The set of features available in the GNU C++ library is shaped by several GCC Command Options. Options that impact libstdc++ are enumerated and detailed in the table below. The standard library conforms to the dialect of C++ specified by the -std option passed to the compiler. By default, g++ is equivalent to g++ -std=gnu++14 since GCC 6, and g++ -std=gnu++98 for older releases. Option Flags Description -std=c++98 or -std=c++03 Use the 1998 ISO C++ standard plus amendments. -std=gnu++98 or -std=gnu++03 As directly above, with GNU extensions. -std=c++11 Use the 2011 ISO C++ standard. -std=gnu++11 As directly above, with GNU extensions. -std=c++14 Use the 2014 ISO C++ standard. -std=gnu++14 As directly above, with GNU extensions. -fexceptions See exception-free dialect -frtti As above, but RTTI-free dialect. -pthread For ISO C++11 <thread>, <future>, <mutex>, or <condition_variable>. -latomic Linking to libatomic is required for some uses of ISO C++11 <atomic>. -lstdc++fs Linking to libstdc++fs is required for use of the Filesystem library extensions in <experimental/filesystem>. -fopenmp For parallel mode.

The C++ standard specifies the entire set of header files that must be available to all hosted implementations. Actually, the word "files" is a misnomer, since the contents of the headers don't necessarily have to be in any kind of external file. The only rule is that when one #includes a header, the contents of that header become available, no matter how. That said, in practice files are used. There are two main types of include files: header files related to a specific version of the ISO C++ standard (called Standard Headers), and all others (TS, TR1, C++ ABI, and Extensions). Multiple dialects of standard headers are supported, corresponding to the 1998 standard as updated for 2003, the 2011 standard, the 2014 standard, and so on. and and show the C++98/03 include files. These are available in the C++98 compilation mode, i.e. -std=c++98 or -std=gnu++98. Unless specified otherwise below, they are also available in later modes (C++11, C++14 etc). algorithm bitset complex deque exception fstream functional iomanip ios iosfwd iostream istream iterator limits list locale map memory new numeric ostream queue set sstream stack stdexcept streambuf string utility typeinfo valarray vector cassert cerrno cctype cfloat ciso646 climits clocale cmath csetjmp csignal cstdarg cstddef cstdio cstdlib cstring ctime cwchar cwctype The following header is deprecated and might be removed from a future C++ standard. strstream and show the C++11 include files. These are available in C++11 compilation mode, i.e. -std=c++11 or -std=gnu++11. Including these headers in C++98/03 mode may result in compilation errors. Unless specified otherwise below, they are also available in later modes (C++14 etc). array atomic chrono codecvt condition_variable forward_list future initalizer_list mutex random ratio regex scoped_allocator system_error thread tuple typeindex type_traits unordered_map unordered_set ccomplex cfenv cinttypes cstdalign cstdbool cstdint ctgmath cuchar shows the C++14 include file. This is available in C++14 compilation mode, i.e. -std=c++14 or -std=gnu++14. Including this header in C++98/03 mode or C++11 will not result in compilation errors, but will not define anything. Unless specified otherwise below, it is also available in later modes (C++17 etc). shared_mutex shows the C++17 include files. These are available in C++17 compilation mode, i.e. -std=c++17 or -std=gnu++17. Including these headers in earlier modes will not result in compilation errors, but will not define anything. Unless specified otherwise below, they are also available in later modes (C++20 etc). any optional string_view variant , shows the additional include file define by the File System Technical Specification, ISO/IEC TS 18822. This is available in C++11 and later compilation modes. Including this header in earlier modes will not result in compilation errors, but will not define anything. experimental/filesystem , shows the additional include files define by the C++ Extensions for Library Fundamentals Technical Specification, ISO/IEC TS 19568. These are available in C++14 and later compilation modes. Including these headers in earlier modes will not result in compilation errors, but will not define anything. experimental/algorithm experimental/any experimental/array experimental/chrono experimental/deque experimental/forward_list experimental/functional experimental/iterator experimental/list experimental/map experimental/memory experimental/memory_resource experimental/numeric experimental/optional experimental/propagate_const experimental/random experimental/ratio experimental/regex experimental/set experimental/source_location experimental/string experimental/string_view experimental/system_error experimental/tuple experimental/type_traits experimental/unordered_map experimental/unordered_set experimental/utility experimental/vector In addition, TR1 includes as: tr1/array tr1/complex tr1/memory tr1/functional tr1/random tr1/regex tr1/tuple tr1/type_traits tr1/unordered_map tr1/unordered_set tr1/utility tr1/ccomplex tr1/cfenv tr1/cfloat tr1/cmath tr1/cinttypes tr1/climits tr1/cstdarg tr1/cstdbool tr1/cstdint tr1/cstdio tr1/cstdlib tr1/ctgmath tr1/ctime tr1/cwchar tr1/cwctype Decimal floating-point arithmetic is available if the C++ compiler supports scalar decimal floating-point types defined via __attribute__((mode(SD|DD|LD))). decimal/decimal Also included are files for the C++ ABI interface: cxxabi.hcxxabi_forced.h And a large variety of extensions. ext/algorithm ext/atomicity.h ext/array_allocator.h ext/bitmap_allocator.h ext/cast.h ext/codecvt_specializations.h ext/concurrence.h ext/debug_allocator.h ext/enc_filebuf.h ext/extptr_allocator.h ext/functional ext/iterator ext/malloc_allocator.h ext/memory ext/mt_allocator.h ext/new_allocator.h ext/numeric ext/numeric_traits.h ext/pb_ds/assoc_container.h ext/pb_ds/priority_queue.h ext/pod_char_traits.h ext/pool_allocator.h ext/rb_tree ext/rope ext/slist ext/stdio_filebuf.h ext/stdio_sync_filebuf.h ext/throw_allocator.h ext/typelist.h ext/type_traits.h ext/vstring.h debug/array debug/bitset debug/deque debug/forward_list debug/list debug/map debug/set debug/string debug/unordered_map debug/unordered_set debug/vector profile/bitset profile/deque profile/list profile/map profile/set profile/unordered_map profile/unordered_set profile/vector parallel/algorithm parallel/numeric

