| 11 Aug 2025 |
 emily | just wanted to check if I should be surprised or not | 13:53:41 |
| emily | seems it's getting completely stuck on the first tarball it fetches. I have no idea why. downloads instantly with curl inside the VM. might be a Lix passta thing. it fetches a Git repo successfully before that. hmm. I could swear it was just being really slow but not totally hanging yesterday | 15:06:43 |
| emily | it seems like session.get(…) is never returning | 15:06:49 |
 Ben Sparks | I'd observed this before as well for a few days back in May / June I think. The vendoring process would slow to a crawl | 15:14:21 |
| emily | DEBUG:urllib3.connectionpool:Starting new HTTPS connection (1): crates.io:443
and then nothing. fun
| 15:14:24 |
| emily | is that the same building-Linux-package-in-remote-builder-VM-on-macOS config? | 15:14:50 |
| Ben Sparks | I think it was; currently trying to find old messages of mine where I had asked in one nix-related channel if crates.io rate-limited | 15:16:18 |
| emily | May / June makes passta an unlikely culprit too (unless it was late June and you use Lix) | 15:17:02 |
| Ben Sparks | $ nix --version
nix (Lix, like Nix) 2.91.3
| 15:17:34 |
| emily | on the builder VM too? :) | 15:17:54 |
| emily | would depend on exact timing then I suppose | 15:18:12 |
| Ben Sparks | I wouldn't rely on this timespan btw until I find further proof of when the issues occurred.
I also never found out why this happened; it just "went away" at some point | 15:19:20 |
 Toma | emily: could you try with the changes in this PR: https://github.com/NixOS/nixpkgs/pull/400865
(while testing if you only build cargoDeps.vendorStaging you won't have many rebuilds)
| 15:33:57 |
| Toma | Currently the timeout logic in fetchCargoVendor is pretty bad (not even sure if it's working properly) and this tries to fix it, while also doing some general improvements (at least it tries to) | 15:34:55 |
| Toma | * Currently the timeout logic in fetchCargoVendor is pretty bad (not even sure if it's working properly) and this tries to fix it, while also doing some general improvements | 15:35:05 |
| emily | unfortunately I rebooted the VM and it fixed it 😅 | 15:35:45 |
| emily | but I will try this next time | 15:35:47 |
 K900 | Any reason we're not just doing async at this point | 15:35:52 |
| K900 | Feels like it'll make things significantly less painful | 15:35:59 |
| emily | oh wait | 15:36:04 |
| emily | it worked after applying your patch and I assumed that was because I rebooted | 15:36:16 |
| emily | but it fails again if I revert your patch | 15:36:20 |
| emily | well, hangs again | 15:36:25 |
| emily | uh | 15:36:42 |
| emily | ok, but then I applied your patch again and it's back to failing | 15:36:47 |
| emily | so I guess it just worked once and then I ^C'd half-way through and tried another build and it re-broke :) | 15:36:56 |
| Toma | hell yeah i love consistency | 15:37:06 |
| emily | I'm doubting this is a bug in your code rather than a VM/passta issue | 15:37:10 |
| emily | wait no | 15:37:24 |
| Toma | in any case, the timeout logic is better with my PR | 15:37:31 |
