I tried to see what was xmake's value proposition, but even xmake's site fails to present anything. Writing something in Lua is far from a compelling reason to switch a build system.
I think it's mostly a cultural thing? I see most Chinese devs prefer xmake, probably out of solidarity because its devs are also Chinese.
As for me, I also don't see the appeal in xmake. CMake, coupled with ninja, just works and for many popular libraries there's a great chance that you can add them to your project simply by doing a find_package because of how popular CMake is. The build systems landscape in C++ is already messy enough and I don't see how introducing yet another CMake-like would improve things. What we need is more .cmake files for popular libraries.
and implement local caching, like ccache but supporting msvc as well as gcc/clang (actually msvc might be disabled at the moment)
.. likewise remote caching, like Mozilla's sscache for teams (again supporting msvc/gcc/clang)
it has its own package manager but also works with pretty much all other popular package managers
the build system will default to any matching installed system libs unless you say otherwise ({ system = false} on a per-included package basis)
it downloads and installed necessary packages when first building a project
package manager can use pre-compiled Windows artefacts
can use the hosted online repo, or our own online or local repos (multiple if we want)
location of alternative repos can be set in xmake.lua file or configured on command line (or config file).
can configure project with a TUI menu xmake f --menu
It just feels more straightforward and logical to me too. I know that's subjective, but the objective part of that is it's a unified build system, cache system and package manager that sets out to work with other tools.
it’s a build system itself, it doesn’t need make or whatever and compilation speeds are similar to ninja – which means it can do things like distributed building – and implement local caching, like ccache but supporting msvc as well as gcc/clang (actually msvc might be disabled at the moment) – … likewise remote caching, like Mozilla’s sscache for teams (again supporting msvc/gcc/clang)
So it offers nothing over cmake, and at best it matches features already offers for ages, such as transparent support for compiler cachês and distributed compilers.
it has its own package manager but also works with pretty much all other popular package managers –
That lies somewhere between matching cmake's features, and placing users in a losing proposition with its own package manager.
can configure project with a TUI menu xmake f --menu
Not only is this irrelevant but it is also something cmake provides.
Also, cmake is the de facto standard and supported natively by IDEs such as Visual Studio.
@lysdexic@leviosa It's always the same with "new" build systems: people think that a new language will make things much easier. But for complex build systems, they then realize how difficult such systems are, which cannot be hidden.
And about the syntax: that's not important.
Supposedly easier to use, with less warts than cmake. However it being developed in China makes it a security risk for me personally.
And even if it wasn't, I'd probably stick with cmake anyways due to cmake being the big standard
I currently find known C/C++ build systems and package managers lacking. Maybe this is caused by the high bar by Cargo.
Its nice to see something similar to Cargo in C/C++ ecosystem. vcpkg and conan are a pain in the ass in my opinion