> For the sake of this discussion, we only need to do three things:
[...snip...] removing lots of *symptom* investigation [...snip...]
You never get to the root cause here, which is removal of connections without
undo recording in the first place.
> which we can summarize as the input/outputs of **Bus-1** are disconnected in
> `bse_container_uncross_undoable` [1], due to this `bse_bus_disconnect` [2]
> fails and finally the undo step is not recorded in `BusImpl::disconnect_bus`
> [3]. The patch that I've proposed fixes this by ignoring the error code
> Error::SOURCE_NO_SUCH_CONNECTION in [3]. Note that this doesn't irgnore all
> possible errors, just specialcase the case that there was no connection.
>
> > * Why should an undo step be recorded for restoring a connection that
> > doesn't exist in the first place?
>
> Because the connection did exist, but was removed in [1].
No, you're just looking at one possible way and one possible time to call a
function, a connection did not *necessarily* exist previously, and only if it
did would it be valid to record an undo step.
> > * Also, with your code, calling disconnect_bus() repeatedly will record
> > more and more undo steps, for a connection that's not existing and could
> > have been restored at most only once anyway.
>
> Honestly, I don't know the answer to this. Maybe the fix should go somewhere
> else, or if we want to keep this, we could add state to the bus that
> indicates whether the bus is connected or not (independantly of whether or
> not the channels are connected).
Thanks for tracing this down, a clear problem understanding is part of what's
needed for a fix. As for the actual patch, we must make sure to only record
undo steps when appropriate though like described earlier.
--
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/tim-janik/beast/pull/82#issuecomment-434403855
_______________________________________________
beast mailing list
[email protected]
https://mail.gnome.org/mailman/listinfo/beast
