[llvm] r293224 - [obj2yaml] Produce correct output for invalid relocations.

[Rafael Avila de Espindola via llvm-commits]

Oh, I don't mean or expect the files to be bit by bit identical. What I
was referring to is that a relocation without a symbol turns into a
relocation with a symbol with an empty name.

Cheers,
Rafael

Davide Italiano <davide at freebsd.org> writes:

> On Fri, Jan 27, 2017 at 10:27 AM, Rafael Avila de Espindola
> <rafael.espindola at gmail.com> wrote:
>> Passing the result object back via yam2obj will produce a different
>> result, no?
>>
>
> A bit. I may be wrong, but I think you can obtain an isomorphism even
> without this change (as sections are laid out differently).
>
> Both files contain the 'correct' reloc, which was the purpose of the test:
>
> $ objdump -r array.o
> RELOCATION RECORDS FOR [.text]:
> OFFSET           TYPE              VALUE
> 0000000000000000 R_X86_64_NONE     *ABS*
>
> $ objdump -r reverse.o
> RELOCATION RECORDS FOR [.text]:
> OFFSET           TYPE              VALUE
> 0000000000000000 R_X86_64_NONE     *ABS*
>
> We now create `.shstrtab` in the inverted relocatable (which doesn't
> appear in the original object).
>
> $ readelf -S array.o
> Section Headers:
>   [Nr] Name              Type             Address           Offset
>        Size              EntSize          Flags  Link  Info  Align
>   [ 0]                   NULL             0000000000000000  00000000
>        0000000000000000  0000000000000000           0     0     0
>   [ 1] .text             PROGBITS         0000000000000000  000001c0
>        0000000000000005  0000000000000000  AX       0     0     4
>   [ 2] .rela.text        RELA             0000000000000000  000001c8
>        0000000000000018  0000000000000018           3     1     8
>   [ 3] .symtab           SYMTAB           0000000000000000  000001e0
>        0000000000000060  0000000000000018           4     2     8
>   [ 4] .strtab           STRTAB           0000000000000000  00000240
>        0000000000000044  0000000000000000           0     0     1
>   [ 5] .shstrtab         STRTAB           0000000000000000  00000284
>        0000000000000026  0000000000000000           0     0     1
>
> $ readelf -S reverse.o
> Section Headers:
>   [Nr] Name              Type             Address           Offset
>        Size              EntSize          Flags  Link  Info  Align
>   [ 0]                   NULL             0000000000000000  00000000
>        0000000000000000  0000000000000000           0     0     0
>   [ 1] .strtab           STRTAB           0000000000000000  00000040
>        0000000000000070  0000000000000000           0     0     1
>   [ 2] .text             PROGBITS         0000000000000000  000000b0
>        0000000000000005  0000000000000000  AX       0     0     4
>   [ 3] .rela.text        RELA             0000000000000000  000000b8
>        0000000000000018  0000000000000018           4     2     8
>   [ 4] .symtab           SYMTAB           0000000000000000  000000d0
>        0000000000000060  0000000000000018           1     2     8
>
> Even without the `.shstrtabe` section emitted, the order of the
> section header is different in the two executables, so I don't think
> you were obtaining the original object back anyway.
>
> --
> Davide