Multiple breaking changes

[ptyork]

Select Topic Area

Bug

Body

Okay, so this is quite worrisome to me. Recent builds of codespaces (and/or perhaps the upstream Docker images; mcr.microsoft.com/devcontainers/base:ubuntu in my case) have broken multiple things.

First (and most "who cares"), my hacks to remove the terminal message have been "patched" (see #43534). Not only do they no longer work, but now they want to show that confusing message TWICE. Cool. I know it was a hack, but it was "fixed" immediately after I was thanked publicly for finding the solution. ...sigh...

Second, and VERY impactful, the remoteUser functionality has been altered in a way that locks my students out of their Codespaces. A couple weeks ago, this worked just fine:

	"features": {
		"ghcr.io/devcontainers/features/common-utils:2": {
			"username": "${localEnv:GITHUB_USER}",
			"userUid": "1999",
			"userGid": "1999"
		}
	},
	"remoteUser": "${localEnv:GITHUB_USER}"

NOW, any container built with this configuration builds and starts up just fine, but you're not able to connect (see #46084). Apparently the common-utils script still creates the proper user from that environment variable, and it can work if you hard code the username, e.g.:

	"remoteUser": "ptyork"

But that entirely defeats the purpose of creating these custom, non-root users in the first place (see #43591).

What happened to break this and why? I now have multiple students with uncommitted changes in containers that are now completely inaccessible to anyone...thus their work is effectively lost to them. Can this change be reverted?

Finally, the last breaking change. The "Codespaces: Rebuild Container" command now fails, making finding and debugging all of above even more difficult. Gives me this error message:

Rebuild Container: F.sendTelemetryEvent is not a function

Fun.

Oh, and I did submit a ticket for my students being locked out 30'ish hours ago, now, but it has yet to be acknowledged.

Sorry if this comes across as snarky. If you can't tell, it's been a frustrating couple of days.

[jge162]

Your post is extremely detailed and warrants attention. My only recommendation is to submit. Bug report to see if they respond faster. Have you exhausted that Avenue yet?

[jge162]

Well the developers occasionally respond here. Hopefully they see this.

[jge162]

I think the call for help -> has gone out loud and wide.

[jge162]

Did you not support a support ticket? You should as they respond very quickly to them. ? Just asking to be sure

[jge162]

Here is link if you did not submit one -> Support

[ptyork]

Yes, ticket number and all (#1994445). FWIW, I was a little unfair earlier. It's only been 26 hours, not 30. ;) They are experiencing "high volume" according to the automated response. Maybe they are. Also maybe they were adversely impacted by recent layoffs (support is usually one of the first and worst hit). Trying to be patient...

[ptyork]

@otter-computer are you still around and able to help with this or with my support ticket? I'm now over 36-hours waiting on a breaking change that has me dead in the water. The remoteUser user change has made all active containers as well as all repositories copied and cloned from my templates completely inoperable.

I just need an answer. If it is "will not fix" then fine. I'll be a bit angry as will my students, but we'll deal with it. If it is "we're actively looking into it" then cool. But something to let me know whether I need to stay the course or to pivot and apologize to my students...PLEASE.

[ptyork]

Version: 1.75.0
Commit: e2816fe719a4026ffa1ee0189dc89bdfdbafb164
Date: 2023-02-01T15:20:55.110Z (5 days ago)
Browser: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/109.0.0.0 Safari/537.36 Edg/109.0.1518.78

And just reverified that I'm still getting error.

[dmgardiner25]

Thanks! I'll do some more investigation and report back.

[dmgardiner25]

I took a look at your codespaces and it looks like the rebuilds were all succeeding in the backend meaning that even though you were seeing the error popup in the client, the rebuilds were actually working. If you run the rebuild command and then refresh your page, I would expect the changes you made to be present. Still digging into the error, just wanted to let you know 🙂 Scratch that, looks like it's not actually working after reproing myself 🤔

Are you seeing this error as well when you connect to the codespace? Something is definitely not behaving correctly with the client.

[dmgardiner25]

@ptyork Can you try removing the cweijan.vscode-mysql-client2 extension from your devcontainer? There seems to be some sort of conflict with our extension and removing that extension solved it. Currently looking into why this is happening and how we can solve it.

[ptyork]

Sho' 'nuff. Actually, I just noticed that it gives that same telemetry error message just trying to connect to the container if that extension is enabled. Though it seems to just ignore it there. The only place I can see that it really impacts things is with the container rebuild.

I can't disable it permanently in the near-term, as it is integrated into all of my lessons. However, as long as it only impacts the rebuilds, that's not really a big deal. And I can definitely confirm that it does seem to be the root of the problem(s). Nice find.

(as an aside, it strikes me that an extension for reading and writing databases would be one heck of a good way to steal sensitive data...sure hope this isn't a sign that something isn't kosher with this thing)