This repository contains the source code and build system for an example implementation of a std::pmr
compatible polymorphic memory resource, intended for use in instrumented tests. The paper which describes
the motivation, operation, and other aspects of this code can be found in the
WG21 paper repository.
- CMake
- MSVC 2017 or gcc-9
At the time of writing the Microsoft Visual Studio 2017 and gcc-9 (trunk) are the only compilers that implement std::pmr as specified
in the C++17 specification, so those are the only to toolchains that have been tested (it's likely to compatible with more recent
versions of those compilers)
The most relevant instructions for installing gcc (at the time of writing, 2019.Apr.20): See https://gcc.gnu.org/wiki/InstallingGCC and replace gcc-4.6.0.tar.gz with a snapshot such as https://gcc.gnu.org/pub/gcc/snapshots/LATEST-9/
libc++ (i.e. the native library for clang) does not yet have a standard pmr implementation. I was able to compile the code in this repository using <experimental/memory_resource> that libc++ implements currently (May 2019).
The repository consists of 5 major parts:
- supportlib -- macros and printing helpers (static lib)
- stdpmr -- the implementations of the proposed types and the exception testing algorithm (static lib)
- pstring -- a series of examples of testing and fixing an imaginary (and quite pathological) string class (executables)
- exception_testing -- an example using the
exception_test_loop - patchpmr -- hacks to make clang with libc++ and older GNU libraries with experimental support work
Please read the paper, or watch the presentation, to better understand the repository contents.
Please use the master branch of this repository unless you are following a presentation from a conference (or C++ standardization meeting), in which case I suggest checking out the corresponding tag.
Git tag: R0 -- Corresponds to the paper revision P1160R0
Git tag: CppNow2019 -- added gcc support, reorganized the code, switched to cmake
Git tag: CppCon2019 -- fixed handling of 0 sized allocations to the standard way (allocating and not returning 0)
If you find any issue or have any concerns please do not hesitate to open an issue.
Pull requests are welcome, however since this code accompanies a paper, changes that would change the essence of the proposal would be better be addressed during WG21 discussions.