realtime recording bug on chained tracks

[Mike]

I just loaded the beta 1.41 to test if this had been fixed, but there is still some issue with realtime recording to chained tracks.

to recreate:

1. chain tracks 1 and 2 in a blank page
2. set track 1 to record enable
3. start recording and play some MIDI into the octopus immediately to record to step 1

You should see that the octopus records a note to step 1 of both tracks in the chain, even though it should be only in track 1.

cheers
ripe

Post edited by: admin, at: 2008/05/11 09:39

[gseher]

I could not reproduce this.
However while trying to do so the Octopus crashed (froze) twice. I have never had the Octo crash before, so there is something fishy here.

[gseher]

Hi guys, the record on step 1 issue is older I guess and while I have identified the cause, it appears like a significant effort to remove. I was hoping we can live with it for just a bit longer, until we have other little buggers out of the way.

But now Lars, you say you were able to crash. Whether that has the same cause is not clear, but more (any!) information on how you did it would be invaluable! Any specifics will help to track down the crashing behaviour and get a fix quickly.

If we find that the step one recording is the cause, priority is obviously going to shift significantly.

[gseher]

I was just realtime recording on chained tracks (1 and 2). It happened twice after recording just a few notes. But I tried again now, and I can not reproduce it…

[Mike]

I don’t know if it helps, but on 1.40 the realtime record was doing something different, it seemed like it was playing all the notes about 5 or 6 steps late. I’m note sure though if it was just skipping the first step because it seemed step 1 had a negative STA (?).

cheers
ripe

[Mike]

Also, it seems that once a track has been chained and armed for realtime record, it is damaged. If you toggle some steps manually, weird things happen to the chained tracks.

cheers
ripe

[gseher]

Thanks John.. I tried to reproduce but did not manage to. Can you reproduce it every time? And give me some more hints?
ripe wrote:

Quote:

Also, it seems that once a track has been chained and armed for realtime record, it is damaged. If you toggle some steps manually, weird things happen to the chained tracks.

Things like what?

[Adam Wilson]

Maybe the actual page number and/or track number matters? That is, the physical place in memory?

[Mike]

I posted a video of the bug happening a couple of times. You have to record a note right at the transition of the tracks. In the video you see that the step for the wrong track is activated, but the sound is in the correct place.

here is the vid:
Link: realtime record bug

Incidentally after recording the video, the octopus froze/crashed and I had to restart.

cheers
ripe

[gseher]

Thanks John, the scenario in the video is reproducible on my end and actually tracked down, as I mentioned earlier.
Is there any more detail you can provide about your findings above re. the "damaged" tracks? That sounds like a very interesting discovery, which I can’t reproduce..

Post edited by: gseher, at: 2008/05/10 16:05

[Mike]

I think it is just an extension of this bug. After it happens, you can no longer activate/deactivate the first step in the second chained track. If you push step 1 in the second track, it activates the first step in the first track. I usually clear a track by toggling steps individually (when working on a specific sequence)

I tried actually clearing the track (with CLR mutator) and it did eliminate the bad step. So I guess it is not permanently damaged.

You said it would take some considerable work to fix this bug though right?

Thanks!

[gseher]

Fixing the setting of the wrong step is a lot of work, as I need to dig quite deep. But preventing the crash should be easier, as soon as I identify it.

Now, I could not reproduce the problem you describe with step 1 of the second track in the chain. I did record the bad step, but then I can turn it off, and back on, and step 1 of track 9 is doing fine as well. Hmm..

[Mike]

I posted a video of the step toggling problem as well, again it looks like it’s just the same issue with the note being incorrectly recorded. When the track is CLRed it goes away.

Link:realtime record 2

I can’t capture the crash on video, it happens fairly randomly.

cheers
ripe

[gseher]

Saw the video – given what I know about the bug, behavior is as expected.
What you see is that the lookahead is not working correctly for track chains.
Lookahead is needed to be able to "pull" the steps. When you play a step in the "lookahead area" it should place it in the next track, which it just doesn’t know yet.

[Adam Wilson]

Uhm, delicate question, but now that the bug videos did their job, maybe it’s a good idea to remove them from YouTube again?

I’m not trying to cover up, I just think that these videos may cause more damage in the end than the bug being discussed, which will no doubt be fixed (in my personal opinion).

[Author]

Posts