Problems with sync from jdk8/jdk8 to ppc-aix-port/stage

Thu Nov 7 07:34:23 PST 2013

Hi Vladimir,

thank you for doing the sync!

Strange that your builds passed on Oracle platforms because because I
think there's a typo in makefiles/lib/CoreLibraries.gmk which should
affect all platforms:

diff -r d152c5b01ea8 makefiles/lib/CoreLibraries.gmk

--- a/makefiles/lib/CoreLibraries.gmk   Tue Nov 05 17:32:53 2013 -0800
+++ b/makefiles/lib/CoreLibraries.gmk   Thu Nov 07 15:36:54 2013 +0100
@@ -269,7 +269,7 @@
         $(WIN_JAVA_LIB), \
     LDFLAGS_SUFFIX_linux := -ljvm -ljava $(LIBZ), \
     LDFLAGS_SUFFIX_solaris := -ljvm -ljava $(LIBZ) -lc, \
-    LDFLAGS_SUFFIX_aix: = -ljvm -ljava $(LIBZ),\
+    LDFLAGS_SUFFIX_aix := -ljvm -ljava $(LIBZ),\
     LDFLAGS_SUFFIX_macosx := $(LIBZ) -ljava -ljvm, \
     VERSIONINFO_RESOURCE := $(JDK_TOPDIR)/src/windows/resource/version.rc, \
     RC_FLAGS := $(RC_FLAGS) \

Maybe that's related to the make version, but with GNU make 3.82 I get
an "*** empty variable name" error.

If I fix this (and together with Goetz's fixes) the build passes on Linux/PPC64.

I think it makes no sense to create a bug for this single character
change so I'm packing it together with the other AIX-related changes
which I'm currently preparing.

Regards,
Volker


On Wed, Nov 6, 2013 at 11:33 PM, Vladimir Kozlov
<vladimir.kozlov at oracle.com> wrote:
> Hi, Volker
>
> I finished the sync.
>
> Hotspot passed clean. JDK passed builds and testing on our platforms with
> few known failures (2 javac regression tests TypeInferenceComboTest.java and
> DefaultMethodsTest.java).
>
> Yes, I did not expect they will do such global cleanup so later in the game
> for jdk8:
>
> https://bugs.openjdk.java.net/browse/JDK-8001931
> http://hg.openjdk.java.net/jdk8/jdk8/rev/174a54ce39c4
> http://hg.openjdk.java.net/jdk8/jdk8/jdk/rev/9c60860b1812
>
> Syncing more frequent will not help with such big changes. I only hope it
> will not happen again in near future until we merge PPC64 code into main
> forest.
>
> As I said I did my best to resolve conflicts and, I think, the only problem
> left is AIX changes in makefiles/CompileNativeLibraries.gmk which you need
> to move to new makefiles/lib/*.gmk files.
> File a bug "PPC64: 8025715 changes broke AIX build after sync" and send me
> the fix. I will have to verify JDK build on our platforms with it.
>
> I am fine with resolving easy conflicts and leaving hard one to you :)
>
> Regards,
> Vladimir
>
> PS: even mailing system freak-out :)
> --------------------------------------------------------------
> Your mail to 'ppc-aix-port-dev' with the subject
>
>     hg: ppc-aix-port/stage/jdk: 740 new changesets
>
> Is being held until the list moderator can review it for approval.
> The reason it is being held:
>
>     Message body is too big: 456885 bytes with a limit of 400 KB
> --------------------------------------------------------------
>
>
> On 11/6/13 8:35 AM, Volker Simonis wrote:
>>
>> Hi Vladimir,
>>
>> thank you for starting syncing again!
>>
>> Please see my comments inline:
>>
>> On Wed, Nov 6, 2013 at 3:15 AM, Vladimir Kozlov
>> <vladimir.kozlov at oracle.com> wrote:
>>>
>>> I was not able to resolve all conflicts. AIX build is definitely broken.
>>>
>>> I have to manually resolve a lot of .m4 and .gmk files (mostly changed
>>> indention).
>>>
>>> I had problem with common/autoconf/platform.m4 when resolving conflicts
>>> at
>>> the end of file (below "Make compilation sanity check"). And for
>>> ADDED_*FLGAS settings I resolved it as in ppc64 repo:
>>>
>>> ppc64 repo:
>>>
>>>    ADDED_CFLAGS=" ${COMPILER_TARGET_BITS_FLAG}${OPENJDK_TARGET_CPU_BITS}"
>>>    ADDED_CXXFLAGS="
>>> ${COMPILER_TARGET_BITS_FLAG}${OPENJDK_TARGET_CPU_BITS}"
>>>    ADDED_LDFLAGS="
>>> ${COMPILER_TARGET_BITS_FLAG}${OPENJDK_TARGET_CPU_BITS}"
>>>
>>
>> Yes, this is the right way how it should be done now that we have
>> ${COMPILER_TARGET_BITS_FLAG}
>>
>>> jdk8 repo:
>>>
>>>    ADDED_CFLAGS=" -m${OPENJDK_TARGET_CPU_BITS}"
>>>    ADDED_CXXFLAGS=" -m${OPENJDK_TARGET_CPU_BITS}"
>>>    ADDED_LDFLAGS=" -m${OPENJDK_TARGET_CPU_BITS}"
>>>
>>>
>>> Merging toolchain.m4 was even more painful.
>>>
>>> An other one was jdk/makefiles/CompileLaunchers.gmk
>>>
>>> But the main problem was jdk/makefiles/CompileNativeLibraries.gmk. Code
>>> from
>>> it was moved into jdk/makefiles/lib files:
>>>
>>> http://hg.openjdk.java.net/jdk8/jdk8/jdk/rev/a5b57fca66da
>>>
>>> I was not able to do anything there. It definitely broke AIX build so you
>>> need to implement it again. Sorry.
>>
>>
>> Well, works as expected:(
>>
>> I know no good way how we could solve these problems (apart from
>> integration shared changes more early into the main repositories of
>> course).
>>
>> I could of course offer to do the merge such that all the AIX changes
>> would still be good. But how could I share a merge change with you?
>> Would you like to get only the files which I had to resolve manually?
>> Or better all the files which are changed by the merge? You could then
>> merge, copy my files over the sources and do the remaining resolving
>> in the closed directories.
>>
>> What do you think, should we try it this way next time?
>>
>>>
>>> I started JDK and Hotspot JPRT test runs. If they passed I will push what
>>> I
>>> have and then you need to fix it for AIX.
>>>
>>
>> I'm ready to start...
>>
>>> Regards,
>>> Vladimir
>>>
>
