On 2016-02-25 17:18, Tanu Kaskinen wrote: > Hi, > > Some time ago I submitted a patch[1] for fixing the HDMI routing > regression[2]. The patch introduces two new variables for cards: > preferred_input_port and preferred_output_port. The variables are > internal to module-switch-on-port-available. It turned out that the > patch doesn't fix the bug properly, because the variables are not saved > on disk, so the user preferences are forgotten on reboot. > > Now my question is what approach is preferred for implementing the > variable persistence. I'd like to keep the logic of setting the > variables in module-switch-on-port-available, but I'd prefer not to > create a new file for storing the variables on disk. So, I propose that > module-card-restore will take care of saving and restoring the > variables. > > How should module-switch-on-port-available and module-card-restore > communicate with each other? My preference would be to add the > variables to pa_card, and add functions > pa_card_set_preferred_input_port() and > pa_card_set_preferred_output_port(). module-switch-on-port-available > would call those when it wants to change the values, and module-card- > restore would call them when restoring the values from disk. There > would also be new hooks PA_CARD_PREFERRED_INPUT_PORT_CHANGED and the > same for output, which module-card-restore would use to update the on- > disk database. > > Another alternative would be to use the card proplist for storing the > variables. > > Opinions? Hmm, is there anything in the above that differs from the strategy used for the recently added pa_device_port->preferred_profile? Adding variables to pa_card seems reasonable to me. Perhaps having the same hook for both input and output (and perhaps adding a parameter telling which one changed) would save a few lines of code here and there? But that's mostly a detail. > > [1] https://patchwork.freedesktop.org/patch/74014/ > [2] https://bugs.freedesktop.org/show_bug.cgi?id=93946 > > -- > Tanu >
