Alright, so: I have deployed Anubis in front of the bug tracker

Let me know if you have trouble getting to it

thanks, seems pretty fast now :)

hooray

I have attempted to make sure git-pbchk will be allowed thorugh

*through

jclulow (LIBERA-IRC): anubs to the rescue. Thanks!

s/anubs/anubis/

may the jackal-eared girl preserve us

[illumos-gate] 17725 pcieadm could decode Virtio vendor-specific capabilities -- Andy Fiddaman <illumos⊙fn>

:o

jclulow: seems to be working. I'm not seeing the anime interstitials that appear on other installations of anubis (and hope it stays that way in our installation...)

I think that will only show up if we have to ramp up the challenge complexity

Or obviously if you somehow trip the bot detection

fenix: illumos#17725

FEATURE 17725: pcieadm could decode Virtio vendor-specific capabilities (Closed)

↳ illumos.org/issues/17725 | code.illumos.org/c/illumos-gate/+/4445

[illumos-gate] 17733 crypto-tests failures when run as non-root -- Gordon Ross <gwr⊙rc>

jclulow: heroic

Eeesh good catch @andyf on fenix illumos#17734

BUG 17734: ZFS fsync can trigger ZIL transaction reordering and data corruption (In Progress)

↳ illumos.org/issues/17734

BTW for some values of Sun login IDs where it's `flNNNNN` (e.g. were I forced to use that back in the day it'd have been `dm30063`) I have data.

for all values in commit messages I have data too

would you like it dan?

That one's taken a village over the past weeks. I only got involved today after it became clear it was likely to be something in ZFS, and colleagues presented a tidy reproducer.

I've been hesitant for other people for privacy reasons, but I know you...

Ack.

/facepalm src.illumos.org/source/xref/illumos…/common/ibpart_rcm.c?r=aab83bb8#314

doh!

based on the comment, they were trying for `ASSERT(node != NULL)`

Maybe ^K'd the assert before committing by accident?

andyf: I'm pretty sure I've seen the underlying failure mode (file corruption due to block write being reordered before a truncate) in a different contexts. NFS, maybe?

Could be. It's always going to be a bad idea. The ZFS ZIL has a handful of places this could happen now I'm looking at it, because it can switch between sync and async records. It's just probably not as bad for most things unless a truncation is involved.

and of course you do need a ZIL replay..