[Bug classpath/108569] New: VMProcess.destroy() throws InternalError for non existing processes

[lgcat76 at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108569

            Bug ID: 108569
           Summary: VMProcess.destroy() throws InternalError for non
                    existing processes
           Product: classpath
           Version: 0.99
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: classpath
          Assignee: unassigned at gcc dot gnu.org
          Reporter: lgcat76 at gmail dot com
  Target Milestone: ---

VMProcess.destroy() may arbitrarily throw InternalError for recently finished
processes.

Taking a look to the implementation, and following it to the native part of the
implementation of VMProcess.nativeKill() (in java_lang_VMProcess.c) it can be
seen that if the value returned by cpproc_kill() is ESRCH (the error code
indicated by the kill() system call for non-existing processes) then an
InternalError is explicitly thrown. But this can happen under normal conditions
if the process is reaped, for example, just before the call to nativeKill().

This can be tested by including an artificial delay in VMProcess.destroy() just
before calling nativeKill(), to ease setting the conditions for a process being
reaped just before the call:

--------------------------------
  public synchronized void destroy()
  {
    if (state == TERMINATED)
      return;

    try { Thread.sleep(10*1000L); } catch (InterruptedException e) {}
    nativeKill(pid);
    ...
--------------------------------


Then a test program like this (tested with jamvm):


--------------------------------
import java.io.IOException;

public class TestDestroy
{
    public static void main(String[] args) throws IOException,
InterruptedException
    {
        System.out.println("Launching process...");
        Process process = Runtime.getRuntime().exec("sleep 5");

        System.out.println("Waiting 1s before destroy...");
        Thread.sleep(1000);

        System.out.println("Calling destroy...");
        process.destroy();
    }
}
--------------------------------

Produces the following output:

--------------------------------
Launching process...
Waiting 1s before destroy...
Calling destroy...
Exception in thread "main" java.lang.InternalError: kill(263): No such process
   at java.lang.VMProcess.nativeKill(Native Method)
   at java.lang.VMProcess.destroy(VMProcess.java:362)
   at TestDestroy.main(TestDestroy.java:14)
--------------------------------


I just think that the ESRCH error code should be handled as success.