22 comments

[ 4.2 ms ] story [ 61.9 ms ] thread
Hey recruiters! This is a somewhat hard working open source software engineer. Did 570k commits in 1 day.

Must be so clever, top class! /s

Let's call them a 10k times engineer.
Ok, I gotta admit, I have no idea what is going on here?

Just a script that commits nothing constantly? The bookmarks are ... non related? I don't understand at all!

Yes, clearly the guy used a script to create as many empty commits as he could. It's now supposedly the repo with the most commits, but it's just a hack. I guess the bookmarks are just things the guy likes and hopes people will click on after reaching the "most commits" repo.
Oh ok, fair enough I guess.

Thought I'd missed some larger joke/meme/issue thing!

I was wondering if there was a way to find the most commits in a repo, and a google showed this as first hit: https://github.com/virejdasani/Commited

I think he/she's just playing with git datastructure, the logs contains only commit messages. That's what we get after the pandemic :D
So this guy is trying hard to break Github, I guess...?
Using a simple shell for-loop, I could create 10000 commits in 25 seconds (= 400 per second) with the code below. The 4.2 million commits should take around 3 hours to create at this pace.

Looking at the most recent 10000 commits in the repository, it doesn't seem like the creator ever got more than 20 commits per second, which means it would take on the order of 60 hours for the creator to make this many commits.

    $ git clone https://github.com/IThinkImOKAY/IThinkImOKAY && cd IThinkImOKAY
    $ t=`git write-tree`
    $ h=`git rev-parse @`
    $ date; for i in `seq 1 10000`; do h=`echo lol|git commit-tree $t -p $h`; done; date
    Wed Jun  9 05:20:11 PM CEST 2021
    Wed Jun  9 05:20:36 PM CEST 2021
Here's a parallel implementation that gets me 950 commits per second:

    manycommits() {
        h=@; t=`git rev-parse @^{tree}`
        for i in `seq 2 "$1"`; do
            h=`echo "$*" | git commit-tree $t -p $h`
        done
        echo $h
    }
    twice() { "$@" a & "$@" b ; wait; }
    twice twice twice twice manycommits 10000 | while read h; do git merge --no-edit $h; done
And here I was thinking matching compensation to LOCs was silly. Clearly more employers need to pay based on commits per second!
Goodhart's Law at it's finest! I'm betting it started as a fun way to get a nice ("S+") score in the new github badges or whatever they are called.
At least the author is committed.
I've always wondered much these types of things degrade or stress out the github backend services and what they do to prevent abuse.
They are ruining / skewing the stats of that plugin in the readme for others, it uses an algo to determine profile rankings based on everyone elses stats.