A few simple rules. First, mixing different dialects of the standard headers is not possible. It's an all-or-nothing affair. Thus, code like #include <array> #include <functional> Implies C++11 mode. To use the entities in <array>, the C++11 compilation mode must be used, which implies the C++11 functionality (and deprecations) in <functional> will be present. Second, the other headers can be included with either dialect of the standard headers, although features and types specific to C++11 are still only enabled when in C++11 compilation mode. So, to use rvalue references with __gnu_cxx::vstring, or to use the debug-mode versions of std::unordered_map, one must use the std=gnu++11 compiler flag. (Or std=c++11, of course.) A special case of the second rule is the mixing of TR1 and C++11 facilities. It is possible (although not especially prudent) to include both the TR1 version and the C++11 version of header in the same translation unit: #include <tr1/type_traits> #include <type_traits> Several parts of C++11 diverge quite substantially from TR1 predecessors.

The standard specifies that if one includes the C-style header (<math.h> in this case), the symbols will be available in the global namespace and perhaps in namespace std:: (but this is no longer a firm requirement.) On the other hand, including the C++-style header (<cmath>) guarantees that the entities will be found in namespace std and perhaps in the global namespace. Usage of C++-style headers is recommended, as then C-linkage names can be disambiguated by explicit qualification, such as by std::abort. In addition, the C++-style headers can use function overloading to provide a simpler interface to certain families of C-functions. For instance in <cmath>, the function std::sin has overloads for all the builtin floating-point types. This means that std::sin can be used uniformly, instead of a combination of std::sinf, std::sin, and std::sinl.

