Adjust UPnP detection to search localhost

[Raymond Wagner]

UPnP backend autodetection only works for systems given a routable address. This set of patches tries to connect the local status server on its default port (​http://127.0.0.1:6544/Myth/GetConnectionInfo) before proceeding with the normal msearch, allowing detection of backends not using UPnP.

This change is needed for MythBuntu? to move away from a shared mysql.txt to the new config.xml in individual user directories.

[superm1@…]

implementation in mythcontext

[rhpot1991@…]

patch for perl bindings

[superm1@…]

Just so these aren't lost (at least for the C++ implementation):

superm1> danielk22, could you comment whether ​http://svn.mythtv.org/trac/ticket/8094 looks too crackful at all during freeze - or if its at least the right direction to be headed? <danielk22> superm1: I think it's to late for that one. As for the patch itself my first concern would be that it might connect to a local master backend when you really want to the popup to choose which master to connect to. <superm1> danielk22, okay thanks. that's what i was figuring, it's just something wagnerrp me and rhpot1991 whipped together. <superm1> that's a situation i certainly didn't think about <superm1> albeit probably uncommon for anyone but a developer <danielk22> I'm not terribly fond of using something as heavyweight as UPnP for detecting backends, but it does mostly work. <superm1> do you think adding a --disable-local switch to the misc binaries using mythcontext like that would be sufficient then to prevent such a type of problem? <danielk22> superm1: Sure, but it's always good to try to address these things from the start, so hacks don't get applied on top of hacks later on. If it isn't tractable, then a switch would be fine. Maybe just a popup that ask "A local master backend was found, do you want to search for remote ones as well?" <superm1> that's a good point. i'll see about making such a popup instead then, sounds like a better solution to avoid hack upon hack being added

[beirdo]

Also related to the mysql.txt/config.xml merge. I'd be happy to work the UPnP side of this if we chose to do this, but meanwhile, I'm just keeping the related tickets together.

[stuartm]

These patches need updating and I think moved to more common code. The backend selection screen fails to find a local backend if the machine is firewalled because the SSDP search uses multicast, so building the local check into the upnp discovery code instead of mythcontext will be more useful.

[stuartm]

I will consider Mario's patch to try localhost first for the service discovery for 0.25, I'm not able to comment on the python and perl patches.

[stuartm]

Patch doesn't work for 0.25 and I think there's a better implementation possible if we wait for 0.26 anyway.

[Rune Petersen <rune@…>]

It is unclear what the plan is... But if you want to use unicast UPNP M-SEARCH you should know:

• It is not possible to receive unicast on port 1900 on windows (windows has a ssdp service, and SO_REUSEADDR with unicast is broken on windows)
• It is possible to use an alternative port for receiving unicast UPNP M-SEARCH (the SEARCHPORT.UPNP.ORG header)

If you simply need libmythupnp to also send M-SEARCH to localhost:1900 I don't mind making a patch for this.

[stuartm]

I can't remember what the "better implementation" I had in mind was. If someone wants to submit a working patch against master I'll do my very best to get it in for 0.28.

[Stuart Auchterlonie]

Moving to 0.29

[Stuart Auchterlonie]

Milestone renamed

[Peter Bennett]

After all of the changes to networking and other things, this is no longer a bug. UPNP detection is working fine in v29 even if your master backend has only 127.0.0.1 as listen address.