sources: update to tokio v1

[webern]

Update all libraries that have tokio dependencies to their tokio v1
versions. Note that Hyper >= 0.14.3 causes a build issue so we have
pinned version 0.14.2. (tracked in #1471)

Issue number:

Closes #1269

Description of changes:

Update all dependencies in the sources workspace to use tokio v1.

Testing done:

• verified that only one version of tokio exists in the lock file
• ensured that webpki-roots-shim is being used
• ran a v1.0.7 version (without my changes) and upgraded into a pseudo-1.0.8 version with my changes and with the 1.0.8 migrations. downgraded, ran pods at each step and logged in to the admin container.
• ran an ami with my changes
  • used the apiclient to set motd
  • checked the journal with journalctl -p 3, no errors

Terms of contribution:

By submitting this pull request, I agree that this contribution is dual-licensed under the terms of both the Apache License, version 2.0, and the MIT license.

[tjkirch]

I guess I'll start the discussion - I'm more comfortable holding off on some dependency updates than taking a beta release of one of our most important dependencies. 3.x has seen a lot of use and presumably 4.x would see a greatly increasing amount after release; we'd be taking on a lot of testing debt by going forward with an early release. It doesn't seem like there's a way to update just actix-rt (which is tokio 1+ and not in beta) without updating actix-web to 4.0 beta, for now, so I'd vote to wait on this PR.

[webern]

The counterargument is that that the issues open for the 4.0 milestone don't seem like blockers:
https://github.com/actix/actix-web/milestone/6

• The bug named "A Stream object in a pending state won't be dropped after the client disconnect" has been open since Jan 2020.
• The bug named "actix-multipart file fields might not be waking properly" is not relevant to us since I don't think we use actix-multipart (?)
• Two api improvements that might require code changes, but don't represent defects in 4-beta
  • "review compression flow"
  • "decouple web and client"

My impression is that 4.0 could be cut right now, but that the project doesn't want to burn a major version number just to get tokio v1 support out when they have some other breaking changes in mind.

On crates.io 4.0.0-beta.X versions are getting at about 1,300 downloads per day compared to 5,500 downloads of all other versions, so about 19% of actix-web users are opting for 4-beta.

[webern]

Update: I checked with a maintainer on the Gitter channel and they said that 4.0.0-beta is no less production ready than v3. The beta label is because there might be a couple of minor breaking changes before 4.0-proper is released.

[webern]

webern force-pushed the webern:tokio branch from ec54e1b to a0767a0

Fix a fixme

[webern]

webern force-pushed the webern:tokio branch from a0767a0 to 62ec7f6

rebase

[webern]

webern force-pushed the webern:tokio branch from 62ec7f6 to 40642da now

rebase

[webern]

webern force-pushed the webern:tokio branch from 40642da to 6e3c139

Rebase.

[webern]

webern force-pushed the webern:tokio branch from 6e3c139 to 3790105

Rebase.

[webern]

webern force-pushed the webern:tokio branch from 3790105 to e32c0ca

Rebase.

[webern]

webern force-pushed the webern:tokio branch from e32c0ca to 282b07b

Rebase.

[webern]

webern force-pushed the webern:tokio branch from 282b07b to 2ebd5d1

Rebase. I will also re-test.

Edit: got at least one unexpected EOF from CI.

[webern]

I retested everything and it checks out. I saw this in the journal Apr 29 18:42:58 ip-192-168-60-200.us-west-2.compute.internal wickedd[2707]: ni_process_reap: process 7043 has not exited yet; now doing a blocking waitpid() but I think it's unrelated.