On 03/14/2013 03:58 PM, Tanu Kaskinen wrote: > On Thu, 2013-03-14 at 14:58 +0100, David Henningsson wrote: >> On 02/20/2013 07:24 PM, Tanu Kaskinen wrote: >>> I was writing function pa_device_port_update_proplist(), and I wanted >>> it to send change notifications only if the proplist actually changes. >>> pa_proplist_update() doesn't provide any indication about whether the >>> proplist changed or not, so some kind of a solution was needed. >>> >>> The simple solution would be to create a copy of the port proplist >>> before calling pa_proplist_update() and check if the copy equals the >>> port proplist after calling pa_proplist_update(). That felt overly >>> wasteful, however: it would mean copying the whole property list and >>> comparing every property in it whenever someone changes even just one >>> property. >>> >>> So, I invented a more complex solution: a pa_proplist_update_info >>> object that holds a description of per-property operations to be >>> applied to a property list. pa_proplist_apply_update_info() iterates >>> through the operations and applies them one by one, keeping track of >>> whether the operations cause actual changes. >> >> I guess that's one way to solve it. I would probably have gone for the >> slightly simpler solution of just keeping a flag inside the proplist. >> The proplist object itself will set the flag whenever a "real" change >> occurs, and it can be manually reset by just calling, say >> pa_proplist_reset_change_flag() or so. > > That sounds pretty sensible. Do you think I should redo the patches? I'd > prefer not to do that, but that's just because I'm lazy. You mean, it's just because you want to get more time for fixing other PulseAudio issues ;-) Anyway, I'm pragmatic. As long a patch improves the current condition and does not cause regressions, APIs we have to commit to in the future, or other implications, I don't mind it going in. It can be rewritten, simplified, or whatever, by someone else later if they feel like it. >>> --- >>> src/map-file | 4 + >> >> I don't think we need to add this to the external API unless somebody >> complains about missing that feature. > > Fair enough, I don't mind hiding this from clients. Ok. -- David Henningsson, Canonical Ltd. https://launchpad.net/~diwic
