Is this "& when" use case a bug?

[matthew-dean]

I've just started experimenting with "& when" to do inline "if" style blocks. However, I can't get them to actually work (EDIT: I got them to work, sort of, but not as I want or expected them to work). I'm not certain if it's my use case or it's just now how they're currently designed.

Reduced test case:

#test {
  .start(@var: true) {
    & when(@var) {
      .output {
        property: value;
      }
    }
  }
}
#test > .start();

Error: Name error: variable @var is undefined

I thought that perhaps this was triggering a bug within a mixin (even though I regularly use & directly inside a mixin) so I reduced it further:

#test {
  @var: true;
    .start when(@var) {
      .output {
        property: value;
      }
    }
}
#test > .start;

Same error. However, I finally noticed that if I removed #test > .start in the reduced case, it output, so it seems like it's a variable scoping bug. Why does it cause a problem in the first mixin? A defined variable in the mixin should be available to a guard, should it not?

[seven-phases-max]

Example 1 is #1750 (I think it was fixed in 1.6.2).
Example 2 is #57, #719 (it also seems to be partially fixed in 1.6.2 but there's still problems with scope visibility for non-conditional mixins):

#test {
    @var: true;

    .non-parametric when (true) {
        b: @var; // OK
    }

    .non-parametric {
        b: @var; // Err, not defined
    }

    .parametric() when (true) {
        b: @var; // Err, not defined
    }

    .parametric() {
        b: @var; // Err, not defined
    }
}

#test > .non-parametric;
#test > .parametric;

[lukeapage]

also although I realise its just a test case, did you realise your first example can be written in a much simpler way..

#test {
  .start(@var: true)  when(@var) {
      .output {
        property: value;
      }
  }
}
#test > .start();

and closing as duplicate

[matthew-dean]

Ah okay, I should have double-checked if it was addressed already in a Less update.

@lukeapage Yes, but as it's a reduced case, there's more that I was intending on putting in the .start() mixin (and outside of the guard). I just reduced it to the parts that threw an error. 😄

[lukeapage]

there is an open issue for the 2nd one. #921