RFR: 8048895: Back out JDK-8027915

[Stefan Karlsson]

On 2014-07-01 16:58, Erik Helin wrote:
> Hi all,
>
> the change JDK-8027915, "TestParallelHeapSizeFlags fails with unexpected
> heap size", did not work as anticipated because of the interaction with
> os::commit_memory on Solaris.
>
> os::commit_memory takes a size_t `alignment_hint` as parameter. This
> parameter is used differently on different operating systems: it is
> ignored on all operating system except for Solaris. For Solaris, the
> hint is used for selecting the large page size. If the alignment_hint is
> smaller than the largest page size available, the hint is assumed to be
> the *exact* page size.
>
> This problem had previously been hidden because due to various
> alignments of heap sizes and generation sizes always made sure that we
> ended up with an alignment_hint of 4 MB by default, which also happens to
> be a valid page size.
>
> The bug can be shown to exist prior to JDK-8027915, for example by
> running:
> java -Xms32m -Xmx128m -XX:LargePageSizeInBytes=256m -version
>
> This patch is the anti-delta of JDK-8027915 and it applied cleanly.
>
> Webrev:
> http://cr.openjdk.java.net/~ehelin/8048895/webrev.00/

combinediff 8027915.patch 8048895.patch returns nothing, so this backout 
looks good.

StefanK

>
> Thanks,
> Erik
>