There are three base header files that are provided. They can be used to precompile the standard headers and extensions into binary files that may then be used to speed up compilations that use these headers. stdc++.h Includes all standard headers. Actual content varies depending on language dialect. stdtr1c++.h Includes all of <stdc++.h>, and adds all the TR1 headers. extc++.h Includes all of <stdc++.h>, and adds all the Extension headers (and in C++98 mode also adds all the TR1 headers by including all of <stdtr1c++.h>). To construct a .gch file from one of these base header files, first find the include directory for the compiler. One way to do this is: g++ -v hello.cc #include <...> search starts here: /mnt/share/bld/H-x86-gcc.20071201/include/c++/4.3.0 ... End of search list. Then, create a precompiled header file with the same flags that will be used to compile other projects. g++ -Winvalid-pch -x c++-header -g -O2 -o ./stdc++.h.gch /mnt/share/bld/H-x86-gcc.20071201/include/c++/4.3.0/x86_64-unknown-linux-gnu/bits/stdc++.h The resulting file will be quite large: the current size is around thirty megabytes. How to use the resulting file. g++ -I. -include stdc++.h -H -g -O2 hello.cc Verification that the PCH file is being used is easy: g++ -Winvalid-pch -I. -include stdc++.h -H -g -O2 hello.cc -o test.exe ! ./stdc++.h.gch . /mnt/share/bld/H-x86-gcc.20071201/include/c++/4.3.0/iostream . /mnt/share/bld/H-x86-gcc.20071201include/c++/4.3.0/string The exclamation point to the left of the stdc++.h.gch listing means that the generated PCH file was used. Detailed information about creating precompiled header files can be found in the GCC documentation.

All library macros begin with _GLIBCXX_. Furthermore, all pre-processor macros, switches, and configuration options are gathered in the file c++config.h, which is generated during the libstdc++ configuration and build process. This file is then included when needed by files part of the public libstdc++ API, like <ios>. Most of these macros should not be used by consumers of libstdc++, and are reserved for internal implementation use. These macros cannot be redefined. A select handful of macros control libstdc++ extensions and extra features, or provide versioning information for the API. Only those macros listed below are offered for consideration by the general public. Below are the macros which users may check for library version information. _GLIBCXX_RELEASE The major release number for libstdc++. This macro is defined to the GCC major version that the libstdc++ headers belong to, as an integer constant. When compiling with GCC it has the same value as GCC's pre-defined macro __GNUC__. This macro can be used when libstdc++ is used with a non-GNU compiler where __GNUC__ is not defined, or has a different value that doesn't correspond to the libstdc++ version. This macro first appeared in the GCC 7.1 release and is not defined for GCC 6.x or older releases. __GLIBCXX__ The revision date of the libstdc++ source code, in compressed ISO date format, as an unsigned long. For notes about using this macro and details on the value of this macro for a particular release, please consult the ABI History appendix. Below are the macros which users may change with #define/#undef or with -D/-U compiler flags. The default state of the symbol is listed. Configurable (or Not configurable) means that the symbol is initially chosen (or not) based on --enable/--disable options at library build and configure time (documented in Configure), with the various --enable/--disable choices being translated to #define/#undef). ABI means that changing from the default value may mean changing the ABI of compiled code. In other words, these choices control code which has already been compiled (i.e., in a binary such as libstdc++.a/.so). If you explicitly #define or #undef these macros, the headers may see different code paths, but the libraries which you link against will not. Experimenting with different values with the expectation of consistent linkage requires changing the config headers before building/installing the library. _GLIBCXX_USE_DEPRECATED Defined by default. Not configurable. ABI-changing. Turning this off removes older ARM-style iostreams code, and other anachronisms from the API. This macro is dependent on the version of the standard being tracked, and as a result may give different results for -std=c++98 and -std=c++11. This may be useful in updating old C++ code which no longer meet the requirements of the language, or for checking current code against new language standards. _GLIBCXX_USE_CXX11_ABI Defined to the value 1 by default. Configurable via --disable-libstdcxx-dual-abi and/or --with-default-libstdcxx-abi. ABI-changing. When defined to a non-zero value the library headers will use the new C++11-conforming ABI introduced in GCC 5, rather than the older ABI introduced in GCC 3.4. This changes the definition of several class templates, including std:string, std::list and some locale facets. For more details see . _GLIBCXX_CONCEPT_CHECKS Undefined by default. Configurable via --enable-concept-checks. When defined, performs compile-time checking on certain template instantiations to detect violations of the requirements of the standard. This macro has no effect for freestanding implementations. This is described in more detail in Compile Time Checks. _GLIBCXX_ASSERTIONS Undefined by default. When defined, enables extra error checking in the form of precondition assertions, such as bounds checking in strings and null pointer checks when dereferencing smart pointers. _GLIBCXX_DEBUG Undefined by default. When defined, compiles user code using the debug mode. When defined, _GLIBCXX_ASSERTIONS is defined automatically, so all the assertions enabled by that macro are also enabled in debug mode. _GLIBCXX_DEBUG_PEDANTIC Undefined by default. When defined while compiling with the debug mode, makes the debug mode extremely picky by making the use of libstdc++ extensions and libstdc++-specific behavior into errors. _GLIBCXX_PARALLEL Undefined by default. When defined, compiles user code using the parallel mode. _GLIBCXX_PARALLEL_ASSERTIONS Undefined by default, but when any parallel mode header is included this macro will be defined to a non-zero value if _GLIBCXX_ASSERTIONS has a non-zero value, otherwise to zero. When defined to a non-zero value, it enables extra error checking and assertions in the parallel mode. _GLIBCXX_PROFILE Undefined by default. When defined, compiles user code using the profile mode. __STDCPP_WANT_MATH_SPEC_FUNCS__ Undefined by default. When defined to a non-zero integer constant, enables support for ISO/IEC 29124 Special Math Functions.

