import-once subdirectory test fail

[dmcass]

@agatronic I noticed a regression today when working on a PR for another bug. Commit a702b28 introduced a bug in @import-once functionality which is visible in the tests. The import-once test demonstrates the issue well enough.

import-once.less

@import-once "import/import-once-test-c";
@import-once "import/import-once-test-c";
@import-once "import/deeper/import-once-test-a";

import/import-once-test-c.less

@import-once "import-test-d.css";
@c: red;

#import {
  color: @c;
}

import/deeper/import-once-test-a.less

@import-once "../import-test-d.css";

output

@import "import-test-d.css";

@import "../import-test-d.css";
#import {
  color: #ff0000;
}

Don't have a chance to look into it right now, and you're more familiar with that part of the codebase than I am.

[lukeapage]

As the commit message says, its not a regression. I fixed a bug that was making less ignore import statements depending on the order of the io... You actually get the fail 1 time in 8 before the fix and this made it deterministic.

The problem is summed up in other bugs.. the CSS path isn't being adjusted, therefore actually import_once is correct as its referencing 2 different paths.

I left the tests failing as import behaviour needs to be sorted out.

[dmcass]

Ah, sorry for the confusion then. I misunderstood the commit message.

[lukeapage]

@cloudhead

Please could you take some time to state how you think it should work? e.g.

1. Should all urls be adjusted to be relative to the base path for both node.js and browser (or was this functionality purposefully removed) and if it should be fixed or put back in, should it be controlled by an option
2. should there be a difference between url("") and "" in terms of the way it is adjusted?

[lukeapage]

basically the node code never adjusts the paths. This is pretty fairly discussed in #331 so closing this as a duplicate. I have fixed the test until 331 is sorted out.