ci: Always checkout sha in build_autoinst.sh

[kalikiana]

There shouldn't be a case where the sha deviates, and if there's a case for it let's make it explicit.

See #4244 for an example where the sha's inexplicably wrong.

[perlpunk]

See #4244 for an example where the sha's inexplicably wrong.

I don't understand where there is something wrong there. Can you link to a CircleCI line where the sha is shown and wrong?

[perlpunk]

I still don't understand the purpose of this PR

[kalikiana]

I still don't understand the purpose of this PR

The sha in #4244 was wrong. I don't fully understand why. This PR makes specifying the sha mandatory, which rules out at least one potential cause.

[perlpunk]

I still don't understand the purpose of this PR

The sha in #4244 was wrong. I don't fully understand why. This PR makes specifying the sha mandatory, which rules out at least one potential cause.

But I asked you for a line showing that "the sha" was wrong.

In #4244 (the dependency PR) apparently the wrong os-autoinst-devel version was installed (and the version has a sha in it), which has nothing to do with the second argument in build_autoinst.sh.
Can you show where build_autoinst.sh is used in the dependency PR and show where exactly the sha was wrong?

[kalikiana]

Can you show where [...]

No. Like I said, this rules out at least one potential cause. The point is that we don't have to doubt this code path next time the question "was the checksum used" is asked.

...