In some cases some reluctance is certainly justified (like here for example), but some projects seem openly hostile to the idea. However, I also see great resistance from many open-source projects to adopt better cross-platform tooling such as CMake. With all that out of the way, I do understand OpenSSL is a huge and complex project and it is used on many platforms so the change of the build system would have to be announced well in advance and planned carfully, not to mention require great effort. Compiling with MSVC compiler produces executables which depend on standard VC Runtime packages which come in the form of nicely packaged redistributables which can be easily updated and do not require manually packing and installing additional runtime DLLs, from another compiler not to mention that you shouldn't mix libraries and executables built with different compilers unless you really know what you are doing. You can build OpenSSL from source with MinGW.ĭevelopers in Windows shops predominantly use Visual Studio. There's nothing stopping people from building CMake there. Regarding those platforms, I only see the explicit mention of OpenVMS and I am not sure if you were refering to the x86-64 port, but if so, that port uses Clang as its official compiler which fully supports C++11 since version 3.3. What's the point of having secure communication if the system using it can be compromised to get to unencrypted data at rest or from memory? Then perhaps it's time to deprecate such platforms? How can you build security software on outdated platforms which probably don't have many hardware security features, and which aren't keeping their compiler and runtime (and who knows what else) up-to-date? Sure you can run tests on your own code and prove it is secure, but the platform itself, the OS, libraries, and software might not be. There are platforms our stuff builds on that don't have C++11, which CMake requires to build, last I checked. Unlike the current system which, at least under Windows, uses nmake and compiles one file at a time, CMake could do parallel builds which would greatly speed up OpenSSL compilation under Windows.ĬMake can generate Visual Studio project and solution files which use native build tools, just like it can generate makefiles for other platforms.ĬMake supports in-place and out-of-place builds, and can therefore support multiple builds from a single source tree. You can do that by calling an API such as SecureZeroMemory on Windows or memset_s from the C11 standard which are both guaranteed to not be optimized away, and I am pretty sure other platforms should have their alternatives provided either by the OS or by the compiler runtime.ĬMake does much more, to name just a few advantages: I came here from another issue (linked above), and I believe I have something constructive to add to this discussion.Īnd securely clearing private keys (how to stop clever compilers from optimizing out memory zeroization).

Openssl cmake windows