I have a strange issue sometimes that crops up where if i connect a new node it does not pass on values to the next node downstream on the patch - as if there is no connection.
anyone come across this before?
Yes, I did, on occaision.
Sometimes, If I make a new connection, the old connection also remains, ending up having 2 connections on 1 input pin.
After deleting both connection (select and press the Delete key) and reconnect again, it seems to work.
Or sometimes I have to delete and make a new node.
I think it is a bug, but have not been able to reproduce it.
right, so its not just my wackey patchn’ style then! :)
ive got some big renderpass loops going on over here, so this problem becomes even bigger!
as you cannot add one of the problem nodes/connections to the loop without re-patching the whole thing it seems!
cheers!
strange issue sometimes
have not been able to reproduce it.
yes, thats a major (if not the major) bug.
a way to reproduce the bug would be a great help.
ok yes. this seems to be an unfriendly bug.
so the problem is that we need to be able to reproduce it, because it is not quite clear where to search for it. does the system output any log message or warning to the tty renderer when this happens? is it just that gui and core graph are not synchronous? (…)
what do you think:
is it somehow related to feedback loops?
is it related to some specific node? can you tell which nodes are involved?
can you save the patch in the moment when it fails and share it with us?
thx
Hi!
@gregsn:
- tty renderer says nothing
- no feedback loops but bug seems to occur more often when connecting subpatches.
- it happens after the patch is loaded. close the (perfectly fine) patch, quit vvvv, open the patch: and there is the bug. not always. just sometimes. But when it happens, it happens at the same place and the same connections.
- Could be node specific: IOBox
@oschatz: I don’t have it here. But it’s on Endo. Don’t use the latest version, but one or two before the last. no need to send textures and stuff. just the patches
ahlo. we do have a case here. it is already in the debugger. a second case to check against would not be bad though.
we had that case here, that values are passed on even when a connection was cut. i could find the bug and solve this particular problem. however if you have a patch with the opposite problem (that values are not passed even if there is a connection) then please hand out that patch to me, so that i can see if this problem persists.
background info:
the problem was somehow connected to the SliceMode in an IOBox (ColsRowsPages). internally SliceMode “Input” is an optimized case where data is passed from nodes above the iobox to nodes below the iobox. actually this case is only really really optimized when the iobox doesn’t have to show the values (that is e.g. when the patch is hidden).
ColsRowsPages cannot be optimized because the output might have another spreadcount (defined over Cols, Rows, Pages) and therefore must be resampled somehow. so this is why “Input” is the default mode for ioboxes.
so. I have tried the patch, that produced the bug with a newly compiled vvvv version. I couldnt’t reproduce the error anymore. With the current beta the bug can still be reproduced.
So it seems to be fixed.
many thanx
nice… no text …