[PATCH FOR REVIEW]: Make source/target options explicit for CORBA bootstrap tools

Tue Sep 1 22:14:54 UTC 2009

2009/8/27 Andrew John Hughes <gnu_andrew at member.fsf.org>:
> 2009/8/27 Jonathan Gibbons <Jonathan.Gibbons at sun.com>:
>>
>> On Aug 27, 2009, at 3:53 AM, Andrew John Hughes wrote:
>>
>>> 2009/8/20 Andrew John Hughes <gnu_andrew at member.fsf.org>:
>>>>
>>>> 2009/8/20 Jonathan Gibbons <Jonathan.Gibbons at sun.com>:
>>>>>
>>>>> Andrew John Hughes wrote:
>>>>>
>>>>> 2009/8/18 Andrew John Hughes <gnu_andrew at member.fsf.org>:
>>>>>
>>>>>
>>>>> 2009/8/18 Jonathan Gibbons <Jonathan.Gibbons at sun.com>:
>>>>>
>>>>>
>>>>> Andrew,
>>>>>
>>>>> If this is a patch for jdk7, it does not appear to be a patch to a
>>>>> recent
>>>>> copy
>>>>> of 7.
>>>>>
>>>>>
>>>>> It's against b69 which is the latest release (from Friday).  The
>>>>> patches are against the IcedTea forest so builds can be tested with
>>>>> IcedTea as well.
>>>>>
>>>>> Specifically, you do not seem to have the recent changeset  to set
>>>>>
>>>>>
>>>>> the source/target used to compile JDK to 7. [1]
>>>>>
>>>>>
>>>>>
>>>>> Er... yes I do:
>>>>>
>>>>> # Add the source level
>>>>> LANGUAGE_VERSION = -source 7
>>>>> JAVACFLAGS  += $(LANGUAGE_VERSION)
>>>>>
>>>>> # Add the class version we want
>>>>> TARGET_CLASS_VERSION = 7
>>>>> CLASS_VERSION = -target $(TARGET_CLASS_VERSION)
>>>>> JAVACFLAGS  += $(CLASS_VERSION)
>>>>> JAVACFLAGS  += -encoding ascii
>>>>> JAVACFLAGS  += -classpath $(BOOTDIR)/lib/tools.jar
>>>>> JAVACFLAGS  += $(OTHER_JAVACFLAGS)
>>>>>
>>>>> but these only cover the rt classes and not the bootstrap classes.
>>>>>
>>>>>
>>>>>
>>>>> While your patch does not directly conflict with any edits in that
>>>>> patch,
>>>>> and
>>>>> while the effect of your patch looks OK, in that patch I was extending
>>>>> the
>>>>> precedent of TARGET_CLASS_VERSION to have an explicit macro for
>>>>> (just) the version number, so that it is easy to change the value of
>>>>> (just)
>>>>> the version number from the command line.
>>>>>
>>>>> With that in mind, I would suggest something like the following for your
>>>>> patch:
>>>>>
>>>>> BOOT_SOURCE_LANGUAGE_VERSION = 6
>>>>> BOOT_TARGET_CLASS_VERSION = 6
>>>>> BOOT_JAVACFLAGS  += -encoding ascii -source
>>>>> $(BOOT_SOURCE_LANGUAGE_VERSION)
>>>>> -target $(BOOT_TARGET_CLASS_VERSION)
>>>>>
>>>>>
>>>>>
>>>>> I didn't copy this for the 6 changes because I didn't immediately see
>>>>> the point of using variables just for this single use.  I forgot that
>>>>> it is possible to override these from the command line, so I've update
>>>>> the patch:
>>>>>
>>>>> http://cr.openjdk.java.net/~andrew/ecj/02/webrev.02/
>>>>>
>>>>>
>>>>>
>>>>> -- Jon
>>>>>
>>>>> [1]
>>>>>
>>>>> http://mail.openjdk.java.net/pipermail/compiler-dev/2009-July/001286.html
>>>>>
>>>>>
>>>>> On Aug 18, 2009, at 5:24 AM, Andrew John Hughes wrote:
>>>>>
>>>>>
>>>>>
>>>>> Currently the javac calls for building the bootstrap tools (not the
>>>>> classes for the final JDK, which correctly now use source and target
>>>>> 7) don't set an explicit source and target version.
>>>>>
>>>>> The webrev:
>>>>>
