Investigate value of precompiled headers

[pajlada]

Precompiled headers are currently causing us to have to recompile big parts of the project any time any header file changes, I think it's worth investigating what sort of value we're getting from precompiled headers (i.e. if it speeds up compilations), and if it's worth the tradeoff.

[fourtf]

I assume this is only relevant to the pajlada/x headers since they are the only ones in the precompiled header?

[pajlada]

I've noticed things being recompiled as soon as any header file is touched, hence the need for investigation

[fourtf]

not sure if this is related to precompiled headers but worth looking into it

[pajlada]

From my tests, precompiled headers shave off about 70 seconds off the build when using cmake and 4 threads

[pajlada]

Currently investigating ccache viability with precompiled headers disabled

[pajlada]

Cached builds (using ccache) work properly when precompiled headers are disabled.

I propose adding an option to the CMake call to disable precompiled headers (so they would be enabled by default to keep the current behaviour).

Thoughts on this @fourtf? The consequence would be adding #include directives where missing alongside the CMake option.
If I recall correctly, having #includes in files that have already been part of the precompiled header shouldn't slow down the compilation for precompiled headers builds

[pajlada]

To add, the original premise of this investigation was false, modifying header files did not cause more file recompiles with precompiled headers than without.