inconsistent case sensivity

[aardgoose]

Importing a module in two places using paths that are different only in case, but handled as case insensitive, results in two instances of the imported module in the resultant output. (Windows 10).
, rollup v 0.34.8

ie. file lookup is case insensitive
ie. module instancing is case sensitive based on path name differences.

Eg.

Module 1:

import { Materials } from './materials/Materials.js';
...

module 2:

import { Materials } from './Materials/Materials.js';
...

Resultant output file has

Materials and Materials$1 objects with expected breakage.

[snuggs]

I feel the lowest common denominator is POSIX compliance which the following volume formats abide by (for the most part):

Linux and Derivatives ( ! OSX)

√ Files MUST be case sensitive. (http://superuser.com/questions/431342/linux-both-case-sensitive-and-case-insensitive-and-always-inconvenient)

Windows (NTFS)

√ Files MUST be case sensitive (https://support.microsoft.com/en-us/kb/100625)

Windows (VFAT | FAT32)

√ Files MUST NOT be case sensitive

OSX

√ Files are SHOULD be case sensitive (but aren't).

Therefore IMHO I feel like we should throw instead of returning a false positive regardless of the solution. "Failing" silently seems too flagrant and would proliforate confusion that is already at a (potential) disagreement on the respective operating system. (I'm lookin' at YOU WIN32 Your thoughts @aardgoose

[pmart123]

@snuggs I like the idea. From prior experience of being burned in similar situations, throwing the error is a nice gesture in comparison to the amount of potential time spent tracking something like this down.

[Rich-Harris]

I actually opened a PR for this yonks ago – #630 – but forgot about it. Just merged it, will cut a release soon. Thanks

[snuggs]

Well that's what i'm here for @Rich-Harris a little healthy nudge. Until can be more productive. Got a great thing going on here. 👍 #FACTS

[aardgoose]

Many thanks, may save me some head scratching time in future.