Thinking this is all beyond my tech abilities and will just use Live365 for my station.
Certainly your choice, but bear in mind that this is a beta you're participating in so any issues you're experiencing are either 1) the result of a damaged installation caused by a past (now-fixed) bug, and just requires a reinstallation, 2) a bug in the current version you're using that hasn't been fixed yet, or 3) expected behavior that's just not well-documented because the formal documentation is not yet released.
So you can't really take personal responsibility for the issues you encounter in v3.
Encountering bugs is, of course, the purpose of a beta -- we do it now so that these issues don't show up in production.
As for this "Unknown" track title issue, I don't recall if you were the first to report it or just one of the many who did, but this ending up being a critical issue and the efforts (and/or suffering
) of yourself and everyone else who submitted info about it were greatly appreciated here.
It was VERY tricky to track down... I spent many 10-hour days this past week trying to figure it out. In the end, it ended up being a bloody PHP bug of some sort. For those interested in the technical details: If ices made an HTTP request to Centova Cast to fetch the next song's info, and that happened to occur while another HTTP request was being made to the Centova Cast web interface, PHP would in some cases close its FastCGI connection to nginx with nothing but an "HTTP/1.1 200 OK". Both script instances would run to completion without error, but one would generate no output beyond its headers, and nginx would record a "connection reset by peer" while reading from the upstream FastCGI server. Nothing in the PHP error logs, and no combination of output buffering settings (including disabling it completely) had any impact. This caused ices to fail over to its static fallback playlist, and for some reason (which I haven't determined yet) it used the blank artist name/song title from the previous, failed request.
The biggest problem was that it didn't occur on our dev servers for the longest time, so I couldn't reproduce it. It wasn't until we upgraded our dev servers to the new PHP v5.4-based release that it started happening here.
Switching to php-fpm eliminated this issue entirely, so while I don't know exactly what triggered this issue, there's clearly something dodgy in the classic PHP FastCGI interface in recent PHP versions.
I'll be pushing out a new php-fpm-based build shortly, and it works very nicely.