How to handle logs interleaving with Hurl progress bar

[Thomasdezeeuw]

Current Hurl overwrites the previous line when it writes its progress bar (relevant code here:

hurl/packages/hurl/src/util/term.rs

Lines 189 to 202 in 4cf44ff

	// We count the number of new lines \n. We can't use the `String::lines()` because
	// it counts a line for a single carriage return. We don't want to go up for a
	// single carriage return.
	let lines = self.progress_bar.chars().filter(|c| *c == '\n').count();
	// We used the following ANSI codes:
	// - K: "EL - Erase in Line" sequence. It clears from the cursor to the end of line.
	// - 1A: "Cursor Up". Up to one line
	// <https://en.wikipedia.org/wiki/ANSI_escape_code#CSI_sequences>
	if lines > 0 {
	(0..lines).for_each(|_| eprint!("\x1B[1A\x1B[K"));
	} else {
	eprint!("\x1B[K");
	}

). This allows it to "progress" the progress bar.

The problems comes when another process logs to the same output.

We use Hurl for our testing (which is great btw -- well done!), but it will overwrite any of the application log which we log to standard out/error by default. Currently we work around this by logging our service logs to a separate file and hope we don't have to look at :) (i.e. no errors happen). But this is less than ideal.

Are there any better solutions for this?

[jcamiel]

Hi @Thomasdezeeuw

With --test, there is currently no explicit way of disabling progress bar. As a workaround, you can either pipe Hurl's stderr to a file (that will disable progress bar), or declare an env var "CI" that will also disable progress bar to be drawn. That said, we plan to add an explicit --progress-bar option to force display of progress bar, we can also add a --no-progress-bar to explicitly disable it.

[Thomasdezeeuw]

Thanks for the response @jcamiel. It's not so much about disabling the progress bar, although that would solve the issue, it's more about it not overwriting the log output.

For a bit more context, when running our tests we expect zero logs from our services as we only log errors. But if a test does fail the log from our service is now (partially) overwritten. The progress bar is quite useful when actively working on the code & running the tests in parallel.

So basically I'm asking for "don't overwrite if another process wrote the previous output", which I don't think is possible. I was wondering if anyone found a better work around than logging to a separate file.