Per Östlund

I had to add another flag `--arrayConnectMinSize` for setting the minimum size of array connections for the array connection handling to be used, since we have test cases that don't...

@casella: Yes, the first point to fix was to only activate the array connection handling if there's actually array connections present. I thought that wasn't a good enough condition since...

> I'm still unsure about this heuristics. I'd say 100 is too large anyway, why would we want to generate 99 separate code chunks to handle 99 connect statements when...

@casella: The error message is from the array connection handling, like I mentioned earlier it seems to have some issues with nested for-loops (the `$i1` is most likely an iterator...

Ok, so this is a bit of a mess. - If you only use `-d=nonfScalarize` then you get the old behaviour which doesn't scalarize anything and only works for some...

@rfranke can perhaps comment on whether the non-scalarized test cases are still relevant for him, and whether he really needs the specific behaviour of the connection handling that they use...

> `--newBackend` does imply not scalarizing > > ``` > // set implied flags to true > FlagsUtil.enableDebug(Flags.SCODE_INST); > FlagsUtil.enableDebug(Flags.ARRAY_CONNECT); > FlagsUtil.disableDebug(Flags.NF_SCALARIZE); > ``` > > is at the start...

@casella: Dividing the splitting and scalarization of connectors into two separate phases in #9733 was mainly to allow such an approach. The analysis should as you say be fairly easy...

The `NFSBGraphUtil.intervalFromExp got unknown expression` error should be fixed in #9838. The issue was just that the array connection algorithm tried to replace the iterators in the crefs, but the...

> @perost did you manage tasks 2 and 3 above? I'm eager to see what happens with the new backend when it is passed correct connection equations. > > Thanks!...