[llvm] r189315 - Move everything depending on Object/MachOFormat.h over to Support/MachO.h.

[Nick Kledzik]

On Aug 28, 2013, at 4:15 PM, David Fang <fang at csl.cornell.edu> wrote:
> Hi all,
> 	Did you have more details about the big-endian failures?  Was it by chance hello-reloc.s, which tests powerpc-darwin8?  I contributed the PPCMachObjectWriter, which has some notes about funny endianness in makeRelocationInfo and makeScatteredRelocationInfo.
> If you have further revisions of this patch that you'd like tested on PowerPC-darwin, I'll be happy to help test.

Fang,

Chip posted a revised patch for review earlier today in this thread.  If you could try that out on a PowerPC machine, that would be most helpful.  Thanks.

-Nick



>> On Aug 27, 2013, at 4:59 PM, Nick Kledzik wrote:
>> 
>>> On Aug 26, 2013, at 10:41 PM, Charles Davis <cdavis5x at gmail.com> wrote:
>>>> On Aug 26, 2013, at 11:32 PM, Charles Davis wrote:
>>>>> On Aug 26, 2013, at 11:19 PM, David Blaikie wrote:
>>>>> 
>>>>>> On Mon, Aug 26, 2013 at 10:00 PM, Charles Davis <cdavis5x at gmail.com> wrote:
>>>>>>> Author: cdavis
>>>>>>> Date: Tue Aug 27 00:00:43 2013
>>>>>>> New Revision: 189315
>>>>>>> 
>>>>>>> URL: http://llvm.org/viewvc/llvm-project?rev=189315&view=rev
>>>>>>> Log:
>>>>>>> Move everything depending on Object/MachOFormat.h over to Support/MachO.h.
>>>>>> 
>>>>>> There seem to be a number of unrelated changes in this patch -
>>>>>> especially for such a large change, please commit the mechanical
>>>>>> portion separately from any other cleanup/changes.
>>>>> Sorry. Would you like me to revert this particular change?
>>>> I went ahead and reverted it (and r189319) in r189321. It looks like it also broke the tests on some platforms (specifically System z and PowerPC--endianness issue? I don't have a big endian machine to test with, so?).
>>> 
>>> Chip,
>>> 
>>> After looking at the failures, the problematic area is with defining relocation_info using bit fields.  Your patch uses:
>>> 
>>> +    struct scattered_relocation_info {
>>> +#if defined(BYTE_ORDER) && defined(BIG_ENDIAN) && (BYTE_ORDER == BIG_ENDIAN)
>>> +      uint32_t r_scattered:1,
>>> +               r_pcrel:1,
>>> +               r_length:2,
>>> +               r_type:4,
>>> +               r_address:24;
>>> +#else
>>> +      uint32_t r_address:24,
>>> +               r_type:4,
>>> +               r_length:2,
>>> +               r_pcrel:1,
>>> +               r_scattered:1;
>>> +#endif
>>> +      int32_t r_value;
>>> +    };
>>> 
>>> 
>>> But whether all three of those preprocessor symbols (e.g. BIG_ENDIAN) are set up is hazy.  My guess is one of the is not set up on the failing big endian builders and causing the relocations to be messed up.
>> Really? I thought it was because the other struct (relocation_info) *doesn't* use them. So now the order of the bitfields depends on the host endianness--which, when we're making an object file, means that the correct order also depends on the *target* endianness. Unfortunately, we're stuck with this, because the definition was cribbed from <mach-o/reloc.h>, and it has the same problem. (The definition of scattered_relocation_info was also cribbed from <mach-o/reloc.h>; that's why I used the endianness macros there. In fact, I believe I originally used Apple's __BIG_ENDIAN__ macro for this, but I changed it because I figured this wouldn't get defined everywhere it needed to be.) As far as I could tell, the relocations that came out wrong on those machines were plain relocations, not scattered relocations. In particular, several x86_64 relocations came out wrong, and x86_64 doesn't use scattered relocations at all.
>>> I don?t see many of the endian conditionals used anywhere in LLVM.  I?d suggest keeping the MC stuff using the word0 and word1 style of access and avoid using bitfields.
>> I figured you'd say that, and it sounds like a good idea regardless: it worked before, after all ;). (It just took me some time to audit my change and make sure there was nothing beyond mechanical renaming so nothing else like this happens.)
>> 
>> New patch attached. In addition to continuing to use MachO::any_relocation_info, it also incorporates David's warning fix from r189319. I don't want to cause any more churn than I already have, so I won't commit until someone goes over it and gives me their LG.
>> 
>> Chip
>> 
> 
> -- 
> David Fang
> http://www.csl.cornell.edu/~fang/