swh/core/config.py
101	That's an info at best, as it'd log an error message if a config file voluntarily doesn't exist (e.g. during tests). `exists_accessible` already raises relevant errors when a file is present but inaccessible.

This revision now requires changes to proceed.Dec 21 2018, 3:17 PM

That's an info at best, as it'd log an error message if a config file voluntarily doesn't exist (e.g. during tests).

exists_accessible already raises relevant errors when a file is present but inaccessible.

We agreed that tests should not try to load config files. And allowing config files to be missing without reporting errors is asking for trouble.

In D878#18891, @vlorentz wrote:

That's an info at best, as it'd log an error message if a config file voluntarily doesn't exist (e.g. during tests).

exists_accessible already raises relevant errors when a file is present but inaccessible.

We agreed that tests should not try to load config files.

Okay?

And allowing config files to be missing without reporting errors is asking for trouble.

All the systems currently in production depend on this behavior: they try to load local configuration (in ~/.config), which fails (expectedly!), then they load the configuration in /etc.

All in all I don't think all workers should immediately start spewing a handful of error messages at each run for something that is _expected_.

I'd even go as far as putting those messages at DEBUG level.

In D878#19188, @olasd wrote:

All the systems currently in production depend on this behavior: they try to load local configuration (in ~/.config), which fails (expectedly!), then they load the configuration in /etc.

Indeed, I forgot about that.