Add --no-trace option, overrides pytrace=True for pytest.fail

[FortStatement]

Hey, I just opened this issue to open a PR for the fork I'll be using for my automated exercise testing.
Hoping this change can be useful to others as well 😄

[blueyed]

The idea/issue is nice, but I think it needs a better / more central method.
Note that _repr_failure_py is bad already (with regard to typing) to return a str here in some cases - this should not get extended. (btw: it also does not handle --full-trace here (blueyed@140d256)).

So I think the option in general makes sense, but should be handled diffrently maybe.

btw: have you tried --tb=line?

[blueyed]

Another idea that comes to mind is enforcing pytrace=False in OutcomeException always when the option is enabled.

[FortStatement]

That's where I started, but the session isn't available there in a clear manner.
The _repr_failure_py is a relatively nice place for it, but perhaps it should be organised a bit more nicely.

[nicoddemus]

Note that _repr_failure_py is bad already (with regard to typing) to return a str here in some cases - this should not get extended.

Not sure I agree... I don't think the method is being extended; it used to return str in some instances, and continues to return str in an additional condition... is that what you mean by "extended"?

I agree that _repr_failure_py is a bit messy, but this is not making it worse than already is.

[blueyed]

Note that _repr_failure_py is bad already (with regard to typing) to return a str here in some cases - this should not get extended.

Not sure I agree... I don't think the method is being extended; it used to return str in some instances, and continues to return str in an additional condition...

I've looked into making it not return str anymore, so this would make that a bit more difficult.

Ref: #3399

@GPery

btw: have you tried --tb=line?

[FortStatement]

@GPery

btw: have you tried --tb=line?

I have not.
This will do nicely, thank you!

[nicoddemus]

So I think --tb=line is sufficient and we can close this @GPery?

[FortStatement]

No, I was mistaken on the last comment, --tb=line still outputs the file path, which doesn't work for me. --tb=none also wouldn't do, because it doesn't output the failure message. I'll keep using the fork, but I'm not sure if my use case is common enough that anyone would find it necessary.

I would appreciate the merge, it's a minor change that'll make things easier for me (keeping up to date with upstream).

[bluetech]

My opinion is that this need is not common enough to deserve a configuration option, so I propose we close this PR. The issue can remain open in case someone comes up with another suggestion for satisfying this need.

[bluetech]

Now that #7100 is merged, we have a better avenue for resolving this issue, as mentioned here: #7100 (review), so I'm closing this PR.