This looks good to me. Thanks for hanging in there!
/Erik
On 2020-03-17 15:07, Roger Riggs wrote:
Hi Magnus, Erik,
Thanks for the pointers, I'm not familiar with those early build
intricacies.
Updated:
http://cr.openjdk.java.net/~rriggs/webrev-stubs-classes-8241073-2/
More cleanup:
- cleanup of ZipSource.gmk and autoconf/spec.gmk.in and Docs.gmk
- The mystery of ActivationGroup_Stub is resolved. The class needed
to be in the spec/javadoc
but it also needed to be generated by RMIC, the version in
src/java.rmi/share/doc
contained the javadoc comments. I merged the javadoc into the
generated stub .java class
and added it to the repo.
- The NetBeans Jmx build script had targets to build the stubs, they
have been removed.
Thanks, Roger
On 3/17/20 10:06 AM, Magnus Ihse Bursie wrote:
On 2020-03-17 14:17, Erik Joelsson wrote:
Hello,
That looks better, but there are still some more things to remove.
This whole block:
################################################################################
# Targets for running rmic.
$(eval $(call DeclareRecipesForPhase, RMIC, \
TARGET_SUFFIX := rmic, \
FILE_PREFIX := Rmic, \
MAKE_SUBDIR := rmic, \
CHECK_MODULES := $(ALL_MODULES)))
ALL_TARGETS += $(RMIC_TARGETS)
And all references to $(RMIC_TARGETS) and $(RMIC_MODULES). In most
cases the whole lines (like the $(foreach) calls that iterate over
them) can just be removed.
This also means killing the entire "rmic" phase of the build. So
please also remove the rmic target from Main.gmk line 1015, and
ALL_TARGETs line 1133. And the reference to "rmic" from the phases in
the help in Global.gmk.
In ZipSource.gmk, there's special handling to include rmic source
code, which is no longer needed.
In Docs.gmk, you can remove $(SUPPORT_OUTPUTDIR)/rmic/* from
MODULES_SOURCE_PATH.
(That actually made me notice another weird rmi thingy. How does
these new stubs relate to
src/java.rmi/share/doc/stub/java/rmi/activation/ActivationGroup_Stub.java?
That file has been present for a long time. It is included when
generating Javadoc, but not when compiling the class files for the JDK.)
It is probably a good idea to make a case-insensitive search for
"rmic" in the make directory afterwards, to confirm that all
remaining cases of rmic should be there (i.e. it relates to compiling
jdk.rmi, not running rmic during the build.)
And btw, I'm eternally grateful to you for removing this. :-) Now the
prospect of a fully warning-free build seems closer than ever!
/Magnus
/Erik
On 2020-03-16 15:19, Roger Riggs wrote:
Hi Erik,
Please review a new webrev that adds the change to remove the
interim build parts.
(Passes Tier 1-3 of CI testing)
http://cr.openjdk.java.net/~rriggs/webrev-stubs-classes-8241073-1/
Thanks, Roger
On 3/16/20 12:22 PM, Erik Joelsson wrote:
Hello Roger,
There is more to be removed in the makefiles.
This file should also be removed:
make/CompileInterimRmic.gmk
In make/Main.gmk, all the targets concerning rmic needs to be
removed as well as any dependencies declared that involves them.
Searching for "rmic" should find all relevant lines.
/Erik
On 2020-03-16 09:02, Roger Riggs wrote:
Please review adding pre-generated RMI stub classes to the jdk repo
and the removal of make files supporting the specific APIs.
It removes a dependency on build time generation invoking RMIC.
RMIC was deprecated in JDK 13 [1].
The source files have been edited to remove or suppress
compilation warnings.
Webrev:
http://cr.openjdk.java.net/~rriggs/webrev-stubs-classes-8241073/
Issue:
https://bugs.openjdk.java.net/browse/JDK-8241073
Thanks, Roger
p.s. A future change will remove the interim build steps
[1] https://bugs.openjdk.java.net/browse/JDK-8217412