In the GCC 5.1 release libstdc++ introduced a new library ABI that includes new implementations of std::string and std::list. These changes were necessary to conform to the 2011 C++ standard which forbids Copy-On-Write strings and requires lists to keep track of their size. In order to maintain backwards compatibility for existing code linked to libstdc++ the library's soname has not changed and the old implementations are still supported in parallel with the new ones. This is achieved by defining the new implementations in an inline namespace so they have different names for linkage purposes, e.g. the new version of std::list<int> is actually defined as std::__cxx11::list<int>. Because the symbols for the new implementations have different names the definitions for both versions can be present in the same library. The _GLIBCXX_USE_CXX11_ABI macro (see ) controls whether the declarations in the library headers use the old or new ABI. So the decision of which ABI to use can be made separately for each source file being compiled. Using the default configuration options for GCC the default value of the macro is 1 which causes the new ABI to be active, so to use the old ABI you must explicitly define the macro to 0 before including any library headers. (Be aware that some GNU/Linux distributions configure GCC 5 differently so that the default value of the macro is 0 and users must define it to 1 to enable the new ABI.) Although the changes were made for C++11 conformance, the choice of ABI to use is independent of the -std option used to compile your code, i.e. for a given GCC build the default value of the _GLIBCXX_USE_CXX11_ABI macro is the same for all dialects. This ensures that the -std does not change the ABI, so that it is straightforward to link C++03 and C++11 code together. Because std::string is used extensively throughout the library a number of other types are also defined twice, including the stringstream classes and several facets used by std::locale. The standard facets which are always installed in a locale may be present twice, with both ABIs, to ensure that code like std::use_facet<std::time_get<char>>(locale); will work correctly for both std::time_get and std::__cxx11::time_get (even if a user-defined facet that derives from one or other version of time_get is installed in the locale). Although the standard exception types defined in <stdexcept> use strings, most are not defined twice, so that a std::out_of_range exception thrown in one file can always be caught by a suitable handler in another file, even if the two files are compiled with different ABIs. One exception type does change when using the new ABI, namely std::ios_base::failure. This is necessary because the 2011 standard changed its base class from std::exception to std::system_error, which causes its layout to change. Exceptions due to iostream errors are thrown by a function inside libstdc++.so, so whether the thrown exception uses the old std::ios_base::failure type or the new one depends on the ABI that was active when libstdc++.so was built, not the ABI active in the user code that is using iostreams. This means that for a given build of GCC the type thrown is fixed. In current releases the library throws a special type that can be caught by handlers for either the old or new type, but for GCC 7.1, 7.2 and 7.3 the library throws the new std::ios_base::failure type, and for GCC 5.x and 6.x the library throws the old type. Catch handlers of type std::ios_base::failure will only catch the exceptions if using a newer release, or if the handler is compiled with the same ABI as the type thrown by the library. Handlers for std::exception will always catch iostreams exceptions, because the old and new type both inherit from std::exception.

