GPUs stuck on Update_Core

[gunnarre]

Resuming a partial download is built into the HTTP protocol which all modern webservers use, but finding out which part of a file is corrupt and needs to re-download would need using a different protocol for the download, like the Rsync protocol or the BitTorrent protocol. What could be done is to download a checksum before the file is downloaded and then test the downloaded file against the checksum. That still doesn't tell you which part of the file is corrupt, only whether or not it is corrupted. That might already be happening inside the client, but since the download never finishes or raises an alarm to the user, nor retries, it never gets solved.
Putting rsync into the client is a bit of an overkill solution, but an error message would be nice.

[[SG]Felix]

... or use an alternative download such a a mobile hotspot to bypass router issues?

Thats the way i got it downloaded, but this can't be the way to go. such a bug will stop potential new users to run fah, because "it doesnt work".

[bruce]

Fixing this limitation would be a really nice improvement to FAH but I don't see anybody who is going to re-write that segment of code. FAHCore_22 made some pretty significant package size increases when CUDA was added (along with other code enhancements. Creating an updating a list of checksums would be a PITA because they keep coming up with scientific enhancements so it continues to grow. After a failure, actually using those checksums after the unpack and then fixing the corrupt component would be a challenge for the novice ... and the unpack would have to work.

Adding the error message might get done if you put an enhancement request in to github but I have my doubts about more grandiose improvements is the download logic.