zero debug still broken after 8189871

Fri Nov 24 17:42:16 UTC 2017

Just built successfully on Debian unstable with:

CONF=linux-x86_64-normal-zero-release MAKE_VERBOSE=y QUIETLY= LOG=debug make clean CONF=linux-x86_64-normal-zero-release ; sh ./configure --with-jvm-variants=zero --with-boot-jdk=/usr/lib/jvm/java-9-openjdk-amd64/ --disable-precompiled-headers --disable-warnings-as-errors && make JOBS=32 MAKE_VERBOSE=y QUIETLY= LOG=debug CONF=linux-x86_64-normal-zero-release

Did a „hg pull && hg update —clean“ prior that.

Luckily, I have an SSH client on my iPhone ;).

Adrian

> On Nov 24, 2017, at 6:26 PM, John Paul Adrian Glaubitz <glaubitz at physik.fu-berlin.de> wrote:
> 
> That’s odd. I just built Zero successfully a few hours ago after „hg pull && hg update —clean“.
> 
> If there had been an issue, I would have run into it.
> 
> Will try to test in an hour, currently on mobile.
> 
> Adrian
> 
>> On Nov 24, 2017, at 6:21 PM, Thomas Stüfe <thomas.stuefe at gmail.com> wrote:
>> 
>> Hi all,
>> 
>> seems that "8191663: Zero variant broken after 8189170 and 8189871" was alone not sufficient to fix zero debug.
>> 
>> When building zero (fast|slow)debug, I get this error:
>> 
>> ---
>> In file included from /shared/projects/openjdk/jdk-hs/source/src/hotspot/share/gc/shared/specialized_oop_closures.hpp:28:0,
>>                 from /shared/projects/openjdk/jdk-hs/source/src/hotspot/share/oops/objArrayOop.cpp:26:
>> /shared/projects/openjdk/jdk-hs/source/src/hotspot/share/oops/access.inline.hpp: In instantiation of ‘void AccessInternal::verify_types() [with long unsigned int decorators = 2048ul; T = volatile oop]’:
>> /shared/projects/openjdk/jdk-hs/source/src/hotspot/share/oops/access.inline.hpp:905:32:   required from ‘T AccessInternal::load(P*) [with long unsigned int decorators = 2048ul; P = volatile oop; T = volatile oop]’
>> /shared/projects/openjdk/jdk-hs/source/src/hotspot/share/oops/access.hpp:450:50:   required from ‘static P Access<decorators>::load(P*) [with P = volatile oop; long unsigned int decorators = 2048ul]’
>> /shared/projects/openjdk/jdk-hs/source/src/hotspot/share/oops/accessBackend.inline.hpp:247:34:   required from ‘static typename EnableIf<AccessInternal::PossiblyLockedAccess<T>::value, T>::type RawAccessBarrier<decorators>::atomic_cmpxchg_maybe_locked(T, void*, T) [with long unsigned int ds = 1030ul; T = oop; long unsigned int decorators = 1030ul; typename EnableIf<AccessInternal::PossiblyLockedAccess<T>::value, T>::type = oop]’ 
>> /shared/projects/openjdk/jdk-hs/source/src/hotspot/share/oops/accessBackend.hpp:325:51:   required from ‘static T RawAccessBarrier<decorators>::atomic_cmpxchg(T, void*, T) [with T = oop; long unsigned int decorators = 1030ul]’
>> /shared/projects/openjdk/jdk-hs/source/src/hotspot/share/oops/access.inline.hpp:619:37:   required from ‘static typename EnableIf<HasDecorator<decorators, 2048ul>::value, T>::type AccessInternal::PreRuntimeDispatch::atomic_cmpxchg(T, void*, T) [with long unsigned int decorators = 289798ul; T = oop; typename EnableIf<HasDecorator<decorators, 2048ul>::value, T>::type = oop]’ 
