Port some commits from from joyent/node@master

[Fishrock123]

Ports a couple of commits over from joyent/node (master) that would be useful here.

Refs: #17, #21, #31

7532d24 repl: Private Buffer object in lib/* files should probably have a test of some sort, or we should somehow check that removing the Buffer global doesn't cause issues in core somewhere else.

Chris suggested running Buffer = {} though a repl instance, but I'm pretty sure that would only catch the readline module without extensive use of other modules.

Generally, I'm still leaning towards making global.Buffer non-writable.

[domenic]

Generally, I'm still leaning towards making global.Buffer non-writable.

Mehhhh, that's so un-JavaScriptey. The repl commit is the better approach.

We could make the linter disallow using Buffer as a global.

[Fishrock123]

We could make the linter disallow using Buffer as a global.

👍 Not exactly sure how though. :P

Maybe cc @silverwind?

[silverwind]

You could disable the Buffer global (and all other predefined ones) by removing this here:

https://github.com/nodejs/io.js/blob/9da168b71fb729635ad71e839630480e815623d0/.eslintrc#L2

Afterwards, specify all globals manually here:

https://github.com/nodejs/io.js/blob/9da168b71fb729635ad71e839630480e815623d0/.eslintrc#L78

If you want something readonly add it as false.

[bnoordhuis]

I agree with the general thrust of 7532d24 but it isn't comprehensive enough, it only updates some of the files in lib/.

$ git grep -lw Buffer lib/
lib/_debug_agent.js  # this the only one that does require('buffer').Buffer
lib/_debugger.js
lib/_http_client.js
lib/_http_outgoing.js
lib/_stream_readable.js
lib/_stream_writable.js
lib/_tls_legacy.js
lib/_tls_wrap.js
lib/assert.js
lib/buffer.js
lib/child_process.js
lib/crypto.js
lib/dgram.js
lib/fs.js
lib/internal/smalloc.js
lib/net.js
lib/querystring.js
lib/readline.js
lib/string_decoder.js
lib/tls.js
lib/util.js
lib/zlib.js

[Fishrock123]

@silverwind that didn't seem to work for Buffer. Also turning off node throws some peculiar errors about returning after a process.exit()..

[silverwind]

@Fishrock123 I think that's stuff for another PR. When disabling node you need to specify globalReturn. And to really make it error on global write access, we need to enable no-undef, which results in a lot of linting failures in test, which need to be investigated first.

[Fishrock123]

Ok, removed that commit since it's not comprehensive and we don't have a good way to keep linting it yet.

[Fishrock123]

@silverwind mind signing off on the remaining two commits here?

[silverwind]

LGTM

On a sidenote, I noticed doc/template.html refers to assets/sh.css while the directory is actually called api_assets.

[Fishrock123]

Weird. Noted.