Dealing with unexpected errors

[jhauberg]

Trying to reach remote repositories can (obviously) fail due to many different factors, but seeing an erroring stacktrace like below is not very helpful to the end-user:

In this particular case I happen to know why it fails (accessing private repo but missing key/bad git configuration), but it could be so many things.

Here's a case where I just pulled the plug:

$ git ls-remote --tags --quiet
fatal: unable to access 'https://github.com/jhauberg/gitdoctor.git/': Could not resolve host: github.com

Here's a case where remote no longer exists:

$ git ls-remote --tags --quiet
remote: Repository not found.
fatal: repository 'https://github.com/[redacted]/[redacted].git/' not found

I think, in most cases, the git command we're trying to execute will just exit with a fatal, unknown, error, like the error code 128 as seen in the above screenshot.

We absolutely should not try to determine the root cause, because that is going to get extremely out of scope as I see it. We should, however, not bail completely, as there is still potential to diagnose other defects (they don't all require remotes access).

I suppose we could point out during diagnosis that something did not work out and just move on, or even just skip it silently. It could also be noted lastly that some defects could not be diagnosed. I'm not sure what is most preferable.

Edit:
As it may be of use, the actual output/error for the above screenshot is:

$ git symbolic-ref --short refs/remotes/origin/HEAD
fatal: ref refs/remotes/origin/HEAD is not a symbolic ref

Turns out that, though still related, is actually a separate issue entirely (see #3).

[jhauberg]

So this issue can essentially be boiled down to whether or not we do check=True in subprocess.run, and, if we don't, how we make it clear, when, and if, something didn't go as expected before we move on.