[x264-devel] Re: bug in x264_cqm_init ?

[David Wolstencroft]

There are ways to affect alignment and padding between struct  
members. (as seen in the #pragma below)

Citing a reference from wikipedia (sorry, it was easy to find)
http://en.wikipedia.org/wiki/Data_structure_alignment


While there is no standard way of defining the alignment of structure  
members, some compilers use #pragma directives to specify packing  
inside source files. Here is an example:

#pragma pack(push)  /* push current alignment to stack */
#pragma pack(1)     /* set alignment to 1 byte boundary */

struct MyPackedData
{
     char Data1;
     long Data2;
     char Data3;
};

#pragma pack(pop)   /* restore original alignment from stack */
This structure would have a compiled size of 6 bytes. The above  
directives are available in compilers from Microsoft, Borland, GNU  
and many others.



- David


On May 3, 2007, at 5:22 PM, Måns Rullgård wrote:

> Loren Merritt <lorenm at u.washington.edu> writes:
>
>> On Thu, 3 May 2007, Fejes Dávid wrote:
>>
>>> Memory is never allocated for h->dequant8_mf, h->quant8_mf and
>>> h->unquant8_mf.
>>>
>>> x264_cqm_init (in common/set.c) will upload this matrices with data,
>>> but it doesn't allocates any memory before that for these  
>>> pointers. In
>>> my build it causes general protection fault (or simply it mess up
>>> another variables). If I paste the corresponding x264_memalloc
>>> function calls onto the beginning of the x264_cqm_init, it goes  
>>> well.
>>>
>>> If I'm wrong, please let me know it. I've downloaded x264 at
>>> 22.07.2007 from the trunk, excuse me if you have corrected this.
>>
>> Does this patch give an assert failure?
>> If so, I'll have to rethink my assumptions about compilers. Though I
>> don't see how to fix it without adding code complexity.
>> If not, I can't see how else it would break.
>>
>> +    assert((void*)&h->  quant4_mf[4] == (void*)&h->  quant8_mf[0]);
>> +    assert((void*)&h->dequant4_mf[4] == (void*)&h->dequant8_mf[0]);
>> +    assert((void*)&h->unquant4_mf[4] == (void*)&h->unquant8_mf[0]);
>> +    assert((void*)&h->quant4_bias[4] == (void*)&h->quant8_bias[0]);
>
> There's no guarantee that those conditions are true.  The compiler is
> allowed to add padding between any struct members.  If you need these
> arrays to be contiguous you could declare them as a single array and
> initialise two pointers into it.  That said, I'd be surprised if any
> compiler actually added any padding here.
>
> -- 
> Måns Rullgård
> mans at mansr.com
>
> --
> This is the x264-devel mailing-list
> To unsubscribe, go to: http://developers.videolan.org/lists.html
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.videolan.org/pipermail/x264-devel/attachments/20070506/fd1b9f8b/attachment.htm