I know that i should care about this when reading data from binary files and with networking but what with a source code? Does it matter for endianness if I assign value, for example int = 42? Will it compile on big endian machine with big endian ordering and then not work properly on little endian machine? Or compiler will take care of this?
Despite its reputation as a low-level language, when writing C++ code, you are not actually writing code for a concrete computer.
Instead, C++ code targets something called the C++ Abstract Machine, and it's the compiler's job to convert the behavior your program would cause on that Abstract Machine into a something (technically anything) that causes the same effect on the actual targeted machine.
In practice, most compilers bypass this and convert code semi-directly to the final target, but that's still the model they have to respect.
Endianness, amongst other things that we take for granted, is not a concept that exists within the C++ abstract machine. Unless you delve into working with raw data, as you pointed out, you don't have to worry about it.
As an extreme-ish example, take >>, the bit-shift operator. It has a well-defined behavior on the abstract machine for multi-byte types, so its behavior cannot depend on the endianness of the target. It's not the actual bits that are bit-shifted, it's the value.