GitLab

In f67d9cd8 we pass these 2 arguments
to the cronjob: --minimum-speed "5 MB/s"

${*} will split this and pass 3 arguments to check-mirrors.rb:

 * --minimum-speed
 * 5
 * MB/s

… which is invalid usage.

So instead, use "$@" that does not split arguments.

Reference: https://github.com/koalaman/shellcheck/wiki/SC2048

For some reason I was expecting a list of args to be passed to exec(),
which actually we're building a shell command line as a string :/

This will help us identify slow mirrors, which:

 - may help find a reasonable timeout value for sysadmin#17702

 - will help with
   https://gitlab.tails.boum.org/tails/blueprints/-/wikis/HTTP_mirror_pool#improve-ux-and-lower-maintenance-cost-2021

I chose 200m as default timeout for the `check-mirrors.rb` runs because
of the following reasoning:

  - There are 4h between first (longest) and second run of
    `check-mirrors.rb.`

  - Each job can sleep up to 2048s (~35m) before actually running the
    command.

  - This leaves ~205m for the check-mirrors.rb command itself.

  - A manual run took ~165m.

  - Let's timeout after 200m then, to ensure the previous job will
    always be killed before the next one starts.

related to: sysadmin#17338

refs !54

For tails#12557 the invocation of faketime is moved out from happening
inside the builder VM to its host, which on Jenkins is the
tails::iso_builder VM.

See: https://github.com/pypa/setuptools_scm/issues/541

This makes our CI setup robust against test suite jobs that abort
in a way they leave files owned by root in the workspace.

One of the branches I've been working on recently does this,
and as a result, since a few days, isotesters are regularly
left offline because jenkins-slave.service fails to start,
because it is not allowed to "/bin/rm -rf /var/lib/jenkins/workspace/"
as the "jenkins" user. Every time, I have to manually re-run
this command as root, then restart jenkins-slave.service.
Let's automate this :)