Hi Kurt, Kurt Miller wrote:
Hi Andrei,
If I understand your email correctly:
1) bsd-port will not build with gcc 4.2 due to the below compiler error
Correct.
2) bsd-port builds fine using soylatte with gcc 4.0.1 on MacOSX and appears to work fine to
Correct.
3) The previous build crashes were due to env vars or other issues which are no longer a problem.
Yes, that happened because of env. variables and g++ 4.2. Thanks, Andrei
Does that summarize the current state correctly?
Regards, -Kurt
Andrei Dmitriev wrote:
Well, I meant that the build itself is crashing with the error below... :(
Also just verified that I have a fresh bsd repo with latest merge with the main OpenJDK:
The thing is that if I change the variant to debug_build then the following compile-time error occur: In file included from ../generated/incls/_ad_x86_32.cpp.incl:1, from ../generated/adfiles/ad_x86_32.cpp:29: ../generated/adfiles/ad_x86_32.hpp: In member function 'virtual int countLeadingZerosINode::ideal_Opcode() const': ../generated/adfiles/ad_x86_32.hpp:4534: error: 'Op_CountLeadingZerosI' was not declared in this scope ../generated/adfiles/ad_x86_32.hpp: In member function 'virtual int countLeadingZerosI_bsrNode::ideal_Opcode() const': ../generated/adfiles/ad_x86_32.hpp:4558: error: 'Op_CountLeadingZerosI' was not declared in this scope ../generated/adfiles/ad_x86_32.hpp: In member function 'virtual int countLeadingZerosLNode::ideal_Opcode() const': ../generated/adfiles/ad_x86_32.hpp:4582: error: 'Op_CountLeadingZerosL' was not declared in this scope ../generated/adfiles/ad_x86_32.hpp: In member function 'virtual int countLeadingZerosL_bsrNode::ideal_Opcode() const': ../generated/adfiles/ad_x86_32.hpp:4607: error: 'Op_CountLeadingZerosL' was not declared in this scope ../generated/adfiles/ad_x86_32.hpp: In member function 'virtual int countTrailingZerosINode::ideal_Opcode() const': ../generated/adfiles/ad_x86_32.hpp:4632: error: 'Op_CountTrailingZerosI' was not declared in this scope ../generated/adfiles/ad_x86_32.hpp: In member function 'virtual int countTrailingZerosLNode::ideal_Opcode() const': ../generated/adfiles/ad_x86_32.hpp:4656: error: 'Op_CountTrailingZerosL' was not declared in this scope ../generated/adfiles/ad_x86_32.cpp: At global scope: ../generated/adfiles/ad_x86_32.cpp:1079: error: no 'const bool Matcher::match_rule_supported(int)' member function declared in class 'Matcher' ../generated/adfiles/ad_x86_32.cpp: In member function 'virtual void countLeadingZerosINode::emit(CodeBuffer&, PhaseRegAlloc*) const': ../generated/adfiles/ad_x86_32.cpp:8642: error: 'class MacroAssembler' has no member named 'lzcntl' ../generated/adfiles/ad_x86_32.cpp: In member function 'virtual void countLeadingZerosI_bsrNode::emit(CodeBuffer&, PhaseRegAlloc*) const': ../generated/adfiles/ad_x86_32.cpp:8660: error: 'class MacroAssembler' has no member named 'bsrl' ../generated/adfiles/ad_x86_32.cpp: In member function 'virtual void countLeadingZerosLNode::emit(CodeBuffer&, PhaseRegAlloc*) const': ../generated/adfiles/ad_x86_32.cpp:8684: error: 'class MacroAssembler' has no member named 'lzcntl' ../generated/adfiles/ad_x86_32.cpp:8686: error: 'class MacroAssembler' has no member named 'lzcntl' ../generated/adfiles/ad_x86_32.cpp: In member function 'virtual void countLeadingZerosL_bsrNode::emit(CodeBuffer&, PhaseRegAlloc*) const': ../generated/adfiles/ad_x86_32.cpp:8708: error: 'class MacroAssembler' has no member named 'bsrl' ../generated/adfiles/ad_x86_32.cpp:8713: error: 'class MacroAssembler' has no member named 'bsrl' ../generated/adfiles/ad_x86_32.cpp: In member function 'virtual void countTrailingZerosINode::emit(CodeBuffer&, PhaseRegAlloc*) const': ../generated/adfiles/ad_x86_32.cpp:8735: error: 'class MacroAssembler' has no member named 'bsfl' ../generated/adfiles/ad_x86_32.cpp: In member function 'virtual void countTrailingZerosLNode::emit(CodeBuffer&, PhaseRegAlloc*) const': ../generated/adfiles/ad_x86_32.cpp:8758: error: 'class MacroAssembler' has no member named 'bsfl' ../generated/adfiles/ad_x86_32.cpp:8760: error: 'class MacroAssembler' has no member named 'bsfl' ../generated/adfiles/ad_x86_32.cpp: At global scope: ../generated/adfiles/ad_x86_32.cpp:22955: error: too many initializers for 'const bool [310]' make[7]: *** [ad_x86_32.o] Error 1 make[6]: *** [the_vm] Error 2 make[5]: *** [jvmg] Error 2 make[4]: *** [generic_build2] Error 2 make[3]: *** [jvmg] Error 2 make[2]: *** [hotspot-build] Error 2 make[1]: *** [generic_debug_build] Error 2 make: *** [build_debug_image] Error 2 real 61.52 user 31.24 sys 17.23
Suspecting that g++ causing this I tried g++4.0.1 (I upgraded the compiler to 4.2 some time ago for some purpose...) and that was the reason so at least java/javac are now built and I could compile a test and show some GUI now.
