After the check in to bug 176501, we got a complaint that the leaks increased by 10 K. http://tinderbox.mozilla.org/showbuilds.cgi?tree=Firefox&hours=24&maxdate=1165472224&legend=0 On the above tinderbox page, look at machine balsa-trunk. Look for checkins from "kaie". Before my bunch of check ins, the Lk count was 268 KB, afterwards it was 279 KB. List of all chechins between checkpoints: http://bonsai.mozilla.org/cvsquery.cgi?module=PhoenixTinderbox&date=explicit&mindate=1165419300&maxdate=1165422239 The patch from bug 176501 looks like the only real candidate. According to my testing/tracing, we call the Load function only once, and we do call the Unload+Destroy on exit. It seems the leak must be somewhere internal in NSS?
This bug appears to be mere speculation of a leak in NSS. There is no stack trace. We can never know if this is fixed. It is not qualified as a confirmed memory leak bug. So I am resolving it incomplete. If it is a real leak, presumably other leak bugs will be filed with real stacks.