I'm basically compiling modules(object files) with several weak function definitions(need to be relocated). I have a running program that brings and consumes those modules in memory, similarly to dlfnc. While inspecting the generated code in the modules with objdump tool, I found out that compiler doesn't generate far calls for functions that need to be relocated. In order to do so, I would have to increase a byte for every unrelocated function. As you can imagine, this patch strategy will potentially break every single PC based code. Curiously, when inspecting the whole build process, I found out that linker patches the near call to a near jump instruction, then through a table (I suppose) it performs another jump, a far one. In other words, two jumps. 1) why performing two jumps over a far call? I also read in docs that there is an implementation -mlong-calls available for ARM. If I understood it correctly, that would be exactly what I'm looking for. 2) Is there a way to induce gcc for doing something similar to me? I tried unsuccessfully to do it through inline asm. I'm about to implement the two jumps procedure, but I would love to know if there is another elegant way. Much appreciated
