Deno: Deno 1.4.0 fmt --quiet option regression

[srackham]

The quiet option is not honoured:

$ deno --version
deno 1.3.3
v8 8.6.334
typescript 4.0.2

$ deno fmt --quiet Drakefile.ts 

$

$ deno --version
deno 1.4.0
v8 8.7.75
typescript 4.0.2

$ deno fmt --quiet Drakefile.ts 
Checked 1 file

$

The Checked N file message appears to be new (fmt was quiet by default in Deno 1.3.3).

[caspervonb]

Yeah this was initially behind a verbose flag, the flag was dropped and we're here.

[magurotuna]

Thanks for the report. I forgot to consider quiet flag when implementing that feature. I will fix it shortly.

[magurotuna]

Seems that outputs of subcommands have been going to _stdout_ since https://github.com/denoland/deno/pull/7142, so the description of quiet option is actually wrong now.

    -q, --quiet
            Suppress diagnostic output
            By default, subcommands print human-readable diagnostic messages to stderr.
            If the flag is set, restrict these messages to errors.

It would be nice to revert it so that logger output will go to _stderr_.

[bartlomieju]

@magurotuna seems like an unintended change, could you fix that in the same PR?

[magurotuna]

@bartlomieju Yes, of course.
Additionally, I'm planning to make the integration tests check stdout/stderr separately. That said, currently the integration tests are checking outputs with stdout and stderr together. I will change it, so actual _stdout_ will be compared with expected _stdout_, and actual _stderr_ with expected _stderr_.
It would help us make sure that outputs are printed to correct (expected) places.

[magurotuna]

Seems that outputs of subcommands have been going to stdout since #7142

my bad, this is wrong. The outputs of subcommands indeed go to _stderr_, as they did before. (I missed env_logger document where we can see the default target is _stderr_.)
But nevertheless I think it should be nice that stdout and stderr are tested separately, so I'll keep implementing it.
And then, I will use the logger functions instead of println! and eprintln! in fmt and lint. It will allow us not to take care of quiet flag in each subcommand implementation.

[bartlomieju]

But nevertheless I think it should be nice that stdout and stderr are tested separately, so I'll keep implementing it.
And then, I will use the logger functions instead of println! and eprintln! in fmt and lint. It will allow us not to take care of quiet flag in each subcommand implementation.

@magurotuna sounds good, let's start with fixing --quiet regression - I don't think there's a lot of need to have stdout/stderr tested separately in itest! - in such cases we just write test case by hand, eg. run_watch.

[magurotuna]

@bartlomieju
okay, I'll fix the quiet regression first.

I don't think there's a lot of need to have stdout/stderr tested separately in itest! - in such cases we just write test case by hand, eg. run_watch.

To be honest I have about 70% finished implementing the feature, so I'd like to open a PR when completing it. And at that time if you and other people think of it as unnecessary, feel free to close it without merging :)