SIGSEGV on PhaseIdealLoop::split_up?

Uwe Schindler uschindler at apache.org
Wed Jan 30 14:06:47 UTC 2019


Hi,

thanks Dawid. The builds on the Jenkins servers don't run with downloadable releases of Lucene. To reproduce one build, a direct GIT checkout of Lucene branch with used commit hash needs to be done.

I have to add to Dawid's explanation: When you run the testsuite of Lucene to reproduce, you should also use the "seed" of the test run. Whenever Lucene runs tests a so called random seed is used which configures the test suite to inject random data or random components during the test run. This allows us to test Lucene with many configurations -- and of course this also finds bugs in the JVM, because every run of the test suite is "unique". If you find a test failure or JVM crush, it's important to reproduce this with the same seed.

The Lucene tests framework prints the hash used before it runs tests. To run it again, you can tell the build script to use the same seed. Type "ant test-help" to get usage help.

Uwe

-----
Uwe Schindler
uschindler at apache.org 
ASF Member, Apache Lucene PMC / Committer
Bremen, Germany
http://lucene.apache.org/

> -----Original Message-----
> From: hotspot-dev <hotspot-dev-bounces at openjdk.java.net> On Behalf Of
> Dawid Weiss
> Sent: Wednesday, January 30, 2019 12:11 PM
> To: Nils Eliasson <nils.eliasson at oracle.com>
> Cc: hotspot-dev <hotspot-dev at openjdk.java.net>; Uwe Schindler (SD
> DataSolutions GmbH) <uwe at thetaphi.de>
> Subject: Re: SIGSEGV on PhaseIdealLoop::split_up?
> 
> Hi Nils,
> 
> Those builds are made straight up from git (from various branches).
> For example this failure:
> 
> https://jenkins.thetaphi.de/job/Lucene-Solr-7.x-Linux/3472/
> 
> with these hs_err and replay files:
> 
> https://jenkins.thetaphi.de/job/Lucene-Solr-7.x-
> Linux/3472/artifact/solr/build/solr-core/test/J1/hs_err_pid27685.log
> https://jenkins.thetaphi.de/job/Lucene-Solr-7.x-
> Linux/3472/artifact/solr/build/solr-core/test/J1/replay_pid27685.log
> 
> comes from rev db57468242 of git at github.com:apache/lucene-solr.git
> (the revision is mentioned on jenkins and in the full log), so:
> 
> git clone git at github.com:apache/lucene-solr.git
> cd lucene-solr
> git checkout db57468242
> 
> then you can compile with:
> 
> cd lucene
> ant jar
> 
> Dawid
> 
> On Wed, Jan 30, 2019 at 12:03 PM Nils Eliasson <nils.eliasson at oracle.com>
> wrote:
> >
> > Hi,
> >
> > With the help of the replay-file I manange to compile the right Class,
> > but the inlining doesn't match. The latest release i see is 7.6, but it
> > looks like the crash is from a 9.0? Is that master? Do you have a link
> > to a build that I can download?
> >
> > Regards,
> >
> > Nils
> >
> > On 2019-01-30 11:55, Dawid Weiss wrote:
> > >> This is release build, right? fastdebug build probably asserts
> somewhere?
> > > I don't think we (or Uwe) runs jobs with fastdebug builds, to be
> > > honest. This isn't a bad idea though.
> > >
> > >> If Nils is not there, let Uwe find me at FOSDEM?
> > > CCing: Uwe.
> > >
> > > D.



More information about the hotspot-dev mailing list