a crack at solving #105

[jameslnewell]

No description provided.

[jameslnewell]

Oh wait, I missed #106

[jameslnewell]

Reopening because #106 is only creating a new engine when a configFile prop is used. The config can also be passed as an object.

[jaythomas]

I like your more-robust solution over #106. One thing you might want to adjust however is the test:

"should report a single error because only one config causes an error" does not infer that "eslint-loader will create an engine for each unique config". Whether or not the second config is ignored, it will always report only one warning (not error). In my test I assured both configs were loading by throwing a unique warning from each: https://github.com/MoOx/eslint-loader/pull/106/files#diff-d1a0eb3a6e5c2e85a9cd54e5f92bfd7fR85

Feel free to borrow my solution or disregard!

[jameslnewell]

Thanks!

Sorry, I'm not sure if I follow what you're saying... With my test, in the case where:

• the first eslint config is used for both instances, there will be 2 warnings generated (and we're expecting 1)
• the second eslint config is used for both instances, there will be 0 warnings generated (and we're expecting 1)
• no eslint configs are used, there will be 0 warnings generated (and we're expecting 1)
• both eslint configs are used, only the first will trigger a warning (and we're expecting 1) so it will pass

[jaythomas]

What I mean to say is here, this rule will never throw an error so your assertion that stats.compilation.warnings.length === 1 is always true even if that second config object is never loaded or if there were another unforeseen problem with the webpack settings. So the test doesn't directly verify that "eslint-loader will create an engine for each unique config" as your test is named.

My suggestion around this would be to change the second config object rule from quotes: [1, "double"] to something that will throw a second warning, (semi: [1, "always"] for instance). You will know the second config is loading correctly if it throws the expected warning from the second config. You can then change the assertion to read stats.compilation.warnings.length === 2 instead of === 1.

[jameslnewell]

Cool, I think I get you, however that introduces the problem where if only the first config is used, you'll still get two (quote) errors and the test will incorrectly pass. Do we need two separate tests? Got any ideas?

[jameslnewell]

^^checking that I get two different types of warnings

[jaythomas]

I have closed PR #106 in favor of this one.

[jameslnewell]

@MoOx thoughts?

[MoOx]

Great work!

[MoOx]

Releasing this as 1.6.1. Thank you @jameslnewell and @jaythomas!