[CLEANUP] Ifdefs inside functions

[QRiary]

So we either do

#ifdef CONDITION
void some_function(...) {
  ...
}
#endif

...
#ifdef CONDITION
  some_function(...);
#endif
...

or

void some_function(...) {
#ifdef CONDITION
  ...
#endif
}

...
some_function(...)
...

and I vote for the second approach, like all the functions which depend on defines should do that

[QRiary]

as for non-void functions - we could do like

int some_function() {
#ifdef ...
...
#else
return 0;  // or whatever useless value there
}

[Ikcelaks]

I think having the ifdefs inside the function works really well for optional void functions that will never alter the state of anything, like debug prints. However, I think that the call site needs ifdefs anyways for some optional feature code, so features should be completely contained in ifdefs, including the function declarations in the header files.

Here is an example. st_handle_backskpace should really return a bool that is true if it successfully performed the backspace and false if the buffer lacked enough keys to perform the undo. Maybe we do some special action in a different feature when the enhanced backspace fails. If enhance backspace is disabled then the call site needs to not do any of that, or else that second feature gets called everytime, because the enhance backspace code is always returning 0.

[QRiary]

Its still debatable - like all those returns are used by void functions at the end of a day, so if we just undef their behavior - all those junk returns will not matter, but void functions definitely good idea

[kamih]

My suggestion:
we wrap whole feature declaration and definitions in #ifdef blocks.
And for #ifndef, we can define stubs for the top level void return functions that simply resolve to nothing, so the top level code doesn't need dirty ifdefs everywhere and can still just call the stub.
Top level code will still have to use #ifdefs around some code, like SEQUENCE_TRANSFORM_ENHANCED_BACKSPACE in process_sequence_transform. No way around stuff like that.