Instruments 10 Signpost Support - Still brittle?

The other option I've thought of is that my app is highly multithreaded. I've started to move signpost stuff onto the main thread (ugh) just in case this is the issue, but it doesn't seem to be making things better, so I'm assuming that might not be it.

Take a look in the "List: Intervals" detail. If the long "n/a" intervals are coming from subsystems with the "com.apple" prefix, then they're coming from the system libraries, and you can safely ignore them.

Thanks! Unfortunetly they're tracing back to me. In list view I can tell that my start message is there, so they aren't coming from nowhere. But it seems like the end is not being captured. I'm setting up a signpost id per worker object/thead, and one of the functions being affected is a simple enough function that only has one exit point. Using the C API in case that matters.

In the betas, signposts in Instruments are spotty after the device has been rebooted. It takes about 9h of up time for an iOS device (depends on the clock resolution) and about 18m and 20sec on Mac before they clear up. There was a bug where garbage was being left in the high order bits of the timestamp and it took about that long before those timestamps were big enough to clear the garbage. Anyway, it can lead to "it works the next morning morning", if that was indeed the case.

Okay, reproduced and found the problem. The os_signpost instrument was not properly pairing using the name, it was just using the ID. So if you had different signposts names but using the same ID, then it could get confused. 41557858