Possible World Record: Most commits to main branch on GitHub
I decided to see how many commits GitHub (and git) could take before acting kind of wonky. At ~19 million commits (and counting) to master: it's wonky.
The GitHub API has periodic issues merging/creating PRs. (I use PRs since that is more reliable than keeping a local master up to date via pulling at this point).
GitHub reports... infinity commits. Doing a full clone from GitHub Actions is taking around 2 hours.
Using:
git pull origin master
to update my local master hangs for a while before even printing anything.
If anyone has seen a public GitHub repo with more commits to its main branch, let me know.
If anyone wants to see how slow wonky git acts at this scale, feel free to clone.
Edit: Update to this post is: https://sh.itjust.works/post/672069
https://github.com/csm10495/commit-mentOpen linkView original on sh.itjust.works
Was this you? 🫤
Some people just want to see the world burn. Haha.
It feels like it's radiating chaotic energy for the sake of learning and I'm a fan
I love it.
These are the types of people to find something "Huh, that's a weird reaction"
So what's next?
Part of me wants to try to look into the git cli to try to see why some of the hangs happen. I don't think any sort of hang with no output (at least with -v) should happen normally.
Who knows if I'll get to that point.
Otherwise I'll let it just keep committing.
This might be good to cross post to [email protected] ... 😱
I wonder if the size of the commits also plays a role.
The latests commits are just an index being incremented, it'd be interesting to update the script to replace and commit lorem impsum text (maybe two or three paragraphs would be a good base)
I'm going to read about GitHub being down (with a link to this repo) on Monday, aren't I?
Infinity commits is interesting. Do you think it's caused by GitHub trying to count commits and timing out?
At this point I think so. Even the API (v3 or graphql) can't get me a commit count anymore. It usually gives something like a 503 with a reference id to give support.
Needless to say, I don't think its worth bugging support about it :P
It might be worth it. Not sure their systems have been stressed in this way before and they might want some engineers to look at it. It could lead to some more efficencies in smaller repos. Or it could lead to nothing. Might be worth contacting them to see. Even if you ultimately don't care about this repo.
Fair enough. I created a thread on the support forum. Since I don't have a paid account, I can't contact support directly.