Actually that's not true, simply because the crash will occur on a code that has not been compiled by GCC, but by gpSP itself.
So there is no easy way to find the cause of a bug when dealing with a dynarec.
Small example, let's say that the recompiled code jumps to the address 0x0, which triggers a segmentation fault. Which part of the recompiler did create the faulty code? Maybe the code which generated the "jump" instruction is incorrect, or maybe the ones which generated other instructions involved in the calculation of the address. There's no easy way to find that, and GDB won't help.