| 26 Oct 2021 |
 Vladimír Čunát | On a quick glance it didn't seem very flaky in the past, unless people were restarting it often: https://hydra.nixos.org/job/nixos/trunk-combined/nixos.tests.hibernate.x86_64-linux/all?page=1
(I didn't bother looking into many of the build steps for repeats) | 10:36:13 |
| Vladimír Čunát | For now I can only hope that a few retries can work around such issues 🤷 | 10:46:05 |
| Vladimír Čunát | * For now I can only hope that a few retries can work around such issues 🤷
EDIT: anyway, this is a wrong channel. | 10:47:41 |
| Vladimír Čunát | * For now I can only hope that a few retries can work around such issues 🤷
EDIT: anyway, this is a wrong channel, I'm sorry. | 10:47:57 |
 das_j | In reply to @r-burns:matrix.org
Thank you! Gonna bisect on jx with this Did you have any luck with the bisect? | 20:32:11 |
 Ryan Burns | Yes it's going great :) Only 3 more commits to check | 20:33:11 |
| Ryan Burns | Needed some manual patch fudging so I wasn't able to run it fully automatically overnight | 20:33:38 |
| das_j | Oh right, I forgot about the patches :/ | 20:33:54 |
| das_j | * Oh right, I forgot about the patches :/ I removed one and added another one | 20:34:10 |
| Ryan Burns | But I'm close! It looks like a change to cp is the culprit | 20:34:34 |
| das_j | Dang it, that was the main reason I wanted the 9.0 release | 20:35:19 |
| das_j | yeah boy | 20:40:48 |
| das_j | it might be this: https://lists.gnu.org/archive/html/coreutils/2021-10/msg00021.html | 20:40:50 |
| das_j | And here's our go tools failing: https://github.com/golang/go/issues/48636 | 20:43:51 |
| Ryan Burns | Interesting that someone else is seeing issues with bootstrapped go. Seems like you need a really specific sequence of events for this issue to manifest | 20:44:21 |
| das_j | And somehow also something that happens on Darwin | 20:45:40 |
| Ryan Burns | Ok yep bisected to the same commit
cp: use SEEK_DATA/SEEK_HOLE if available | 21:07:23 |
| Ryan Burns | Also note this coreutils commit after the 9.0 release: https://git.savannah.gnu.org/cgit/coreutils.git/commit/?id=61c81ffaacb0194dec31297bc1aa51be72315858 | 21:08:18 |
| Ryan Burns | Run the SEEK_DATA check on the
1TiB empty file to exclude both FreeBSD 9.1 which takes 35s,
and ZFS which requires a delay of about 5s between file creation
and use of SEEK_DATA to correctly determine it's empty (return ENXIO).
| 21:09:08 |
| Ryan Burns | If you have to modify the unit test, isn't that a sign that something broke?? | 21:09:43 |
| das_j | In reply to @r-burns:matrix.org
If you have to modify the unit test, isn't that a sign that something broke?? Nono of course not :D | 21:09:55 |
| das_j | we could manually unset SEEK_HOLE, maybe that would help? | 21:11:37 |
| das_j | For reference, this is the commit: https://github.com/coreutils/coreutils/commit/a6eaee501f6ec0c152abe88640203a64c390993e | 21:12:12 |
| das_j | Redacted or Malformed Event | 21:20:39 |
| das_j | I'll open a bug report on coreutils-bugs | 21:21:21 |
| das_j | 
Download image.png | 21:46:13 |
| das_j | Is there anything anyone wants to add/change here? | 21:46:19 |
| Ryan Burns | Are we sure that peertube is the same problem? I wouldn't want to confuse them if it is not actually related | 21:47:07 |
| Vladimír Čunát | Is there evidence that peertube is broken by the same coreutils commit? | 21:47:15 |
| das_j | Good point | 21:47:23 |
