Weird skipping of instruments in the rendering the same code

[anton-k]

The same code is rendered. But produced results are different. Sometimes it skips the instrument.
The strange thing is that naming of the arguments changes from run to run.

The problem is reported on GHC-7.10.3

same-run-diff-files.zip
with-haskell-source.zip

[anton-k]

My guess it has something to do with Hashing algorithm. I use hashing a lot to prevent from rendering the same instrument twice. But it seems that hashing of the expression trees is done wrong or the whole idea of using cashes for this type of problems is unreliable.

[Ebmtranceboy]

Here is a little smaller set
bug14-03-2017-17.41.zip

I tried to remove as much as possible but the bug seems to disappear as soon as I remove even a single note. With this version I normally have an output similar to tmp_ref.csd, but an output similar to tmp_weird.csd can occur from time to time (~ 1 of 6). This is the worst : an instrument is not rendered. More often, I get wrong panning or wrong amplitude (~1 of 4) : tmp_wrongpan.csd.

[Ebmtranceboy]

Smallest example with wrong panning (no instrument loss here)

bug14-03-2017-20.57.zip