Arch Linux

?

moneymanagerex lists boost in it's makedepends array
From the src directory extracted with makepkg -Codd
$ git describe
v1.5.2
$ grep -r boost
3rd/rapidjson/include/rapidjson/rapidjson.h: \li http://www.boost.org/doc/libs/1_42_0/boost/detail/endian.hpp
3rd/rapidjson/include/rapidjson/rapidjson.h:// Adopt C++03 implementation from boost
3rd/rapidjson/include/rapidjson/msinttypes/stdint.h:// These #ifndef's are needed to prevent collisions with <boost/cstdint.hpp>.
3rd/rapidjson/include/rapidjson/msinttypes/stdint.h:// These #ifndef's are needed to prevent collisions with <boost/cstdint.hpp>.
3rd/fmt/test/gtest/gmock/gmock.h:types (like for example boost::function).
3rd/fmt/test/gtest/gtest/gtest.h:// boost::filesystem::path, whose iterator has a value_type that is equal to
3rd/fmt/test/gtest/gtest/gtest.h:// boost::filesystem::path.
3rd/fmt/README.rst:Boost Format 1.67 boost::format 7.24
3rd/fmt/doc/index.rst: a long time. I’ve used both <code>boost::format</code> and
3rd/fmt/ChangeLog.rst: to 5x speed up on long format strings and significant performance boost on
3rd/wxsqlite3/sqlite3secure/src/sqlite3.c:** show that method (3) with N==100 provides about a 5% performance boost for
3rd/wxsqlite3/sqlite3secure/src/sqlite3.c:** performance boost to bulk delete or insert operations. However, if
3rd/wxsqlite3/sqlite3secure/src/sqlite3.c: ** row without seeking the cursor. This can be a big performance boost.
3rd/wxsqlite3/sqlite3secure/src/sqlite3.c:** performance boost.
3rd/wxsqlite3/sqlite3secure/src/sqlite3.c: ** yield a significant performance boost.

Nothing includes any boost header or checks for any boost library. The matches are all comments.
Removing boost from makedepends does not produce any changes.

Adding rapidjson to makedepends and that appears to be used in place of the bundled copy.

PKGBUILD.diff removed boost from makedepends. Added rapidjson to makedepends. Added all git submodules to sources array and configure the git clone to use them in prepare().
Edit:
cmake --build . --target package
The custom target package calls makepkg within makepkg to use upstream's PKGBUILD to generate the package mmex?

Thanks for the details. Makes sense now. I've removed boost from depends.
I cannot be bothered with providing the submodules separately, as I'd have to take care of pinning them to a particular commit/version. So unless upstream takes care of that I'll just stick to the current PKGBUILD.
Cheers!