[OpenJDK 2D-Dev] [8] request for review: 7150594: VM chash in JCK api/java_awt/Image/ConvolveOp/ tests for 64 bit jdk8 on linux.

[Andrew Brygin]

Hello Phil,

  thanks for the review.

  Regarding posix_memalign(), originally I was going to use it instead of
  memalign() but found that it is not available on Solaris platform.
  I did not like an idea to introduce yet another platfrom-specific code
  piece, and and decided to leave memaling() routine.

  I do not think that there are another issues caused by the fix for
  CR 7113017, and the fix suggested here probably can be used
  for 7u8 (where we have to resolve the JCK failure as well).

Thanks,
Andrew

On 02.08.2012 20:40, Phil Race wrote:
> I am OK with this. The submitter of the fix for 7113017 apparently
> did not pay attention in the updating to use POSIX functions that
> posix supplies its own function called posix_memalign so does not
> define memalign. I don't know off hand what library that would come
> from on all platforms and I do not think it worth cycles to figure 
> this out.
> I hope this is the only such issue but don't know if for a fact.
> Some of these portability requests seem like they are more trouble 
> than they are
> worth and using malloc.h works just fine for me !
>
> If there are any more issues I suggest we just revert 7113017.
>
> -phil.
>
>
> On 8/2/2012 7:36 AM, Andrew Brygin wrote:
>> Hello,
>>
>>  could you please review a fix for 7150594?
>>
>>  This problem is triggered by the fix for CR 7113017. In particular,
>>  this fix replaces the malloc.h with stddef.h in mlib_types.h. This
>>  change leads to compiling mlib_sys.c without forward declaration for
>>  memalign() routine, and cause following warnings:
>>
>> mlib_sys.c:96: warning: implicit declaration of function 'memalign'
>> mlib_sys.c:96: warning: cast to pointer from integer of different size
>>
>>  This cause the problem on systems where size of integer is less than
>>  size of pointers: the pointer value is clamped, and usage of the
>>  clamped pointer causes the observed crash.
>>
>>  Proposed solution is to include malloc.h header directly to mlib_sys.c.
>>
>>  Please take a look.
>>
>> Bug: http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=7150594
>> Webrev: http://cr.openjdk.java.net/~bae/7150594/8/webrev.00/
>>
>> Thanks,
>> Andrew
>