GCC is sort of infamously pedantic about -Os. This is my favorite example:

  {0}[calvinow ~] cat test.c
  int foo(int a)
  {
        return a / 4;
  }
  {0}[calvinow ~] gcc -c -O2 test.c -o opt.o
  {0}[calvinow ~] objdump -d opt.o 

  opt.o:     file format elf64-x86-64


  Disassembly of section .text:

  0000000000000000 <foo>:
   0:   85 ff                   test   %edi,%edi
   2:   8d 47 03                lea    0x3(%rdi),%eax
   5:   0f 49 c7                cmovns %edi,%eax
   8:   c1 f8 02                sar    $0x2,%eax
   b:   c3                      ret
  {0}[calvinow ~] gcc -c -Os test.c -o size.o
  {0}[calvinow ~] objdump -d size.o 

  size.o:     file format elf64-x86-64


  Disassembly of section .text:

  0000000000000000 <foo>:
   0:   89 f8                   mov    %edi,%eax
   2:   b9 04 00 00 00          mov    $0x4,%ecx
   7:   99                      cltd
   8:   f7 f9                   idiv   %ecx
   a:   c3                      ret

...because the division saves a single byte on x86! It's hard to believe anybody would ever want that trade.