Jeremiah Fieldhaven4d ago

So my systems recently updated to rsync 3.4.3, and as soon as that happened my backup system - which does incremental backups using multiple --compare-dest= arguments - started to fail on anything but a full backup.

Revert to 3.4.1 and it works.

So I go look at the source in GitHub to see what might have changed, because there doesn't seem to be anything relevant in the changelog.

Since 3.4.1, 36 commits by "tridge and claude"

Oh for fuck's sakes.

Show threadThe Seven Voyages Of Steve4d ago
@JeremiahFieldhaven Christ if it’s coming for rsync of all things software is clearly done
Show threadsosquqer
@sinbad @JeremiahFieldhaven
It's in `curl` also, see https://codeberg.org/small-hack/open-slopware#networking. Not that I fully agree with the sentiment of that list, but it is useful to keep track of such things
open-slopware

Free/Open Source Software choosing to use and/or support LLM usage/AI, as well as alternatives and tips to requesting better policies or forking.

Codeberg.org
May 29 at 8:24amWeb
Show threadltning4d ago
@sosquqer Maybe don't cite that list .. they're not being honest, and in the curl case it's worth reading or watching the authors own musings on this (he's on here). I don't agree with everything but it is at least an honest and informed take. Quite unlike this list.
Cc @sinbad @JeremiahFieldhaven
Show threadDr. Bruce Simpson4d ago
@sosquqer
@anyia
@sinbad @JeremiahFieldhaven The "open-slopware" list mischaracterizes FreeBSD as having a "permissive" policy. There is no official policy, yet, to my knowledge. Although LLM-driven bug discovery is tolerated, LLM code generation is informally not tolerated within src, but tolerated for ported software (but not ports metadata). That is how I've seen things playing out so far. Documentation remains totally human written to the best of my knowledge.
Show threadTurboTrain3d ago
@bms48 @sosquqer @sinbad @JeremiahFieldhaven @anyia "not tolerated within src", well maybe i can't read it correctly but https://github.com/freebsd/freebsd-src/commit/e1cff854997884ed9b7251d409d9c9c7a025606d and https://github.com/freebsd/freebsd-src/commit/6495dafd58b94a44fc9bc966ef47d6bc6916f5b9
pass(4): Allowlist CCB func_codes to harden passthrough ioctls · freebsd/freebsd-src@e1cff85

The pass(4) driver's CAMIOCOMMAND and CAMIOQUEUE ioctls accept arbitrary CCBs from userland. This device requires root to open, and thus send these commands. Previously, the only func_code fil...

GitHub
Show threadDr. Bruce Simpson3d ago
@TurboTrain @sosquqer @sinbad @JeremiahFieldhaven @anyia OK this is news to me. I should point out that of these two changes, one is a one-liner, and the other, has no "tells" of so-called "slop". Did you look? cam is something I have coded for (ufdformat, a USB floppy formatter) but not since 2005. I agree with the position some take: "Co-Authored-By: is reserved for humans". It may be that these two changes crept in inadvertently. What I said is not official policy but still seems mostly true.
Show threadDr. Bruce Simpson3d ago
@TurboTrain @sosquqer @sinbad @JeremiahFieldhaven @anyia I can't say with certainty that there is no LLM-generated code in FreeBSD-src, merely that it is very likely there isn't. I know Warner is using Claude Code largely because his employer subsidizes it. I don't agree with this because it is known Anthropic prompt their models to lie that they are human, and that they have been exfiltrating customer data (Jonny Saunders' reverse engineering of the leak he published here.) I do not use it ever
Show threadax67613d ago

(removed larger list of accounts not related to this subthread)

@TurboTrain Well 😒 ...

"Oh, No! Not #OpenZFS?!"

"Et tu, OpenZFS?"

"Slop comes for every project."

@[email protected] With https://mastodon.social/@bms48/116662724883486066 in consideration, you are now moving goal posts, or accepting the increasing use bit by bit -- pick as suitable.

#whatHaveYouSloppedLately?🤨

Show threadDr. Bruce Simpson2d ago
@ax6761 @TurboTrain You are being intellectually dishonest here and arguing via false dichotomy. That which can be asserted without evidence can be refuted without evidence. But I already pointed out why the commits were not "slop". Read the diffs. I'm not responsible for what other FreeBSD developers do. Read what I actually wrote. If someone left Claude Code switched on, that reflects on them, not the Project. I don't use it thanks. If you have a problem with OpenZFS, cite commits.
Show threadDr. Bruce Simpson2d ago
@ax6761 @TurboTrain I mean, just the fact you raised the point at all lends weight to the central thrust of the thesis in https://burdentennis.com/ as the situation has us all screaming into our fists. As a scientist I prefer to refute empirically. But the very nature of LLMs, makes it hard, because they produce convincing output. Without empirical data (in this case, direct code inspection), it's burden-of-proof argument tennis. Principal-agent problem: you have to trust human code reviewers.
Burden Tennis: Refutation of Generative AI