Is there a known issue with OS_CallAfter ?

Wow. That took a long time to build from scratch. ;-)

Okay, running now.

Okay, a quick update…

Appears to work fine now. The system has been up for two days now with no signs of ticker chain corruption.

As for my module, I have implemented two ways of performing the twice-a-second check. If the OS build date parses as 12th August or later, it’ll CallEvery 50cs as it has in the past.
If the build date is prior¹, it will hang off TickerV and increment a counter for 49 ticks, and on the 50th tick reset the counter and do the necessary.
Hopefully that will provide reliable operation in either use case.

¹ Also if build date is not available; but at this time the software only works on RISC OS 5.

Hopefully the final follow-up. ;-) Only just realised, but NetTime may be exhonerated for its alleged sins. My Pi has been keeping good time, currently it is 0.050434 seconds fast, and appears to have been checking every half hour.

I am beginning to suspect that the problems exhibited by NetTime might not have been down to the CallEvery failing. That said, I wonder why its effects were not more widely noticed? EtherUSB is called back every 2cs, USBDriver every 3001cs, LanManFS every 4500cs, DWCDriver every 11cs, and the Wimp every 100cs. Surely if that lot failed, there should have been more fall-out than “NetTime sucks and my server won’t respond”? ;-)

I am also wondering, given the amount of USB/Networking stuff implicated in the above list, if this might not have been in part to blame for some of the problems David Feugey was seeing with his server? I was running a clone of his site on my Pi with WebJames and it was being spidered by the Russians, but WebJames kept going. Luck, perhaps? Unfortunately for this, David has switched to a non-RISC OS server, as it would be interesting to see if the fix cures that issue…

PS: David, if you’re reading, “Hi! Hope you had a nice holiday!”. (France grinds to a halt for the first few weeks of August; and it is fun when the Juilliards (July holiday people) come home at the same time as the Aoutiens (August holiday people) leave. The weekend overlap is endless traffic jams, exploding radiators, and the usual sorts of fun.
Me? I sat in the garden and read stuff for most of my holiday. I don’t miss the beach, anything resort-like is too far away, and frankly, overpriced. So I enjoyed the weather with mom+cat.

I think most of the ticker events listed are just there to kick things back into life if the programs get stuck with data pending or the program misses an event. I suspect generally they don’t do anything

“Hi! Hope you had a nice holiday!”

Holly-what?

The server issues were really strange. Now that I switched to a traditional server, I hope to provide new services for RISC OS users. 2015-2016 should be challenging and fun.

… And perhaps I’ll have some holidays this time :)

My Pi, running my server and WebJames has been running reliably since I last rebooted it.

So thank you Jeffrey for tracking down the ticker chain problem.

I just hope you are too busy with more important things than to read this until the day after Christmas. ;-)