Closed OleksandrKvl closed 6 months ago
@OleksandrKvl Thank you for submitting this issue! We will inestigate it and deal with it in the subsequent releases.
We do not use Conan with our testing, and, unfortunately, can not guarantee compatibility with any given framework. As such, We're closing this issue.
I didn't ask you to do any Conan-specific change but to use things in a more standard way to make it more friendly to whatever approach people use.
Seems that RTSDK makes some assumptions which makes it hard to integrate with our internal code base that uses Conan for dependency management. Below are the issues I've got so far.
addExternal_zlib.cmake:501
relies onZLIB_VERSION_STRING
variable which is legacy/deprecated. You should useZLIB_VERSION
instead which is provided by the standardfind_package()
call (or at least support both).ZLIB_VERSION_STRING
seems to be provided only byFindZLIB
which is not the only way to get zlib, it doesn't work with Conan because it sets only standard variables.$<TARGET_FILE:ZLIB::ZLIB>
. This doesn't work with Conan because it provideszlib
as anINTERFACE
library which doesn't have any direct artifacts. It's not clear why client even need thatlibrssl_tmp.arch
, if this is a part of some installation routine then it should not be performed during configuration/build steps.For now I've solved those problems simply by installing RTSDK (using
FetchContent
) before including Conan generated files. But this works only because we are using old way of consuming Conan dependencies viainclude(${CMAKE_BINARY_DIR}/conan_paths.cmake)
directly in our CMake file. Modern way is to consume it via CMake toolchain file. In that case there's simply no way for the above solution. Honestly, I don't know if everything will work as expected in this project because now I got some dependencies from RTSDK and I'm not sure how well they will work with the rest of the project. Please consider to use more canonical ways of consuming dependencies to be friendly to different approaches.