[OpenJDK 2D-Dev]  request for review: 7150594: VM chash in JCK api/java_awt/Image/ConvolveOp/ tests for 64 bit jdk8 on linux.
philip.race at oracle.com
Thu Aug 2 16:40:17 UTC 2012
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
worth and using malloc.h works just fine for me !
If there are any more issues I suggest we just revert 7113017.
On 8/2/2012 7:36 AM, Andrew Brygin wrote:
> 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/
More information about the 2d-dev