If you get linker errors about undefined references to symbols that involve types in the std::__cxx11 namespace or the tag [abi:cxx11] then it probably indicates that you are trying to link together object files that were compiled with different values for the _GLIBCXX_USE_CXX11_ABI macro. This commonly happens when linking to a third-party library that was compiled with an older version of GCC. If the third-party library cannot be rebuilt with the new ABI then you will need to recompile your code with the old ABI. Not all uses of the new ABI will cause changes in symbol names, for example a class with a std::string member variable will have the same mangled name whether compiled with the old or new ABI. In order to detect such problems the new types and functions are annotated with the abi_tag attribute, allowing the compiler to warn about potential ABI incompatibilities in code using them. Those warnings can be enabled with the -Wabi-tag option.

There are three main namespaces. std The ISO C++ standards specify that "all library entities are defined within namespace std." This includes namespaces nested within namespace std, such as namespace std::chrono. abi Specified by the C++ ABI. This ABI specifies a number of type and function APIs supplemental to those required by the ISO C++ Standard, but necessary for interoperability. __gnu_ Indicating one of several GNU extensions. Choices include __gnu_cxx, __gnu_debug, __gnu_parallel, and __gnu_pbds. The library uses a number of inline namespaces as implementation details that are not intended for users to refer to directly, these include std::__detail, std::__cxx11 and std::_V2. A complete list of implementation namespaces (including namespace contents) is available in the generated source documentation.

One standard requirement is that the library components are defined in namespace std::. Thus, in order to use these types or functions, one must do one of two things: put a kind of using-declaration in your source (either using namespace std; or i.e. using std::string;) This approach works well for individual source files, but should not be used in a global context, like header files. use a fully qualified name for each library symbol (i.e. std::string, std::cout) Always can be used, and usually enhanced, by strategic use of typedefs. (In the cases where the qualified verbiage becomes unwieldy.)

Best practice in programming suggests sequestering new data or functionality in a sanely-named, unique namespace whenever possible. This is considered an advantage over dumping everything in the global namespace, as then name look-up can be explicitly enabled or disabled as above, symbols are consistently mangled without repetitive naming prefixes or macros, etc. For instance, consider a project that defines most of its classes in namespace gtk. It is possible to adapt namespace gtk to namespace std by using a C++-feature called namespace composition. This is what happens if a using-declaration is put into a namespace-definition: the imported symbol(s) gets imported into the currently active namespace(s). For example: namespace gtk { using std::string; using std::tr1::array; class Window { ... }; } In this example, std::string gets imported into namespace gtk. The result is that use of std::string inside namespace gtk can just use string, without the explicit qualification. As an added bonus, std::string does not get imported into the global namespace. Additionally, a more elaborate arrangement can be made for backwards compatibility and portability, whereby the using-declarations can wrapped in macros that are set based on autoconf-tests to either "" or i.e. using std::string; (depending on whether the system has libstdc++ in std:: or not). (ideas from Llewelly and Karl Nelson)

Or as close as it gets: freestanding. This is a minimal configuration, with only partial support for the standard library. Assume only the following header files can be used: cstdarg cstddef cstdlib exception limits new exception typeinfo In addition, throw in cxxabi.h. In the C++11 dialect add initializer_list type_traits There exists a library that offers runtime support for just these headers, and it is called libsupc++.a. To use it, compile with gcc instead of g++, like so: gcc foo.cc -lsupc++ No attempt is made to verify that only the minimal subset identified above is actually used at compile time. Violations are diagnosed as undefined symbols at link time.

