diag(cts_runner): raise log level to info by default

[ErichDonGubler]

No description provided.

[andyleiserson]

I'm not sure this is the right way to do this... won't this suppress INFO-level logs in cts_runner?

[ErichDonGubler]

@andyleiserson: I'm not sure what you mean. cts_runner only uses env_logger::init right now, which only prints at error-level messages by default (see this playground). We should only be increasing the log level with a change like this.

Now that you mention it, it would be better to just fix our env_logger init. 😅 I hadn't fully mentally loaded the control flow here with the graph of components before your pushback.

[cwfitzgerald]

I think the CI failures are legit, it's getting thrown by the log message it seems

[andyleiserson]

cts_runner only uses env_logger::init right now, which only prints at error-level messages by default

You're right, I was thinking of the INFO messages logged by the xtask, but they wouldn't be affected by this.

I think the CI failures are legit, it's getting thrown by the log message it seems

May be fixed by https://github.com/gfx-rs/wgpu/pull/9709/changes#diff-1cd4ad01cffde4d05952fd3b0450f2a5db3b9fbd2af913f707940db2efc8a5b6R683. (The message was there even before this PR, we just didn't see it because the logging wasn't enabled.)

[ErichDonGubler]

@andyleiserson:

May be fixed by https://github.com/gfx-rs/wgpu/pull/9709/changes#diff-1cd4ad01cffde4d05952fd3b0450f2a5db3b9fbd2af913f707940db2efc8a5b6R683. (The message was there even before this PR, we just didn't see it because the logging wasn't enabled.)

I opted to set RUST_LOG to error on the cts_runner invocations in the broken snapshot tests with the latest force-push. Info-level log entries seem like something that aren't useful to snapshot, but I could be persuaded that setting it to warn and updating snapshots to include the Vulkan errors could make sense (as follow-up).