Hi Michael,
Partial review of jdk changes. Sorry but it is too large to go through
everything in detail.
General meta comment: When this stuff went into 7u my biggest complaint
was that it introduced a mass of new platform-specific stuff in places
that weren't otherwise platform specific. That still seems to be the
case. I was really hoping/expecting that these issues could be resolved
before it was merged with JDK8 otherwise they will never be fixed. Eg
jdk/make/common/Rules.gmk, jdk/make/common/shared/Defs-java.gmk. It
seems to me that many of the specializations could be placed into the
platform-specific gmk file(s) instead of polluting the shared ones. In
some case though (eg Makefiles where we have to change SUBDIR) it
indicates we need a better way to specialize things.
Ditto for the Java code. I really hate to see platform specific changes
added to Java code that previously was platform independent.
One other general nit (because it made it much harder to see real
changes!): there is no need to add comments to every "else" and "endif"
in the build files. When they are long sections sure, but not for one or
two lines. That makes it harder to read things in my opinion. (Ref:
jdk/make/common/Program.gmk in particular) Plus this was done
inconsistently anyway (ref: jdk/make/common/Release.gmk)
A few specific comments:
make/common/Release.gmk:
This:
975 # The launcher source files we need for src.zip
976 FILES_launcher = $(wildcard $(SHARE_SRC)/bin/*) \
977 $(wildcard $(PLATFORM_SRC)/bin/java_md*)
is picking up a couple of extra files compared to the original file
list: defines.h, emessages.h, main.c
---
make/common/Rules.gmk
Can we fix the ToDos in here? Either do them or delete them.
---
/jdk/make/com/sun/nio/Makefile
35 ifneq ($(PLATFORM), macosx)
36 include $(BUILDDIR)/common/Subdirs.gmk
37 SUBDIRS = sctp
38 endif
39
40
41 all build clean clobber::
42 $(SUBDIRS-loop)
Don't we need the include of Subdirs.gmk to get the definition of
SUBDIRS-loop?
Is sctp an optional JDK component? Or is it provided some other way on OSX?
---
java/lang/Thread.java:
private native void resume0();
private native void interrupt0();
+ private native void setNativeName(String name);
I don't see the change in setName to call setNativeName ???
---
David
----------------
On 28/02/2012 7:07 AM, Michael McMahon wrote:
This is an updated webrev for the Mac OS X changes after comments so far:
http://cr.openjdk.java.net/~michaelm/7113349/jdk8/webrev.3/
The main changes from the original webrev are:
- jprt configuration for macosx
- problem list updates
- removal of dud source files brought in from 7u-dev by mistake
- langtools test updates
- synchronized with TL now as opposed to master
- removal of fix for 7087956, which wasn't part of mac os work
Comments welcome.
Thanks,
Michael.
On 22/02/12 21:50, Michael McMahon wrote:
This is the webrev for the Mac OS changes for JDK8. It is pretty much
a straight merge of the current jdk7u-dev changes. It builds and
passes most
JCK and regression tests with only a small number of failures.
http://cr.openjdk.java.net/~michaelm/7113349/jdk8/webrev.1/
There are three separate webrevs at the link above for the three separate
repositories affected (the top-level, corba and jdk). Obviously most
of the changes
are in the jdk webrev. Comments welcome!
Thanks,
Michael.
