Memory leak in Dispatch.cpp

[EJP286CRSKW]

There is a memory leak around the call to name2ID(), in the case where it fails. Code reads:

  if (name != NULL) 
  {
    const char *nm = env->GetStringUTFChars(name, NULL);
    HRESULT hr;
    if (FAILED(hr = name2ID(pIDispatch, nm, (long *)&dispID, lcid))) {
      char buf[1024];
      sprintf_s(buf, 1024, "Can't map name to dispid: %s", nm);
      ThrowComFail(env, buf, -1);
      return NULL;
    }
    env->ReleaseStringUTFChars(name, nm);
  }

It should read:

  if (name != NULL) 
  {
    const char *nm = env->GetStringUTFChars(name, NULL);
    HRESULT hr = name2ID(pIDispatch, nm, (long *)&dispID, lcid));
    env->ReleaseStringUTFChars(name, nm);
    if (FAILED(hr)) {
      char buf[1024];
      sprintf_s(buf, 1024, "Can't map name to dispid: %s", nm);
      ThrowComFail(env, buf, -1);
      return NULL;
    }
  }

[freemansoft]

You are saying that the leak is because the release string is skipped if it fails because the throw() exits the function before ReleaseString is called?

Ugh. I'll have to look at putting my dev environment back together.

[freemansoft]

How did you find this?

[EJP286CRSKW]

The ReleaseStringUTFChars() is skipped because of the premature return NULL; in the failure block.

I found it by reading the code.

[freemansoft]

I think

    HRESULT hr = name2ID(pIDispatch, nm, (long *)&dispID, lcid));

Should be

    HRESULT hr = name2ID(pIDispatch, nm, (long *)&dispID, lcid);

5853381

[EJP286CRSKW]

Correct, well spotted.