Don't warn on prefer global for devDependencies

[saper]

Windows Powershell assumes the command has failed
whenever something is written to stderr.

Followup-To: #1648
Fixes: #8517

[saper]

I think that da015ee might have uncovered an issue by nuking the cache, is this correct @othiym23 ?

[othiym23]

@saper This looks good, but can you add a test, please? Given that you ultimately identified #8517 as a regression, a test case that makes sure that nothing gets printed to stderr for preferGlobal dependencies and devDependencies ought to do it.

[zkat]

I worked for a while on writing tests for this, but ran into issues with getting the test to actually fail. The plan is to take care of it next thing tomorrow and definitely get it in with next week's release, but I'm bumping this down the line one more time. Gotta figure out what's up with our test setup.

[othiym23]

@zkat found another bug related to this functionality while writing the tests for it, so we're going to need a little more time to address it adequately. Next week's release, this time for sure! :D

[saper]

I have suspected there is some other issue, since da015ee alone shouldn't be enough to cause the problem...

[othiym23]

@zkat is out for most of this week, and she has a branch with some of the work done for fixing this, so it's getting bumped to the next next release one more time. ;)

[saper]

Ah the benefits of distributed version control ...

/me hides :)

[zkat]

Thanks, @saper! I'm moving discussion + review to #9409 cause I wrote a bunch of tests and added some bugfixes over this. :) I'm closing this PR for now.

[othiym23]

This appears to have come undone in npm@3, so I'm going to reopen this so that @iarna and / or @zkat can take a look at it (and I can re-close an issue that was originally opened in 2011). I've noticed this behavior too, and am not sure exactly what's up.

[othiym23]

Whoops, not going to reopen the PR, will reopen #8517 instead.