webrevs for macosx changes to jdk7u-osx

Alan Bateman Alan.Bateman at oracle.com
Tue Nov 22 03:48:52 PST 2011


On 21/11/2011 15:46, Michael McMahon wrote:
> Hi,
>
> The following webrevs are an initial set of changes
> taken from the macosx-port forest to be applied
> to the jd7u-osx forest at: http://hg.openjdk.java.net/jdk7u/jdk7u-osx/
> This will eventually be merged into the main jdk7u-dev forest.
>
> Hotspot has already integrated into this forest. This changeset includes
> the changes (in the jdk and corba) repositories that are needed to 
> build and run
> openjdk on MacOSX. The AWT/client code is however taken from the BSD 
> forest. So, the
> native Apple AWT/client code will be integrated later. By default, the 
> VM will run in headless
> mode. To run with the X windows AWT, set the environment variable 
> AWT_TOOLKIT to XToolkit.
>
> This is a work in progress and much remains to be done. But, it does 
> build and run.
> All comments welcome.
>
Thanks for getting this going. I haven't done a detailed review but I 
have skimmed through the webrevs and have a couple of comments, some of 
which are just to note areas that need closer examination.

I had expected to see lots of updates to the tests but they aren't in 
the webrev.

It would be good to get Kelly or someone from build-dev to do a pass 
over the make file changes.

I think the launcher changes should be re-examined and probably 
refactored into platform specific code for the launcher.

System.java/ClassLoader.java - I think we need to re-examine these 
changes aswell to see whether they are needed and where the best place 
is to support it.

I agree with Phil's comment that we need to be consistent on the value 
of os.name.

In check_code.c then _ck_ntohl seems a bit odd. I assume it has been 
renamed because of a conflict.

src/share/native/java/lang/System.c - this needs to be re-examined/moved 
to platform specific code.

src/share/transport/socket/socketTransport.c - this seems to be fixing a 
bug that isn't Mac specific and should be fixed upstream.

src/solaris/native/com/sun/management/UnixOperatingSystem_md.c - this is 
missing implementations of several functions and I assume is on the list 
to examine.

src/solaris/native/java/io/io_util_md.c - the addition of code to 
convert to Unicode NFD is a reminder that this is not the only place 
where this need to do done (looks like it's missing from the new code 
added in jdk7).

src/solaris/native/java/net/Inet4AddressImpl.c - is this necessary? 
Would it be easier to get 7112670 back-ported into 7u?

src/macosx/classes/sun/nio/ch/DefaultSelectorProvider.java - the kqueue 
Selector fails quite a few tests so I would suggest default to 
sun.nio.ch.PollSelectorProvider until those issues are fixed. Also no 
need for the java.nio.preferSelect property as there is a standard 
property (java.nio.channels.spi.SelectorProvider) defined in the 
specification to selector the provider.

src/solaris/classes/sun/nio/fs/BsdFileStore.java - looks like the 
commented out code can be removed (not needed).

src/solaris/classes/sun/nio/fs/BsdFileSystem.java - looks like 
supportedFileAttributeViews() and copyNonPosixAttributes() can be removed.

src/solaris/classes/sun/nio/fs/BsdFileSystemProvider.java - looks like 
the getFileAttributeView and readAttributes can be removed.

I hope to do a second and more detailed pass over the changes once I get 
time.

-Alan.






























More information about the macosx-port-dev mailing list