RFR: JDK-8160285: Bootcycle builds are broken on jdk9/hs for windows i586

[David Holmes]

Hi Erik,

On 28/06/2016 1:57 AM, Erik Joelsson wrote:
> Hello,
>
> When doing a bootcycle build for a 32bit target, but using a 64bit
> bootjdk, the JVM memory arguments found suitable for the 64bit bootjdk
> are still used when running the bootcycle part of the build. When using
> a 32bit bootjdk, we limit the mx flag to 1100M while a 64bit bootjdk
> gets 1600M. So far, the 32bit JDK 9 Hotspot has been able to cope with
> -Xmx1600M, but it has now started to fail on Windows. While that may be
> a bug to investigate on its own, the build has not reason for requiring
> such big heaps from a 32bit JVM.

This may coincide with the switch from the client VM to server VM on 
32-bit Windows.

> This patch adds more special treatment in the bootcycle-spec for when
> building a 32bit target that makes sure the 32bit JVM gets at most
> mx1100M. I would like to push this to jdk9/dev ASAP to avoid the
> bootcycle build breaking when jdk9/hs next gets integrated.
>
> Bug: https://bugs.openjdk.java.net/browse/JDK-8160285
> Webrev: http://cr.openjdk.java.net/~erikj/8160285/webrev.top.01/

Fix looks fine to me.

Thanks,
David


> /Erik