If the only library built is the static library (libstdc++.a), or if specifying static linking, this section is can be skipped. But if building or using a shared library (libstdc++.so), then additional location information will need to be provided. But how? A quick read of the relevant part of the GCC manual, Compiling C++ Programs, specifies linking against a C++ library. More details from the GCC FAQ, which states GCC does not, by default, specify a location so that the dynamic linker can find dynamic libraries at runtime. Users will have to provide this information. Methods vary for different platforms and different styles, and are printed to the screen during installation. To summarize: At runtime set LD_LIBRARY_PATH in your environment correctly, so that the shared library for libstdc++ can be found and loaded. Be certain that you understand all of the other implications and behavior of LD_LIBRARY_PATH first. Compile the path to find the library at runtime into the program. This can be done by passing certain options to g++, which will in turn pass them on to the linker. The exact format of the options is dependent on which linker you use: GNU ld (default on GNU/Linux): -Wl,-rpath,destdir/lib Solaris ld: -Wl,-Rdestdir/lib Some linkers allow you to specify the path to the library by setting LD_RUN_PATH in your environment when linking. On some platforms the system administrator can configure the dynamic linker to always look for libraries in destdir/lib, for example by using the ldconfig utility on GNU/Linux or the crle utility on Solaris. This is a system-wide change which can make the system unusable so if you are unsure then use one of the other methods described above. Use the ldd utility on the linked executable to show which libstdc++.so library the system will get at runtime. A libstdc++.la file is also installed, for use with Libtool. If you use Libtool to create your executables, these details are taken care of for you.

GCC 5.3 includes an implementation of the Filesystem library defined by the technical specification ISO/IEC TS 18822:2015. Because this is an experimental library extension, not part of the C++ standard, it is implemented in a separate library, libstdc++fs.a, and there is no shared library for it. To use the library you should include <experimental/filesystem> and link with -lstdc++fs. The library implementation is incomplete on non-POSIX platforms, specifically Windows support is rudimentary. Due to the experimental nature of the Filesystem library the usual guarantees about ABI stability and backwards compatibility do not apply to it. There is no guarantee that the components in any <experimental/xxx> header will remain compatible between different GCC releases.

This section discusses issues surrounding the proper compilation of multithreaded applications which use the Standard C++ library. This information is GCC-specific since the C++ standard does not address matters of multithreaded applications.

All normal disclaimers aside, multithreaded C++ application are only supported when libstdc++ and all user code was built with compilers which report (via gcc/g++ -v ) the same thread model and that model is not single. As long as your final application is actually single-threaded, then it should be safe to mix user code built with a thread model of single with a libstdc++ and other C++ libraries built with another thread model useful on the platform. Other mixes may or may not work but are not considered supported. (Thus, if you distribute a shared C++ library in binary form only, it may be best to compile it with a GCC configured with --enable-threads for maximal interchangeability and usefulness with a user population that may have built GCC with either --enable-threads or --disable-threads.) When you link a multithreaded application, you will probably need to add a library or flag to g++. This is a very non-standardized area of GCC across ports. Some ports support a special flag (the spelling isn't even standardized yet) to add all required macros to a compilation (if any such flags are required then you must provide the flag for all compilations not just linking) and link-library additions and/or replacements at link time. The documentation is weak. On several targets (including GNU/Linux, Solaris and various BSDs) -pthread is honored. Some other ports use other switches. This is not well documented anywhere other than in "gcc -dumpspecs" (look at the 'lib' and 'cpp' entries). Some uses of std::atomic also require linking to libatomic.

