - For example, if a type is required to support the preincrement - operator but doesn't, then you will see - _STL_ERROR::__postincrement_operator_requirement_violation, which - should give you a hint as to the nature of the problem. + <P>The new version imposes a much smaller space overhead on the generated + object file. The checks are also cleaner and easier to read and + understand. </P> - <P>Other names might be seen for more specific errors, for example, - _ERROR_IN_STL_CONTAINER::__begin_iterator_accessor_requirement_violation. + <P>Concept checking can be disabled when you build your code, for example, + to save space during a production build. Just define (via -D or + #define) any of the macros + <TT>_GLIBCPP_NO_CONCEPT_CHECKS (yes, with the leading underscore), + <TT>_STL_NO_CONCEPT_CHECKS</TT> (also with the leading underscore), + or <TT>NDEBUG</TT>. The first macro is specifically for this + feature, the second is the disabling macro for the replaced SGI + version (some code may assume SGI's version is in use), and the third + is the usual macro to disable <TT>assert()</TT>, which is often turned + off for production builds. </P> - <P>You will probably also see other errors as the malformed code is - actually used. The concept checking error messages should be printed - before the others, so start at the top and work your way down. - </P> - <P><STRONG>Bad news:</STRONG> The current checking code is somewhat - messy. It results in no runtime overhead, but there is a space - penalty (sometimes a very large one) in the generated code. And the - code itself has bugs. - </P> - <P>Concept checking can be disabled when you build your code. Just - define (via -D or #define) the macro - <TT>_STL_NO_CONCEPT_CHECKS</TT> (yes, with the leading underscore). - In fact, this may already be defined in the library by default, - depending on what decision we come to. - </P> - <P><STRONG>More good news:</STRONG> Replacement code has already been - written by the same author of the original code. It's available at - Boost and we hope to integrate it into the library. + <P>Return <A HREF="#top">to top of page</A> or <A HREF="../faq/index.html">to the FAQ</A>. </P> @@ -129,9 +115,8 @@ <HR> <P CLASS="fineprint"><EM> Comments and suggestions are welcome, and may be sent to -<A HREF="mailto:pme@sources.redhat.com">Phil Edwards</A> or -<A HREF="mailto:gdr@gcc.gnu.org">Gabriel Dos Reis</A>. -<BR> $Id: howto.html,v 1.1 2000/12/10 04:04:54 pme Exp $ +<A HREF="mailto:libstdc++@gcc.gnu.org">the mailing list</A>. +<BR> $Id: howto.html,v 1.2 2001/03/25 00:01:56 pme Exp $ </EM></P> |