Thanks, Andrei
Kurt Miller wrote:
Hi Andrei,
Hmm, it would be helpful to do a debug build. I see you used VARIANT=DBG, but doesn't appear to be working as expected. Can you rebuild with this for your last line of the script?
time -p make debug_build $* 2>&1 | tee $log
That will get some debug info into libjvm.dylib and hopefully we can see the function and line number that is crashing.
Regards, -Kurt
Andrei Dmitriev wrote:
Kurt,
yep, I used these variables since that time because noticed that they don't really affect the build process itself - I got some binaries and never tried how they work. Now I see that variables... do affect JDK somehow. :( So here is the script: -------------------------------- #!/bin/bash
export ALT_BOOTDIR=/export/devtools/binary.plugs/soylatte16-i386-1.0.3/ export ANT_HOME=/usr/share/ant export NO_DOCS=true export HOTSPOT_BUILD_JOBS=1; export ALT_FREETYPE_HEADERS_PATH=/usr/X11R6/include; export ALT_FREETYPE_LIB_PATH=/usr/X11R6/lib; export ALT_JIBX_LIBS_PATH=/export/devtools/jibx/lib/
log=bsdport.build.log
time -p make VARIANT=DBG $* 2>&1 | tee $log -------------------------------- and it leads to the same error:
make[7]: `/Users/dav/bsd-port-reference/build/bsd-i586/hotspot/outputdir/bsd_i486_compiler2/product/../generated/sa-jdi.jar'
is up to date. make[6]: `incls/_precompiled.incl.gch' is up to date. if [ -d /Users/dav/bsd-port-reference/hotspot/agent -a "x86" != "ia64" ] ; then \ make -f vm.make libsaproc.dylib; \ fi make[7]: `libsaproc.dylib' is up to date. All done. cd bsd_i486_compiler2/product && ./test_gamma java full version "1.6.0_03-p3-landonf_19_aug_2008_14_55-b00" # # A fatal error has been detected by the Java Runtime Environment: # # SIGBUS (0xa) at pc=0x00e8d50f, pid=40704, tid=2685097760 # # JRE version: 6.0_03 # Java VM: OpenJDK Server VM (16.0-b03 mixed mode bsd-x86 ) # Problematic frame: # V [libjvm.dylib+0xb150f] # # An error report file with more information is saved as: # /Users/dav/bsd-port-reference/build/bsd-i586/hotspot/outputdir/bsd_i486_compiler2/product/hs_err_pid40704.log
# # If you would like to submit a bug report, please visit: # http://java.sun.com/webapps/bugreport/crash.jsp # ./test_gamma: line 12: 40704 Abort trap ./${gamma:-gamma} -Xbatch -showversion Queens < /dev/null make[4]: *** [product] Error 134 make[3]: *** [generic_build2] Error 2 make[2]: *** [product] Error 2 make[1]: *** [hotspot-build] Error 2 make: *** [build_product_image] Error 2 real 41.36 user 32.39 sys 15.49
Also verified that JAVA_HOME, ALT_HOTSPOT_IMPORT_PATH, ALT_JDK_IMPORT_PATH not set. I'm building from the bsd-port/ and bsd-port/hotspot/ folders with similar results listed above. :(
Thanks, Andrei
Kurt Miller wrote:
Andrei Dmitriev wrote:
Hi folks,
I'm experiencing a crash with the latest javac and java build on MacOSX: Any ideas what's going on there? I think it's not an issue with the latest merge because I've seen that crash that week too.
../build/bsd-i586/bin/javac # # An unexpected error has been detected by Java Runtime Environment: # # Internal Error (554E534146450E43505004DA), pid=97421, tid=0xb0051000 # # Java VM: Java HotSpot(TM) Client VM (1.6.0_03-p3-landonf_19_aug_2008_14_55-b00 mixed mode)
^^^^^^^^^^^^^^^^^^^^^^
Hi Andrei,
That looks like the soylatte hotspot not the bsd-port/openjdk7 hotspot. Did you build with ALT_HOTSPOT_IMPORT_PATH?
ALT_HOTSPOT_IMPORT_PATH and ALT_JDK_IMPORT_PATH can not point to a soylatte build. Leave them unset to build bsd-port. See the following email for an example of how to get a clean build of bsd-port on OS X:
http://mail.openjdk.java.net/pipermail/bsd-port-dev/2009-June/000735.html
Regards, -Kurt