Shadowing Practice: Github are you joking? - Learn English Speaking with Video

GitHub just had its magnum opus of a bug.

Now, I've heard lore that people have dropped databases,

just some showstopper bugs, but I did not think it was possible for GitHub to cease being Git.

Now, we're gonna go through what actually happened,

but more so, I just have a lot of questions as to how this was even possible.

All right, so before we begin about GitHub ceasing to be Git,

let's first go over a bit of lore because it's important.

In 2005, the creator of Linux decided that he was going to take the Brendan Eich challenge,

which is to create a programming language in seven days,

but instead create version control in 10.

And thus, Git was born.

Upon seeing this beautiful distributed version control,

software developers did the only thing that made sense,

immediately create a single point of failure.

GitHub became so popular that other services decided to use the hub naming convention.

Of course, the most popular being Docker Hub,

where you can store all your dockers at.

Microsoft, seeing that GitHub attracted developers,

developers, developers, decided to purchase GitHub for $7.5 billion and proceeded to generationally fumble the bag.

Microsoft, you could have done like nothing.

Like if you would have just left GitHub,

it would have operated better than whatever has happened today.

But the impressive part is that Microsoft has discovered the fourth E in the triple E strategy of course,

which is embrace, extend, extinguish, and shitify.

Yesterday, we had a regression in the merge queue behavior where,

in some cases, squash or rebase commits were generated from the wrong base state,

making earlier changes appear reverted in branch history.

You had one job, just one singular job,

GitHub, which is of course to be git.

Now that's only about half the story because it really isn't kind of actually illustrating what happened.

Now, here's a better story.

On April 23rd, GitHub's merge queue started silently reverting code on customers' main branches.

Now, reverting is actually a Git operation.

It did not revert code.

That's very important to understand.

Not a handful of lines, in some cases, thousands.

Nothing looked wrong.

A PR with a plus 29 minus 34 diff got reviewed, approved, and queued.

What actually landed on main was a single commit with plus 245 minus 1137.

Thousands of lines of unrelated already shipped code were quietly removed.

Every merge that followed went in on top of that broken history.

That means that you would look at the PR.

Your PR would actually look correct.

And when you hit the merge button,

what actually went in was not what you hit merge on.

This is absolutely diabolical behavior.

I mean, to be lied to,

like the UI just telling you, nah, everything's good, man.

This is exactly what's going to happen.

And for you to have no idea is crazy.

I still have a lot of questions,

but this is roughly how it's being reported.

Now, we do not have the official GitHub kind of postmortem that explains exactly what went wrong.

This is just what is being purported by large companies.

Effectively, what happens is you have history that looks like this in Git,

and this would be your main branch.

So let's just say you have four commits on your main branch.

And for whatever reason, you branched off back here to make your new branch,

and then you've added two commits.

So the actual root cause of what would happen is that when you'd go to merge,

what would end up happening is they would run something called a merge queue,

which would take your two merges,

which if you just drew this in a straight line would be three circles,

because you'd have where you branch from and your two changes.

And it would squash these two into a singular change.

So they are just one single branch.

This is actually a whole new branch that's created internally in one of Microsoft's CI.

Let's just call this singular squashed merge A.

What should happen is that A should be merged to the tip of main.

But what actually ended up happening is

that they would take this history right here and overwrite whatever was on main with this right here.

So that means if this thing was called B,

that means what would be on main now would simply be B to A.

And that is it.

That means these three commits were just gone.

Now, this obviously caused teams to scramble all afternoon.

There are hundreds of tweets of people saying,

hey, I had to spend all afternoon

or multiple days trying to figure out how to detangle whatever has happened

and what is actually missing from all of their PRs.

GitHub, of course, has said there's only 2,804 pull requests that were messed up out of 4 million merged.

Then later on said there was only 800K PRs either way.

That's the difference between 0.07% and 0.3 plus percent.

It's a big difference.

So which one is it, GitHub?

And obviously even more confusing thing,

if you go to April 23rd, that's weird.

I don't, do you see like some major outage or a partial outage going up there?

No, they don't show anything went wrong on the 23rd.

Obviously the reason for that is that the only downtime was you

and your team trying to figure out how to undo what just happened.

But I also just have a lot of questions and maybe you guys can help me on this one.

So the first thing I have is that if this is B

and this is A and this is the temporary branch that we have created,

if history looked like this,

wouldn't pulling down cause you to have some sort of,

hey, the remote branch is like out of sync with your local branch,

meaning your two histories have diverged.

You wouldn't even be able to pull down main anymore.

100

Or the other option is that instead of just simply putting A right here,

101

it actually had revert commits,

102

meaning that it would have these three commits right here,

103

and then it would have another commit that was like,

104

hey, I'm reverting these three commits right here,

105

and then you would have something like A,

106

which would make this not too difficult to be able to go,

107

okay, well, here's the only revert we have,

108

unreverted, which is not like how anything sounds like.

109

But even more importantly, Git is still Git.

110

You cannot change history like this easily.

