<div class="gmail_quote">On Fri, Jun 19, 2009 at 8:06 PM, John Arbash Meinel <<a href="mailto:john@arbash-meinel.com">john@arbash-meinel.com</a>> wrote: <blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"> Are you using the line written on stdout to determine what test is running? IIRC that uses the 'progress_bar' code which is designed to 'debounce' the output. (It only updates the screen every 200ms or so.) Since the test that is passing is only taking 49ms, and I would imagine the next few tests are also probably fast, that in the end it is still <div class="im"> test_http.TestAuth.test_wrong_pass(pycurl,HTTP/1.1,basicdigest) </div>which is hanging, it just didn't update the screen first. John =:-></blockquote><div> Ooh! Yes, in the non-verbose case I am using the last thing on stdout. And, no, I did not know about the debounce buffering. When I ctrl-C, I imagine the SIGHUP is not trapped so there's no chance to flush buffers. So your theory makes a lot of sense. </div></div> But hold on... wouldn't the failing test be the other way around? non-verbose (uses progress bar) stalls at test_http.TestAuth.test_wrong_pass(pycurl,HTTP/1.1,basicdigest) but can't be trusted because of debounce *and* verbose (does not use progress bar) stalls at test_http.TestActivity.test_get(urllib,HTTP/1.0,https) which can be trusted because a separate line is written to stdout for every test. *and* in non-verbose mode test_http.TestAuth.test_wrong_pass(pycurl,HTTP/1.1,basicdigest) actually *passes* So your theory applies but I think the test actually failing in both cases is test_http.TestActivity.test_get(urllib,HTTP/1.0,https) I'm guessing that's what you actually meant. So now I not only have pycurl problems (did I mention I also get a few XFAIL on pycurl?) I also have a urllib problem. Yikes! Time to soak brain. -M