Fix local host detection for remotes with user

[raindev]

Strip the user when comparing the remote with the hostname of the local machine to prevent repeatedly running topgrade over SSH on localhost.

Details of the issue

Set-up

topgrade is run on a machine with a hostname equal to host, which has user@host present in the remote_upgrades list in topgrade.toml.

Current behavior

topgrade will trigger remote execution for user@host over SSH connection to host. This will repeat indefinitely due to the host recursively connecting to itself during each remote execution.

New behavior

user@host remote will be recognized as being local to host and skipped by topgrade.

Standards checklist:

• The PR title is descriptive.
• I have read CONTRIBUTING.md
• The code compiles (cargo build)
• The code passes rustfmt (cargo fmt)
• The code passes clippy (cargo clippy)
• The code passes tests (cargo test)
• Optional: I have tested the code myself

[codecov]

Codecov Report

Attention: Patch coverage is 90.47619% with 4 lines in your changes are missing coverage. Please review.

Project coverage is 4.05%. Comparing base (728ea26) to head (6e9cd71).

Files	Patch %	Lines
src/config.rs	92.68%	3 Missing ⚠️
src/main.rs	0.00%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff            @@
##            main    #755      +/-   ##
========================================
- Coverage   5.03%   4.05%   -0.99%     
========================================
  Files         37      32       -5     
  Lines      12085    6194    -5891     
========================================
- Hits         609     251     -358     
+ Misses     11476    5943    -5533     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[SteveLauC]

Hi, thanks for your interest in contributing to Topgrade!

From what I understand, this seems to be a new feature rather than a bug-fix as it is explicitly documented that the remote host list should only contain hostnames

to repeatedly running topgrade over SSH on localhost.

Could you please show me the benefit of this new feature?

[raindev]

Hi,

Sorry for the confusion, the description was missing a word:

to prevent repeatedly running topgrade over SSH on localhost

The ability to specify a user is handy in situations when it differs between the local and the remote machine, both having the a copy of the same topgrade config.

Despite the example config file only describes specifying hostnames of remotes, user@host form already works (except that host will attempt to connect to itself when the remote is present in its own configuration, which the PR resolves). Given remote execution is over SSH, this would likely be what the users expect (an example from a random config found on GitHub) - other tools that leverage SSH would usually also allow to specify a user.

[raindev]

I see two approaches for improving how topgrade handles user@ included into a remote: 1) detect this and let the user know that this is not supported 2) take user into account when determining if topgrade should be run remotely. The option 2 is what the pull request implements. Right now including users into a remote will work as expected in some cases but not in the others (see the screenshot below). @SteveLauC let me know what do you think.

[SteveLauC]

Hi, I didn't reply, it is not because I am opposed to this change, it is simply because I am busy with my work and couldn't have time to handle these OSS things, I am sorry about this.

[SteveLauC]

Ok, I just gave this PR a closer look, and the issue described in the PR description indeed exists(thanks for it), the approach proposed in this PR looks good to me!

[SteveLauC]

I am curious why we added this new argument (hostname), looks like the previous way how we handled it still works

[SteveLauC]

Ahh, I get it, it is for tests

[raindev]

No need to apologise, I just wasn't sure if I explained the problem clearly enough previously.

Thanks for your work on topgrade!