From: FFADO <ffa...@ff...> - 2012-07-16 12:28:54
|
#242: ntp interferes with ffado ------------------------+--------------------------------------------------- Reporter: ppalmers | Owner: ppalmers Type: bug | Status: new Priority: major | Milestone: FFADO 2.x Component: | Version: FFADO 2.0-rc2 (1.999.42) Resolution: | Keywords: Device_name: all | ------------------------+--------------------------------------------------- Comment (by jwoithe): Thanks again for the testing. I'm a little unsure of how to proceed from here. As far as I can tell, unmodified r2178 (and presumedly later) works fine on the PC with a good RTC - right? Then on the bad PC, unmodified r2178 fails to start while the modified variant starts but still gives xruns and the like. Please correct where I've misunderstood. Assuming the above is vaguely accurate it does mirror what I'm seeing here. On my test PC (which has a reasonable RTC) I am seeing no practical difference between occasions where CLOCK_MONOTONIC and CLOCK_MONOTONIC_RAW are used. This doesn't surprise me because as far as I can tell both clock sources should be more than capable of supporting FFADO start up. My understanding of these two sources is that for FFADO CLOCK_MONOTONIC_RAW is probably the appropriate one to use. In any case, I am puzzled why with a bad RTC there is such a dramatic difference between the two. My present theory is that perhaps CLOCK_MONOTONIC_RAW - being uncorrected in every way - may be so far off that the values used to initialise the DLLs result in the DLLs not locking. I would have expected them to eventually, but maybe we don't wait long enough for the cases where there's a dramatic difference between theory and practice. Or perhaps it doesn't have a regular period due to some hardware problem (which is possibly related to the inaccurate RTC timekeeping). The start up fail log you attached produces a lot of negative time steps in the cycle timer helper. I find the existence of these interesting but don't know what to make of it yet. However, I do suspect that these negative time steps may be at least partly to blame for the "reconstructed CTR counter discrepancy" messages you get in abundance later on. -- Ticket URL: <http://subversion.ffado.org/ticket/242#comment:59> FFADO <http://subversion.ffado.org/index.fcgi> Free Firewire Audio Drivers |