The assembly generated in case of clang-3.5 is
ldr r0, .LCPI0_0
ldr r1, .LCPI0_1
add r0, pc, r0
ldr r0, [r1, r0]
ldr r0, [r0]
.size indirect_call, .Ltmp0-indirect_call
Hmm, never mind. This looks correct. This is tail call optimization, which is possible because indirect_call() doesn't have anything in its frame.