>>>>> http://cr.openjdk.java.net/~andrew/ecj/02/webrev.01/
>>>>>
>>>>> sets these to 6 explicitly, as happens in the Ant builds performed by
>>>>> langtools/jaxp/jaxws.  This is noticeable especially when using ecj as
>>>>> the bootstrap javac as it defaults to a version < 1.5, and the build
>>>>> fails.
>>>>>
>>>>> Ok to push?
>>>>>
>>>>> Thanks,
>>>>> --
>>>>> 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
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> 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
>>>>>
>>>>>
>>>>>
>>>>> Is this version now ok?  If so, I'll push it to the build gate using
>>>>> the bug ID Kelly allocated for the same fix in JDK.
>>>>>
>>>>>
>>>>> Andrew,
>>>>>
>>>>> I approve your webrev
>>>>> http://cr.openjdk.java.net/~andrew/ecj/02/webrev.02/
>>>>
>>>> Thanks.  Pushed this last night:
>>>> http://hg.openjdk.java.net/jdk7/build/corba/rev/8001ba2bf10d
>>>>
>>>>> My earlier confusion was caused by the fact that the corba Makefile is
>>>>> not
>>>>> consistent with the jdk Makefile with respect to the use of
>>>>> SOURCE_LANGUAGE_VERSION.
>>>>> It would be good to (separately) fix that inconsistency, but that does
>>>>> not
>>>>> affect the validity of what you propose here.
>>>>>
>>>>
>>>> Yes, I see what you mean now.  Here's another webrev:
>>>>
>>>> http://cr.openjdk.java.net/~andrew/consistency/01/webrev.01/
>>>>
>>>> which should make the two fairly consistent.  It brings in a lot of
>>>> changes to the version in JDK that seem to have been missed from
>>>> CORBA, namely:
>>>>
>>>>  * Turning off option outputs on fastdebug builds
>>>>  * Supporting USE_HOTSPOT_INTERPRETER_MODE
>>>>  * Supporting JAVAC_MAX_WARNINGS and JAVAC_WARNINGS_FATAL
>>>>  * The SOURCE_LANGUAGE_VERSION sync above
>>>>  * Include JAR_JFLAGS in BOOT_JAR_JFLAGS
>>>>
>>>> The following differences remain, which didn't seem appropriate to
>>>> include:
>>>>
>>>> -JAVACFLAGS  += -classpath $(BOOTDIR)/lib/tools.jar
>>>> +JAVACFLAGS  += "-Xbootclasspath:$(CLASSBINDIR)"
>>>>  JAVACFLAGS  += $(OTHER_JAVACFLAGS)
>>>>
>>>>  # Needed for javah
>>>> -JAVAHFLAGS += -classpath $(CLASSBINDIR)
>>>> +JAVAHFLAGS += -bootclasspath $(CLASSBINDIR)
>>>>
>>>> I wasn't sure of the pros/cons of these changes but can easy add them if
>>>> needed.
>>>>
>>>>> -- Jon
>>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> 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
>>>>
>>>
>>> Does this change look ok? Can I push it?
>>>
>>> Thanks,
>>> --
>>> 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
>>
>> Andrew,
>>
>> The changes regarding flags for javac look OK.  I can't speak to the Hotspot
>> options.
>>
>> -- Jon
>>
>>
>
> Ok, thanks Jonathan.  I'll forward a copy to hotspot-dev for approval
> on supporting USE_HOTSPOT_INTERPRETER_MODE.
> --
> 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
>

The HotSpot developers don't see any issue with USE_HOTSPOT_INTERPRETER_MODE:

http://mail.openjdk.java.net/pipermail/hotspot-dev/2009-September/002032.html

and it is already in the JDK makefiles.

If someone could allocate me a bug ID for this, I'll push it.

Thanks,
-- 
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