111

Does that mean internally is Microsoft like pushing with force, right?

112

Because the only way that you could push this right here,

113

this history onto main would be that you have to force push because you're changing history.

114

And if that's the case,

115

if that's the case that they're force pushing,

116

makes me kind of nervous.

117

Like what other issues are there?

118

What other things have been silently happening?

119

And of course this, there's no shot it was 2,804 total.

120

Please double check this stat.

121

We probably had 200 plus as a single customer.

122

So I'm actually just pretty curious.

123

Did you lose any commits?

124

Are you using the kind of the auto squash feature and thus you have things gone?

125

Also, this has opened up the most legendary excuse of all time for the next year.

126

When any company gets hacked and stuff gets leaked or something goes horribly wrong,

127

they can go, oh yeah,

128

well, actually we actually did implement perfect security.

129

We actually had this exact situation already covered,

130

but GitHub decided to silently drop our commit.

131

Brother, it's not our fault, it's GitHub's.

132

It's also funny because there's probably a bunch of vibe coders

133

that they just simply just kept on going on with their day.

134

Some weird thing happened and they're getting,

135

they're just like, oh, fix it,

136

oh, AI, just fix it.

137

And then when a bug appeared,

138

they're like, stupid AI, always reverting previous changes.

139

Hey, fix that one now again.

140

You know that happened.

141

Absolutely beautiful.

142

And the worst part is for once, it actually wasn't AI.

143

They're just gonna be blamed for part of this for a whole group of people.

144

Actually, I mean, to be fair,

145

it could have been AI.

146

It could, hey, it could have been.

147

There's some reports going around that it was a faulty feature flag.

148

Of course, Hacker News had its heyday and people were viciously discussing what happened to GitHub.

149

Four people spent hours putting a repo back together at my company after this.

150

GitHub has been unreliable and now they are breaking core tenant of what I expect from this service.

151

Core tenant?

152

You mean tenant, not tenant.

153

You must be fun at parties.

154

I don't care to associate with people who are offended by being corrected.

155

They can sod off and I never go to what you Americans call parties.

156

Whenever you use quotes and they are directly followed by punctuation,

157

you must include the punctuation within the quotes, parties, period.

158

Reply, don't you love Hacker News?

159

It's just such a lovely place where you see really thoughtful and amazing discussions about topics.

160

But at the end of the day,

161

we still do not have the exact post-mortem,

162

so we don't know the exact cause or reasoning,

163

or will GitHub even provide us with the exact details?

164

I don't know.

165

So if anybody can answer these questions,

166

does Microsoft actually use a force underneath the hood to be able to align histories?

167

Were there revert commits, which I don't think there was because nobody is saying that on the internet,

168

wouldn't this cause all sorts of history being absolutely destroyed?

169

Wouldn't this be kind of easy to be able to kind of resolve?

170

Maybe, yeah, it would take a little bit of time.

171

Maybe you have to ask Claude Opus for,

172

you know, mythos to be able to resolve it,

173

but you should be able to get this done pretty quick.

174

And really, was it only 2,800 PRs?

175

Is that it?

176

I mean, Mr. 200 over there doesn't seem to make that to be the case.

177

The name is that I think the most impressive thing out of all of this is

178

that GitHub created such a monumentously good name that even after GitHub ceased to be Git for a little bit,

179

everybody's probably still going to keep on using it.

180

I mean, now that, my friends,

181

is one of the most impressive products of all time.

182

At this point, it seems like GitHub is more addictive than cigarettes.

183

A jet.

184

Hey, is that HTTP?

185

Get that out of here.

186

That's not how we order coffee.

187

We order coffee via sshterminal.shop.

188

Yeah, you want a real experience?

189

You want real coffee?

190

You want awesome subscriptions so you never have to remember again?

191

Oh, you want exclusive blends with exclusive coffee and exclusive content,

192

then check out Kron.

193

You don't know what SSH is?

194

Well, maybe the coffee is not for you.

¿Qué es la Técnica de Shadowing?

Shadowing es una técnica de aprendizaje de idiomas respaldada por la ciencia, desarrollada originalmente para la formación de intérpretes profesionales y popularizada por el políglota Dr. Alexander Arguelles. El método es simple pero poderoso: escuchas audio en inglés nativo y lo repites en voz alta de inmediato, como una sombra que sigue al hablante con solo 1-2 segundos de retraso. A diferencia de la escucha pasiva o los ejercicios de gramática, el shadowing obliga a tu cerebro y músculos de la boca a procesar y reproducir simultáneamente patrones de habla reales. Las investigaciones muestran que mejora significativamente la precisión de la pronunciación, la entonación, el ritmo, el habla conectada, la comprensión auditiva y la fluidez al hablar, convirtiéndola en una de las metodologías más efectivas para la preparación del IELTS Speaking y la comunicación en inglés en el mundo real.

Shadowing Practice: Github are you joking? - Learn English Speaking with Video

Context & Background

Top 5 Phrases for Daily Communication

Step-by-step Shadowing Guide

¿Qué es la Técnica de Shadowing?