Unhandled Composer Stager exceptions leave the update process in an indeterminate state

Per #3331355-23: [PP-1] Change exception class Architecture → , #3331355 is itself blocked. Reflecting that here.

📌 Refactor exception architecture Fixed is unblocked; it's being worked on 👍 Reflecting that here.

📌 Refactor exception architecture Fixed is 99% done — the direction is locked in, and it's down to nitpicks! 🥳

This issue is actionable even before that lands:

1. create a new branch/MR, against 3.0.x (the existing one is against 8.x-2.x)
2. on that new branch: curl https://git.drupalcode.org/project/automatic_updates/-/merge_requests/718.diff | git apply -3v && git commit -nam 'Issue #3331355'
3. → now you can build on top of that!

Looks like that works! But … I don't think tests have been written yet? 🤓 Without those, I don't have that much confidence in this 😅

We'll need to simulate hard crashes in all of the different calls to composer-stager!

📌 Refactor exception architecture Fixed landed.

That means we should revert the commit that we did here that added an in-progress version of that 😊

Tests exist now, but some things are not yet clear to me 😅

I asked for one more clarifying comment, but … this is dealing with a part of the codebase that I'm not familiar enough with. Leaving to Ted or Adam to review!

I made a few changes here:

• I'm not super comfortable with very large try blocks. If a lot is happening in there, it becomes a lot muddier to understand when we expect exceptions to maybe happen, and why. So I tried to focus that scope down with an anonymous helper function.
• This pattern of catching a throwable and then re-throwing it is used frequently in Stage, so I decided to make a helper method for it. That way, we don't have to remember to always pass the previous throwable to the re-thrown exception.
• In this case, I think instanceof is more readable than in_array().

phenaproxima → closed merge request !630

I think @phenaproxima meant to mark this on Friday.

A few minor things, and then a couple of blocking questions.

No nits found, but the blocking logic issue is still not clear to me. If someone can explain it to me (or just show me what I'm missing), and maybe add a comment, I'd be happy to RTBC and commit this.

A few minor things, and then I think this is good to go. It makes a lot more sense!

Fixed the nitpicks. I think this looks good; asking @tedbow for final review.

Exciting to see this one land — this definitely helps make Package Manager more robust! 🚀

Found a few nits, but keeping RTBC because they're minor remarks.

See MR comments

Looks good!

Nice to get this fixed!

Automatically closed - issue fixed for 2 weeks with no activity.