[LAD] Strange Jack1 problem

[Simon Jenkins]

> On 11 Aug 2015, at 19:48, Fons Adriaensen <fons at linuxaudio.org> wrote:
>> 
>> Simon (last name forgotten right now) added a variation on topological
>> sort years ago, specifically to deal with some issues. His comment
>> was:
>> 
>> /* How the sort works:
>> ...
> 
> ???

Fons,

I wrote that comment whilst implementing what the list agreed was the best way to deal with cycles in the graph. Its not “how the sort works” so much as “how we fudged the sort to deal with cycles”.

I think the underlying sort mechanism was the same before I did that: Call the sort function using the transitive closure of the who-feeds-who relation as the comparison. (Except now its the who-feeds-who-after-inverting-feedback-connections relation).

That was always *supposed* to achieve a topological sort (before and after the feedback mods) but you’re saying it doesn’t: The merge sort behind jack_slist_sort can outright fail with a partial ordering because, if A isn’t connected to B and B isn’t connected to C, it may never even ask about the relationship between A and C.

In which case yes, its a bug and it goes way back.

Simon Jenkins
Bristol, UK


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linuxaudio.org/pipermail/linux-audio-dev/attachments/20150811/6db070a7/attachment.html>