HotSpot 16 and OpenJDK6

[Andrew John Hughes]

On 23 February 2010 20:10, Joseph D. Darcy <Joe.Darcy at sun.com> wrote:
> Hello.
>
> Andrew John Hughes wrote:
>>
>> Here's the merge:
>>
>> http://cr.openjdk.java.net/~andrew/jdk6-hs16-merge/webrev.01/
>>
>> which now builds, thanks to this additional changeset from Daniel
>> Daugherty in baseline:
>> http://hg.openjdk.java.net/hsx/hsx16/baseline/rev/c9740f5ed5b4
>>
>
> Good.
>
>> It also includes an additional fix, also listed separately at:
>>
>> http://cr.openjdk.java.net/~andrew/jdk6-hs16-merge/webrev.02/
>>
>> which reverts an erroneous change from SIZE_FORMAT to %d in
>>
>> changeset:   734:dbbe28fc66b5
>> user:        twisti
>> date:        Fri Feb 27 03:35:40 2009 -0800
>> summary:     6778669: Patch from Red Hat -- fixes compilation errors
>>
>> fixing the build on x86_64.  We'll need a bug ID for this reversion.
>>
>
> The webrev generally looks good, but I have a few questions and comments
> before this goes back.
>
> Do you know why webrev shows so many files with zero changes?  I assume this
> is an artifact of the merge.
>

No, you're right that does seem strange.  All I did was a pull from
hs16 and then a merge.  Perhaps this is some webrev oddity?

If you look at http://mail.openjdk.java.net/pipermail/jdk6-dev/2010-February/001230.html
I included a diff of the merged OpenJDK6 against the hs16 master
(which includes Daniel's patch that is currently only in baseline).
It actually makes the other questions below a bit clearer.

> In hotspot_version, what are the semantics of JDK_PREVIOUS_VERSION?
>

I'm not sure.  It doesn't seem to do much.  Its only use is to set a
default BOOTDIR AFAICS:

# Find JDK used for javac compiles
BOOTDIR=$(SLASH_JAVA)/re/j2se/$(PREVIOUS_JDK_VERSION)/latest/binaries/$(PLATFORM)

(from hotspot/make/defs.make)

We actually differ from hs16 by still using 1.5.0 rather than 1.6.0.
1.5.0 implicitly seems to make more sense to me, because 1.6.0 is the
version being built.  But in practice, it doesn't seem to matter
unless you're relying on the contents of a /java tree.

> In terms of fixing the %d vs SIZE_FORMAT in vmError.cpp, I've filed 6929005
> "Fix format specifier in vmError.cpp."  However, I notice that the JDK 7
> master still has %d in this location.  I'll approve going back from %d ->
> SIZE_FORMAT in OpenJDK 6 conditional on the review and approval of a HotSpot
> engineer.
>

Interesting; I didn't check OpenJDK7 and indeed you're right.  Looking
again at the diff, it seems that we don't apply part of the same
changeset:

--- ../hs16/src/share/vm/utilities/vmError.hpp	2010-01-13
14:35:26.651211783 +0000
+++ hotspot/src/share/vm/utilities/vmError.hpp	2010-02-17
11:07:40.899854566 +0000
@@ -50,7 +50,7 @@

   // additional info for VM internal errors
   const char * _filename;
-  int          _lineno;
+  size_t       _lineno;

   // used by fatal error handler
   int          _current_step;

keeping the size_t rather than using int.  So I'll change the fix to
be making vmError.hpp match hs16 and OpenJDK7, rather than changing
vmError.cpp.

The remaining diffs are typos and license fixups from the previous
merge which need to go back to the main HotSpot tree.

> Thanks,
>
> -Joe
>



-- 
Andrew :-)

Free Java Software Engineer
Red Hat, Inc. (http://www.redhat.com)

Support Free Java!
Contribute to GNU Classpath and the OpenJDK
http://www.gnu.org/software/classpath
http://openjdk.java.net

PGP Key: 94EFD9D8 (http://subkeys.pgp.net)
Fingerprint: F8EF F1EA 401E 2E60 15FA  7927 142C 2591 94EF D9D8