3 ms·
Let's make the comparison with pkg-config. If a library (either as source, or as a compiled package) provides a (correct) pkg-config file, then using that libr
by PaulDavisThe1st 7y ago
Let's make the comparison with pkg-config.
If a library (either as source, or as a compiled package) provides a (correct) pkg-config file, then using that library from within a project is pretty easy, assuming the build system knows how to use pkg-config files (we use waf, and it does).
But if the library doesn't provide a pkg-config file, or provides a broken one (yep, that happens), then we either have to create one or use a different method to tell the build system where to find and how to use the library.
We don't really want a package manager per se, because we're cross-platform, and package managers for compiled libraries are almost certainly going to be platform specific.
So for our release builds, and for some of our developers, we need to start from the source code, build the library using flags we provide (because we know what we want and what we're doing), install the library (somewhere) and only then set about using it from within our project. That process has to work more or less identically for our linux, macOS and mingw/windows builds (we x-compile for windows).
Consequently, even if there's a pkg-config file and even if it's contents are correct, none of that addresses (1) the version of the library (2) the compile time flags for the library (3) actually downloading the library (4) actually compiling and linking it. It does help with "installation" because the pkg-config file goes to a known location even if the installation is somewhere non-standard.
So in summary I would say that even if libman became wildly successful, and was used by the dozens of C libraries that we use in addition to the C++ ones, it would really only function as a replacement for pkg-config and would not address the harder, bigger problems we have with 3rd party library dependencies.