OK....I won't be surprised if it fails then.
I know I was told that the OP cpu app would not work on XP.

I am surprised and a bit worried when I hear that BOINC is apparently not protected against running modified binaries.

The only change is to the header of the executable that says what version it needs as a minimum.
As it is compiled on a later version that is what is demanded but it does work on earlier versions.

Paul.

I do not agree here. With the original binaries being signed, BOINC can detect that the code it's going to execute wasn't sent by the project. No more and no less. There's no possibility to tell how and by whom it was modified. I would expect that BOINC simply refuses to proceed in that situation. But apparently the signature is not verified. Maybe once on initial download but not after that.


In this case you'll either have no signature at all or the binary is self-signed with a valid signature. In the other case you have a signature which is clearly invalid but not checked.

IMO it's a surprising and arguably unreasonable feature, very close to a bug.

There is some server side protection with the use of replication=2 to verify result. Any modified or wrong data becomes invalid. Modified binary or data can modify the results which can have invalid result. Faulty RAM can also cause invalid data, I had to replace DDR3 RAM on my old AMD FX 4100 computer to fix my random invalid results.

Edit: To add, I guess modification which doesn't change to results data is probably ok. BOINC is open source so anything is possible at the client side.