| 8 Aug 2022 |
 Winter (she/her) | it looks like it's only ever applied to arm | 05:48:38 |
| Winter (she/her) | (see https://github.com/MagicStack/uvloop/issues/412) | 05:48:50 |
 fabianhjr | eventlet failing on hydra: https://hydra.nixos.org/build/186525852
uvloop no attempt on hydra: https://hydra.nixos.org/build/186601863
| 05:49:23 |
| fabianhjr | Unsure about what propagated the issue | 05:49:44 |
| fabianhjr | eventlet has a new test failure but was a single case so disabled it and didn't investigate further | 05:50:10 |
| fabianhjr | uvloop i identified the hang while building on my machine and upon reading the comments and noticing where it got stuck I just went and tested if I had the same issue and I did have it | 05:50:52 |
| Winter (she/her) | on x86_64? | 05:51:07 |
| fabianhjr | Yeah | 05:51:13 |
| fabianhjr | zen ver. 2 | 05:51:18 |
| fabianhjr | AMD 3800X | 05:51:25 |
| fabianhjr | Can be also kernel / ipc related | 05:51:55 |
| fabianhjr | Reading the upstream issue it seems tricky | 05:57:12 |
| fabianhjr |
Our builders wait for 120 minute, but the test suite is really stuck there. I cannot reproduce the issue on a Raspberry Pi 4 (Linux 5.11.14, Python 3.8.9), but it is perfectly reproducible on an APM X-Gene 2 (Linux 4.19.188, Python 3.8.9).
| 05:57:15 |
| fabianhjr | Maybe some specific hardware + software that creates a really odd edge case. | 05:57:45 |
| fabianhjr |
No luck - all tests passed. And the stack info is not really telling a lot. I think it should be safe to disable in your case, as test_remote_shutdown_receives_trailing_data (tests.test_tcp.Test_AIO_TCPSSL) is particularly testing asyncio not uvloop.
| 05:59:19 |
| fabianhjr | Alpine also disabled it: https://github.com/alpinelinux/aports/commit/1cbbc2effee69cb63fe596de5ad4f7c8c7191448 | 06:00:06 |
| Winter (she/her) | Alright, LGTM then, as long as we're not the only ones 😅 | 06:07:40 |
| Winter (she/her) | In reply to @fabianhjr:matrix.org
eventlet has a new test failure but was a single case so disabled it and didn't investigate further i don't see why it would suddenly be failing, no recent changes :/ broke pipe doesn't give us much to work with | 06:10:55 |
| Winter (she/her) | s/broke/broken | 06:11:09 |
 Vladimír Čunát | Note... for myself? This needs decision before the next staging-next iteration: https://github.com/NixOS/nixpkgs/pull/185613 | 10:02:21 |
 Sandro 🐧 | the decision must not be the same for staging and stable | 12:00:48 |
| Winter (she/her) | In reply to @vcunat:matrix.org
Note... for myself? This needs decision before the next staging-next iteration: https://github.com/NixOS/nixpkgs/pull/185613 I merged it, it's a zlib issue. | 15:19:49 |
| Winter (she/her) | Not really sure what you want to do with that info though, I don't remember if the original commit is staging-next | 15:20:31 |
| Winter (she/her) | it isn't -- what do you mean by that, then? | 15:21:47 |
| Vladimír Čunát | It is not. | 15:21:47 |
| Winter (she/her) | oh i see what you mean | 15:22:06 |
| Winter (she/her) | the patch to zlib to fix this seems like it would be pretty simple but at the same time i'm only looking at an isolated piece of the code so who knows if that would cause other things to blow up | 15:22:50 |
| Vladimír Čunát | I'd hope that something is proposed upstream very soon. (not necessarily by madler) | 15:25:28 |
| Winter (she/her) | https://github.com/madler/zlib/pull/688 this makes all the curl tests pass, though who knows what i've missed -- hopefully it gets merged or madler makes a proper fix | 16:56:29 |
| Sandro 🐧 | you should add s description to that | 17:04:42 |
