[PATCH] D43632: [ELF] - Add testcase documenting flags assigned when empty synthetic section is removed.

[George Rimar via llvm-commits]

>LGTM

>It is not immediately clear what is the best behavior, but having a test
>for it is a good thing.
>
>BTW, what does ld.bfd outputs?
>
>Cheers,
>Rafael

bfd 2.30.51.20180214 marks .bar as writable. But that is because
of BYTE command which seems turns on "W" flag. 
In all cases below .bar is writable (.got.plt is absent):

1) SECTIONS { .bar : {  BYTE(0x11) }} and there is no any sections in .o (empty file).
  [ 0]                   NULL             0000000000000000  00000000
       0000000000000000  0000000000000000           0     0     0
  [ 1] .bar              PROGBITS         0000000000000000  00200000
       0000000000000001  0000000000000000  WA       0     0     1
  [ 2] .symtab           SYMTAB           0000000000000000  00200008
       0000000000000030  0000000000000018           3     2     8

2) SECTIONS { .foo : { *(.foo) } .bar : {  BYTE(0x11) }} and there is .foo in .o.
3) SECTIONS { .foo : { *(.foo) } .bar : { *(.got.plt) BYTE(0x11) }} and there is .foo in .o.

Output for 2, 3:
  [ 1] .foo              PROGBITS         0000000000000000  00200000
       0000000000000008  0000000000000000  AX       0     0     1
  [ 2] .bar              PROGBITS         0000000000000008  00200008
       0000000000000001  0000000000000000  WA       0     0     8

And gold just makes .foo section to be allocatable:
  [ 1] .foo              PROGBITS         0000000000000000  00001000
       0000000000000008  0000000000000000  AX       0     0     1
  [ 2] .text             PROGBITS         0000000000000008  00001008
       0000000000000000  0000000000000000  AX       0     0     1
  [ 3] .data             PROGBITS         0000000000000008  00001008
       0000000000000000  0000000000000000  WA       0     0     1
  [ 4] .bar              PROGBITS         0000000000000024  00001024
       0000000000000001  0000000000000000   A       0     0     0
  [ 5] .bss              NOBITS           0000000000000025  00001025
       0000000000000000  0000000000000000  WA       0     0     1

At least given the different behaviors we are more consistent with ourselves now, what is probably good.

George.