In the terms of the 2011 C++ standard a thread-safe program is one which does not perform any conflicting non-atomic operations on memory locations and so does not contain any data races. The standard places requirements on the library to ensure that no data races are caused by the library itself or by programs which use the library correctly (as described below). The C++11 memory model and library requirements are a more formal version of the SGI STL definition of thread safety, which the library used prior to the 2011 standard. The library strives to be thread-safe when all of the following conditions are met: The system's libc is itself thread-safe, The compiler in use reports a thread model other than 'single'. This can be tested via output from gcc -v. Multi-thread capable versions of gcc output something like this: %gcc -v Using built-in specs. ... Thread model: posix gcc version 4.1.2 20070925 (Red Hat 4.1.2-33) Look for "Thread model" lines that aren't equal to "single." Requisite command-line flags are used for atomic operations and threading. Examples of this include -pthread and -march=native, although specifics vary depending on the host environment. See Command Options and Machine Dependent Options. An implementation of the atomicity.h functions exists for the architecture in question. See the internals documentation for more details. The user code must guard against concurrent function calls which access any particular library object's state when one or more of those accesses modifies the state. An object will be modified by invoking a non-const member function on it or passing it as a non-const argument to a library function. An object will not be modified by invoking a const member function on it or passing it to a function as a pointer- or reference-to-const. Typically, the application programmer may infer what object locks must be held based on the objects referenced in a function call and whether the objects are accessed as const or non-const. Without getting into great detail, here is an example which requires user-level locks: library_class_a shared_object_a; void thread_main () { library_class_b *object_b = new library_class_b; shared_object_a.add_b (object_b); // must hold lock for shared_object_a shared_object_a.mutate (); // must hold lock for shared_object_a } // Multiple copies of thread_main() are started in independent threads. Under the assumption that object_a and object_b are never exposed to another thread, here is an example that does not require any user-level locks: void thread_main () { library_class_a object_a; library_class_b *object_b = new library_class_b; object_a.add_b (object_b); object_a.mutate (); } All library types are safe to use in a multithreaded program if objects are not shared between threads or as long each thread carefully locks out access by any other thread while it modifies any object visible to another thread. Unless otherwise documented, the only exceptions to these rules are atomic operations on the types in <atomic> and lock/unlock operations on the standard mutex types in <mutex>. These atomic operations allow concurrent accesses to the same object without introducing data races. The following member functions of standard containers can be considered to be const for the purposes of avoiding data races: begin, end, rbegin, rend, front, back, data, find, lower_bound, upper_bound, equal_range, at and, except in associative or unordered associative containers, operator[]. In other words, although they are non-const so that they can return mutable iterators, those member functions will not modify the container. Accessing an iterator might cause a non-modifying access to the container the iterator refers to (for example incrementing a list iterator must access the pointers between nodes, which are part of the container and so conflict with other accesses to the container). Programs which follow the rules above will not encounter data races in library code, even when using library types which share state between distinct objects. In the example below the shared_ptr objects share a reference count, but because the code does not perform any non-const operations on the globally-visible object, the library ensures that the reference count updates are atomic and do not introduce data races: std::shared_ptr<int> global_sp; void thread_main() { auto local_sp = global_sp; // OK, copy constructor's parameter is reference-to-const int i = *global_sp; // OK, operator* is const int j = *local_sp; // OK, does not operate on global_sp // *global_sp = 2; // NOT OK, modifies int visible to other threads // *local_sp = 2; // NOT OK, modifies int visible to other threads // global_sp.reset(); // NOT OK, reset is non-const local_sp.reset(); // OK, does not operate on global_sp } int main() { global_sp.reset(new int(1)); std::thread t1(thread_main); std::thread t2(thread_main); t1.join(); t2.join(); } For further details of the C++11 memory model see Hans-J. Boehm's Threads and memory model for C++ pages, particularly the introduction and FAQ.

This gets a bit tricky. Please read carefully, and bear with me.

A wrapper type called __basic_file provides our abstraction layer for the std::filebuf classes. Nearly all decisions dealing with actual input and output must be made in __basic_file. A generic locking mechanism is somewhat in place at the filebuf layer, but is not used in the current code. Providing locking at any higher level is akin to providing locking within containers, and is not done for the same reasons (see the links above).

