Debugging segmentation faults in the JVM on linux-powerpc

coleen.phillimore at oracle.com coleen.phillimore at oracle.com
Sat Jun 10 19:21:13 UTC 2017


It looks like you're building the zero interpreter which doesn't get the 
same level support as other platforms, well, it gets actually almost no 
support.

What I would do is create two repositories, build one without the zero 
options.   Build the other with --with-build-jdk=<the just built jdk> 
and zero.  Debug zero like:

gunzip libjvm.diz where it is built
gdb --args <path to build/jdk>java -version  (or whatever fails).

Good luck,
Coleen


On 6/10/17 1:09 PM, John Paul Adrian Glaubitz wrote:
> On 06/09/2017 08:57 PM, John Paul Adrian Glaubitz wrote:
>> It's still building, let's see if it actually succeeds.
> It does. And it fails again with "--with-debug-level=fastdebug":
>
> #  Internal Error (/<<PKGBUILDDIR>>/src/hotspot/src/os_cpu/linux_zero/vm/os_linux_zero.cpp:260), pid=46604, tid=46606
> #  fatal error:
> #
> #    /--------------------\
> #    | segmentation fault |
> #    \---\ /--------------/
> #        /
> #    [-]        |\_/|
> #    (+)=C      |o o|__
> #    | |        =-*-=__\
> #    OOO        c_c_(___)
> #
> # JRE version: OpenJDK Runtime Environment (9.0) (fastdebug build 9-Debian+0-9b170-2)
> # Java VM: OpenJDK Zero VM (fastdebug 9-Debian+0-9b170-2, interpreted mode, serial gc, linux-ppc)
> # No core dump will be written. Core dumps have been disabled. To enable core dumping, try "ulimit -c unlimited" before starting Java again
>
> Logs:
>
> --with-debug-level=slowdebug (succeeds):
>
>> https://buildd.debian.org/status/fetch.php?pkg=openjdk-9&arch=powerpc&ver=9%7Eb170-2&stamp=1497054989&raw=0
> --with-debug-level=fastdebug (fails):
>
>> https://buildd.debian.org/status/fetch.php?pkg=openjdk-9&arch=powerpc&ver=9%7Eb170-2&stamp=1497076948&raw=0
> Does any of the JVM wizards have an idea? I haven't looked at the code yet myself :).
>
> Adrian
>



More information about the hotspot-dev mailing list