| 27 Oct 2025 |
 emily | I'm having to fix it because of dropping Transmission 3 breaking it. or well, "having to", I could just mark it broken. | 17:17:42 |
| emily | I was surprised Hydra is running it. | 17:17:48 |
 Vladimír Čunát | Ah, so we end up with thousands copies of Sintel video in the S3. | 17:18:00 |
| emily | but I suppose it's not really any different from fetchurls, mostly. I expect it can't actually receive connections through the firewall. | 17:18:06 |
| emily | probably a tiny contributor to cache growth compared to ISOs etc. in fairness 🙃 | 17:18:30 |
 hexa | we didn't build this on hydra before, it if it was marked unfree | 17:19:04 |
| hexa | we started building this once it was changed to sintel | 17:19:17 |
| hexa | so more likely just hundreds of copies 😄 | 17:19:31 |
| Vladimír Čunát | So far. | 17:19:41 |
| emily | another consideration is that with Transmission 4 it starts logging things like
source-salted-fym81xbq7b5j> [2025-10-27 17:23:14.340] Sintel: [66.56.81.6]:32243 [qBittorrent 5.1.0]: got unrequested piece 364:49152->16384
| 17:28:37 |
| emily | I am not sure what the GDPR implications of temporarily recording the IP addresses of torrent swarm peers in public Hydra logs are | 17:28:57 |
| emily | I'd guess that publicly announcing your IP to serve data to anyone who asks via a tracker means that your IP is no longer personal information | 17:29:22 |
| emily | but I am not a lawyer… | 17:29:29 |
| emily | anyway, if there is a strong feeling about not running these on Hydra I can adjust it in my drop PR | 17:37:02 |
| emily | otherwise I'll leave the status quo | 17:37:05 |
| emily | fwiw, I assume that a postFetch could turn these into zero-byte FODs | 17:37:45 |
| emily | although of course it would still be downloading Sintel like 6 times per platform on every staging-next, just not pushing it to cache | 17:38:04 |
| Vladimír Čunát | With FOD you have a problem to force rebuilds. | 17:51:56 |
| Vladimír Čunát | Normally FODs don't get rebuilt. | 17:52:05 |
| Vladimír Čunát | i.e. it wouldn't be testing anything beyond a single attempt. Not even on stdenv rebuilds, basically never. | 17:52:45 |