>> /shared/projects/openjdk/jdk-hs/source/src/hotspot/share/oops/access.inline.hpp:637:71:   required from ‘static typename EnableIf<(! HasDecorator<decorators, 2048ul>::value), T>::type AccessInternal::PreRuntimeDispatch::atomic_cmpxchg(T, void*, T) [with long unsigned int decorators = 287750ul; T = oop; typename EnableIf<(! HasDecorator<decorators, 2048ul>::value), T>::type = oop]’ 
>> /shared/projects/openjdk/jdk-hs/source/src/hotspot/share/oops/access.inline.hpp:821:67:   required from ‘oop AccessInternal::atomic_cmpxchg_reduce_types(oop, HeapWord*, oop) [with long unsigned int decorators = 287748ul]’
>> /shared/projects/openjdk/jdk-hs/source/src/hotspot/share/oops/access.inline.hpp:942:60:   required from ‘T AccessInternal::atomic_cmpxchg(T, P*, T) [with long unsigned int decorators = 262148ul; P = volatile HeapWord; T = oop]’
>> /shared/projects/openjdk/jdk-hs/source/src/hotspot/share/oops/access.hpp:492:78:   required from ‘static T Access<decorators>::oop_atomic_cmpxchg(T, P*, T) [with P = volatile HeapWord; T = oop; long unsigned int decorators = 262144ul]’
>> /shared/projects/openjdk/jdk-hs/source/src/hotspot/share/oops/objArrayOop.cpp:40:78:   required from here
>> /shared/projects/openjdk/jdk-hs/source/src/hotspot/share/utilities/debug.hpp:184:29: error: incomplete type ‘STATIC_ASSERT_FAILURE<false>’ used in nested name specifier
>>   typedef char PASTE_TOKENS(STATIC_ASSERT_DUMMY_TYPE_, __LINE__)[ \
>>                             ^   
>> /shared/projects/openjdk/jdk-hs/source/src/hotspot/share/utilities/macros.hpp:47:33: note: in definition of macro ‘PASTE_TOKENS_AUX2’
>> #define PASTE_TOKENS_AUX2(x, y) x ## y
>>                                 ^   
>> /shared/projects/openjdk/jdk-hs/source/src/hotspot/share/utilities/macros.hpp:45:28: note: in expansion of macro ‘PASTE_TOKENS_AUX’
>> #define PASTE_TOKENS(x, y) PASTE_TOKENS_AUX(x, y)
>>                            ^   
>> /shared/projects/openjdk/jdk-hs/source/src/hotspot/share/utilities/debug.hpp:184:16: note: in expansion of macro ‘PASTE_TOKENS’
>>   typedef char PASTE_TOKENS(STATIC_ASSERT_DUMMY_TYPE_, __LINE__)[ \
>>                ^   
>> /shared/projects/openjdk/jdk-hs/source/src/hotspot/share/oops/access.inline.hpp:873:5: note: in expansion of macro ‘STATIC_ASSERT’
>>     STATIC_ASSERT((HasDecorator<decorators, INTERNAL_VALUE_IS_OOP>::value || // oops have already been validated
>>     ^   
>> --
>> 
>> with
>> 
>> CONFIGURE_COMMAND_LINE:=--with-boot-jdk=/shared/projects/openjdk/jdks/openjdk9 --with-debug-level=fastdebug --with-jvm-variants=zero --with-native-debug-symbols=internal --disable-precompiled-headers --with-build-jdk=../output-release/images/jdk
>> 
>> on Ubuntu 16.4. x64
>> 
>> The error also happens in slow and fastdebug, with precompiled headers enabled or not. Release is fine (in a fashion, I need disable-warnings-as-errors to get it built, but I guess that is another bug).
>> 
>> Thanks, ThomasIt seems triggered by "8189871: Refactor GC barriers to use declarative semantics", but I stared at this error message for half an hour now with not much progress... I do not really understand why the STATIC_ASSERT is failing here. Does anyone have an idea?
>> 
>> 
>> Thanks, Thomas 
>> 