FFmpeg is not happy with AI generated patches sent by AMD(twitter.com)

39 pointsby itvision8 days ago5 comments

Alifatisk8 days ago
The dialog was something to read. Doesn’t sound like ai but rather something a developer has practiced for years and no one has challenged the dev on this, until now.
Also, there is discussions on what a commit message should contain, apparently the patch had user guidance in the commit message. Literally on how to install dependencies such as gcc using pacman, in the commit message.
https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/21595
- yesco8 days ago
  I get maintainers have their own issues to deal with, and respect that they are trying to keep the project clean. At work I have had many times where I spent more of my day reviewing MRs than actually writing code, and sometimes my cold blunt replies can unintentionally rub people the wrong way.
  Still, I feel like they were pretty rude to this guy for no real reason. I don't think I'd want to work with them.
dryarzeg8 days ago
const int EIGHT = 8 lol
I really doubt any AI (even some small local models) would actually generate something like this :)
- SomeUserName4328 days ago
  I've run into something akin to `const int EIGHT = 7`.
  Courtesy of TCS.
- bobx118 days ago
  Agreed. This reads more like a very junior dev reads static analysis warnings and extracts it into a constant to satisfy the ide. An LLM would at least give the constant a slightly abstracted name.
- hulitu7 days ago
  > const int EIGHT = 8 lol
  > “The primary purpose of the DATA statement is to give names to constants; instead of referring to pi as 3.141592653589793 at every appearance, the variable PI can be given that value with a DATA statement and used instead of the longer form of the constant. This also simplifies modifying the program, should the value of pi change.”
  > — Early FORTRAN manual for Xerox Computers, attributed to David H. Owens.
- Alifatisk8 days ago
  Yeah, I don’t think this is a case of ai slop. LLMs tend to be verbose with the comments but are fine with magic constants, at least from my experience.
high_na_euv8 days ago
I dont see problem with user manual in commit as long as reasoning for the commit is clearly written
- emsign2 days ago
  It's not about the commit itself but that due to it they discovered vibe coding by AMD. They object to the sending commits that have not been checked by a human dev.
- potbelly838 days ago
  yeah, they seemed to be nitpicking there, better to focus attention on what someone's actually trying to commit
- mort968 days ago
  Really? You think a git commit message in a C project is the right place to document how to install make and gcc on arch linux? That's not better suited as part of a readme file?
  - high_na_euv8 days ago
    It should be in readme but I wouldn't complain if commit with new feature also had it
    8 days ago
    undefined
    hulitu7 days ago
    > if commit with new feature also had it
    Maybe reading his OS documentation will be more appropiate. What's next ? How to install Arch ? Or Mesa ?
    emsign2 days ago
    But but but it could be helpful for the next coding agent in case how to use a compiler has already fallen out of its context window! Stay positive, man./s
- guilhas7 days ago
  -Where are the build instructions? -Please go check that commit message from 2011
estimator72928 days ago
The xitter post consists of only a screenshot, a link, and a drive-by comment.
Here's the actual thing we're talking about https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/21595#issuecomme...
guilhas7 days ago
AI code, junior dev or distracted person, either way there is no defense for AMD. As a billion dollar company you would at least have a simple QA before pushing slop into open source projects
FFmpeg maintainers have to keep pointing out incorrect parts. And conclude the entire is redundant
The PR interaction are clearly AI filler, at least on the second one
At this point open source projects should charge big companies to review their PRs