I can't see the WU at the moment as the Database is messed up, but I'll look at it later.

I am noticing something different in my ARP behavior. It will horrify you to know that I was getting a high percentage of invalids (30-50% I wasn't scientific in my measurements, just looking at my results list now and then and counting). The number of invalids has dropped to an occasional one in a list of 30 or so valids. So something has changed. My best guess is that they fine tuned my wingman choice. But I admit to this being a wild guess.

Unixchick

I suspect it has stuck after getting an error. Do you know its name?

Mike

Unixchick -- I'd just checked your latest rogue unit (now the server is back up) but I see you've already spotted that it ended up as a lost cause...

It would be nice to know what causes a situation where 6 notionally equivalent systems all produce results that the validator thinks are different. [At risk of stating the obvious] in the case of Apple systems there may be some very subtle issues regarding "Apple Silicon + Rosetta" versus "Intel" as well as possible differences because of different O/S versions (not that the latter should've applied in this case...)

And yes, it's frustrating when one puts in the time and doesn't get the credit. As an aside, I wonder how many of the users who are/were getting "Too Late" in the 2025-04-25 MAM1 Beta batches are feeling the same way...

Cheers - Al.

Unixchick

In this case, you were not too late in the usual sense. It means that either you had not replied by the time they decided to chop the unit or they dhopped it after you replied and no-one had valid results.

Mike

I'm not 100% sure whether "chop the unit" was intended to indicate human intervention here, but in case that was the intention here's some information...

I've just had a look at the validator control code of a standard BOINC server, and it will automatically mark the WU as in error if there are more "viable" results than a specified maximum number of "success" results but there still isn't a canonical result. The usual number is (I believe) 6, so it is quite likely that in both those odd Darwin WU cases it was an automated fail.

The use of "Too Late" as the tag in this situation has been criticised as misleading[*1], for obvious reasons, though I can't think of anything much better to use! I suspect that it's a case of keeping the code simple, as the behaviour is the same if there's one valid result returned to a WU but all the others are marked as Errors for one reason or another.

Cheers - Al.

*1 I last saw that a long time ago, and I can't remember whether it was here (perhaps when we had all those ARP1 tasks failing in IBM days?) or elsewhere, so "Sorry, no citation"...

[Edited the paragraph on the use of "Too Late" as a tag...]