There can be extensions written in Java installed in StarOffice/OpenOffice.org. This Java code is executed in a JVM that is started in-process via the Java Invocation API.

Now, I have been approached by a customer who claims that in the above scenario, such Java extension code like

  try {
    f1().f2();
  } catch (Exception e) { ... }

where the call to f1 returns null does not lead to a NullPointerException caught in the third line, but instead to a SIGSEGV. They claim this works on Windows but fails on Solaris x86, JRE 1.5. I could not (easily) reproduce this, however. They say that changing the code to

  T x = f1();
  if (x != null) x.f2();
  else ...

works as expected and does not cause the SIGSEGV crash. Attached is a hs_err_pid1208.log of theirs.