We haven't. Elasticsearch does this before calling Tika. Given that the SecurityManager is slated to be removed in a version of Java we'll get to by 2030, I was hesitant.
We have to defend against oom and infinite loops; if we protected against literal System.exit with the SecurityManager, it frankly wouldn't offer much. On Wed, Jul 28, 2021 at 12:05 PM Ken Krugler <kkrugler_li...@transpac.com> wrote: > > Hi Tim, > > Did we ever try this approach of catching the System.exit() call? > > https://stackoverflow.com/questions/5549720/how-to-prevent-calls-to-system-exit-from-terminating-the-jvm > > Though I imagine calling System.setSecurityManager() isn’t appropriate for a > library :) > > — Ken > > > > On Jul 28, 2021, at 7:18 AM, Tim Allison <talli...@apache.org> wrote: > > > > All, > > > > I'm trying to upgrade the surefire plugin in Apache Tika, and I'm > > running into "forked vm terminated without properly saying goodbye" > > [0]. > > > > Tika uses a bunch of dependencies we have no control over. At one > > point, we had a parser with a "feature" that called System.exit. We > > have code that defends against this kind of thing by forking and > > monitoring the forked process for early termination, and we'd like to > > test this forking code. Is there any way to test this with more > > recent versions of the surefire plugin, or is there a recommended > > workaround? > > > > Thank you. > > > > Best, > > > > Tim > > > > [0] > > http://maven.apache.org/surefire/maven-surefire-plugin/faq.html#vm-termination > > -------------------------- > Ken Krugler > http://www.scaleunlimited.com > Custom big data solutions > Flink, Pinot, Solr, Elasticsearch > > >