The __basic_file type is simply a collection of small wrappers around the C stdio layer (again, see the link under Structure). We do no locking ourselves, but simply pass through to calls to fopen, fwrite, and so forth. So, for 3.0, the question of "is multithreading safe for I/O" must be answered with, "is your platform's C library threadsafe for I/O?" Some are by default, some are not; many offer multiple implementations of the C library with varying tradeoffs of threadsafety and efficiency. You, the programmer, are always required to take care with multiple threads. (As an example, the POSIX standard requires that C stdio FILE* operations are atomic. POSIX-conforming C libraries (e.g, on Solaris and GNU/Linux) have an internal mutex to serialize operations on FILE*s. However, you still need to not do stupid things like calling fclose(fs) in one thread followed by an access of fs in another.) So, if your platform's C library is threadsafe, then your fstream I/O operations will be threadsafe at the lowest level. For higher-level operations, such as manipulating the data contained in the stream formatting classes (e.g., setting up callbacks inside an std::ofstream), you need to guard such accesses like any other critical shared resource.

A second choice may be available for I/O implementations: libio. This is disabled by default, and in fact will not currently work due to other issues. It will be revisited, however. The libio code is a subset of the guts of the GNU libc (glibc) I/O implementation. When libio is in use, the __basic_file type is basically derived from FILE. (The real situation is more complex than that... it's derived from an internal type used to implement FILE. See libio/libioP.h to see scary things done with vtbls.) The result is that there is no "layer" of C stdio to go through; the filebuf makes calls directly into the same functions used to implement fread, fwrite, and so forth, using internal data structures. (And when I say "makes calls directly," I mean the function is literally replaced by a jump into an internal function. Fast but frightening. *grin*) Also, the libio internal locks are used. This requires pulling in large chunks of glibc, such as a pthreads implementation, and is one of the issues preventing widespread use of libio as the libstdc++ cstdio implementation. But we plan to make this work, at least as an option if not a future default. Platforms running a copy of glibc with a recent-enough version will see calls from libstdc++ directly into the glibc already installed. For other platforms, a copy of the libio subsection will be built and included in libstdc++.

Don't forget that other cstdio implementations are possible. You could easily write one to perform your own forms of locking, to solve your "interesting" problems.

This section discusses issues surrounding the design of multithreaded applications which use Standard C++ containers. All information in this section is current as of the gcc 3.0 release and all later point releases. Although earlier gcc releases had a different approach to threading configuration and proper compilation, the basic code design rules presented here were similar. For information on all other aspects of multithreading as it relates to libstdc++, including details on the proper compilation of threaded code (and compatibility between threaded and non-threaded code), see Chapter 17. Two excellent pages to read when working with the Standard C++ containers and threads are SGI's http://www.sgi.com/tech/stl/thread_safety.html and SGI's http://www.sgi.com/tech/stl/Allocators.html. However, please ignore all discussions about the user-level configuration of the lock implementation inside the STL container-memory allocator on those pages. For the sake of this discussion, libstdc++ configures the SGI STL implementation, not you. This is quite different from how gcc pre-3.0 worked. In particular, past advice was for people using g++ to explicitly define _PTHREADS or other macros or port-specific compilation options on the command line to get a thread-safe STL. This is no longer required for any port and should no longer be done unless you really know what you are doing and assume all responsibility. Since the container implementation of libstdc++ uses the SGI code, we use the same definition of thread safety as SGI when discussing design. A key point that beginners may miss is the fourth major paragraph of the first page mentioned above (For most clients...), which points out that locking must nearly always be done outside the container, by client code (that'd be you, not us). There is a notable exceptions to this rule. Allocators called while a container or element is constructed uses an internal lock obtained and released solely within libstdc++ code (in fact, this is the reason STL requires any knowledge of the thread configuration). For implementing a container which does its own locking, it is trivial to provide a wrapper class which obtains the lock (as SGI suggests), performs the container operation, and then releases the lock. This could be templatized to a certain extent, on the underlying container and/or a locking mechanism. Trying to provide a catch-all general template solution would probably be more trouble than it's worth. The library implementation may be configured to use the high-speed caching memory allocator, which complicates thread safety issues. For all details about how to globally override this at application run-time see here. Also useful are details on allocator options and capabilities.