corrupt messages in build logs
Kelly.Ohair at oracle.com
Thu Feb 3 02:23:38 UTC 2011
On Feb 2, 2011, at 6:10 PM, Dr Andrew John Hughes wrote:
> On 16:43 Wed 02 Feb , Kelly O'Hair wrote:
>> On Feb 2, 2011, at 1:56 PM, Jonathan Gibbons wrote:
>>> Some messages in the build logs are getting corrupted, perhaps
>>> because of parallism. For example,
>>> ../../../src/plugin/solaris/jvm_natives/server.c:2031: warning: cast
>>> to pointer from integer of different
>>> warning: ignoring #pragma i
>>> ../../../src/plugin/solaris/jvm_natives/native.c:165: warning: type
>>> defaults to 'int' in declaration of 'getAw
>>> warning: /localtools/linux/gcc29/usr/bin/gcc -O2 -fPIC -DCC_NOEX -W
>>> -Wall -Wno-unused -Wno-parentheses -fno-omit-frame-pointer -
>>> D_LITTLE_ENDIAN -I../../../../src/plugin/solaris/nscore -
>>> I../../../../src/plugin/solaris/nscore/../common -I/HUDSON/
>>> jdk7-2-build-linux-i586-product/jdk7/build/linux-i586/tmp/ -I/
>>> deploy/headers/plugin -I../../../../src/plugin/solaris/common -
>>> I../../../../src/plugin/share/jpiapi -I../../../../src/plugin/
>>> solaris/nscore/../nsstub -I/HUDSON/workspace/jdk7-2-build-linux-
>>> product/jdk7/jdk/make/../src/solaris/javavm/export -I/HUDSON/
>>> javavm/export -D_GNU_SOURCE -DXP_UNIX -DNO_XEMBED_MOZILLA -DSVR4 -
>>> DPLUGIN_NODOTVERSION=\"170\" -DPLUGIN_VERSION=\"1.7.0\" -DVERSION=
>>> \"1.7.0-ea-b122\" -DRELEASE=\"1.7.0-ea\" -DARCH=\"i586\" -Di586 -
>>> DLIBARCH=\"i386\" -B /localtools/linux/gcc29/usr/bin//../lib/gcc-
>>> lib/ -I /localtools/linux/gcc29/usr/bin//../lib/gcc-lib/i386-redhat-
>>> linux/egcs-2.9*/include -DRAPTOR_API -DNEW_STREAMING_API -DXP_UNIX -
>>> DLINUX -c -o /HUDSON/workspace/jdk7-2-build-linux-i586-product/
>>> QueueRunnable.o ../../../../src/plugin/solaris/nscore/
>>> Is this known and/or fixable?
>> This is due to the 'make -j 4 *.o' that happens when the sharedly
>> library is built.
>> I though GNU make did a better job on keeping the messages grouped
>> with the compile line, but maybe not.
> I haven't seen an issue with this from GNU make before, but isn't
> the above log
> on Solaris rather than GNU/Linux?
We use GNU make on both, but I think the above is on Linux.
> It also doesn't look like OpenJDK to me.
Yup, looks like mozilla plugin code to me, using the old gcc29 compiler.
Maybe it's a gcc29 compiler issue?
>> It may have to echo out the compile line first. Not sure it can be
>> fixed easily.
>> Maybe build with ALT_PARALLEL_COMPILE_JOBS=1
> Nice and slow... ;-)
>>> -- Jon
> Andrew :)
> Free Java Software Engineer
> Red Hat, Inc. (http://www.redhat.com)
> Support Free Java!
> Contribute to GNU Classpath and IcedTea
> PGP Key: F5862A37 (https://keys.indymedia.org/)
> Fingerprint = EA30 D855 D50F 90CD F54D 0698 0713 C3ED F586 2A37
More information about the build-dev