Print a usable reproduction command of failed tests.

[sigurdm]

Currently a failing test prints a line like so:

08:28 +54 ~9 -1: test/integration.shard/command_output_test.dart: flutter run --machine uses AppRunLogger [E] 

Running that single test again from the command-line is kind of cumbersome. One has to

• write flutter test or dart test
• copy-paste the test filename
• write something like -n ""<backarrow>
• copy paste the test name.

It would be awesome if the whole command could be printed. Something like:

08:28 +54 ~9 -1: test/integration.shard/command_output_test.dart: flutter run --machine uses AppRunLogger [E] 
To run this test again: `flutter test test/integration.shard/command_output_test.dart --plain-name="flutter run --machine uses AppRunLogger"`

With escaping of tricky characters in the name string.

[jakemac53]

#329 is another related feature, not sure if we would want both?

[sigurdm]

#329 is another related feature, not sure if we would want both?

I would personally use both of these features.

Sometimes you want to drill down into a single failed test.
Sometimes you want to see if you managed to fix all the last failures

[jakemac53]

This one is also significantly easier to implement I think (although it won't be perfect, a basic implementation should work in the 99% case)

[jakemac53]

I looked into this a bit and this is more challenging than it might appear - at least to get the line to consistently show up in a good place.

I could pretty reliably get it to show up as the first line after the failure line, something like this:

00:11 +0 -1: some test [E]  
To run this test again: `dart test test/some_test.dart -p vm --plain-name "some test"`
< failure output here>

But I don't think that is really what you want, I think you probably want it as the last line (after the failure output). The problem is some tests tend to output more text a bit after this failure event comes through, so it ends up in the middle of the output. I think that is a result of printOnFailure calls printing slightly later or something, not quite sure.

[jakemac53]

Ah, so the problem appears to be that a test is marked as "complete" as soon as the first error happens. Really we want to log this after all async expectations are done as well. I figured out how to make sure it doesn't print until the very end of the test.