deps: update http_parser to 2.4.0

[indutny]

n/t

cc @bnoordhuis@rvagg

[yosuke-furukawa]

+1 !!! Thanks for quick release.

[rvagg]

+1 just for the goto fail additions

how safe is this to land straight into 1.0.0 @indutny? I don't have a feel for how major (and potentially dangerous) these changes are.

[indutny]

@rvagg I feel safe about it, every commit was reviewed by either me or @bnoordhuis

[rvagg]

ok LGTM then

[indutny]

Landed in d790f61, thank you.

[yosuke-furukawa]

I just benchmarked using benchmark/http.sh .
http is improved! Thanks.

	node v0.11.14	iojs v1.x brunch(d790f61)
1	15724.83	17600.49
2	16801.4	18409.31
3	17080.68	18736.1
4	17347.16	18726.99
5	16961.31	18175.39
6	16508.48	17537.85
7	17193.63	18224.03
8	17047.39	18455.47
9	17002.99	18342.36
10	17735.59	18652.9
11	17559.13	18593
12	17607.01	18704.86
13	17159.27	18571.3
14	17524.88	18468.55
15	17302.25	18570.18
16	17604.3	16193.09
17	17684.03	15473.57
18	17282.97	17837.95
19	17567.96	17696.15
20	17471.96	17748.02
Average	17208.361	18035.878

machine spec:
CPU : 2.3 GHz Intel Core i7
Memory : 16GB
OSX Yosemite

[indutny]

Sorry, this is going to be postponed until next release. It breaks build on windows and some tests became flacky (according to users).

[yosuke-furukawa]

hm, no problem. i am looking forward to next release.

[jbergstroem]

Test fail with http_parser 2.4.0 on OS X:

=== release test-http-1.0 === Path: parallel/test-http-1.0
assert.js:100
throw new assert.AssertionError({
^
AssertionError: 1 == 2
at response_validator (/private/tmp/io.js/test/parallel/test-http-1.0.js:154:12)
at Socket.<anonymous> (/private/tmp/io.js/test/parallel/test-http-1.0.js:48:7)
at Socket.emit (events.js:117:20)
at _stream_readable.js:888:16
at process._tickCallback (node.js:337:11)
Command: out/Release/iojs /private/tmp/io.js/test/parallel/test-http-1.0.js

[indutny]

@jbergstroem is it always failing? Appears to be working on my side.

[jbergstroem]

@indutny Just tried once (as well as after pulling the revert). Let me check.

Edit: Nope, can't reliably reproduce.

[indutny]

@jbergstroem I think this is just a test problem. I built http-parser with ASAN, and sanitizers for various overflows, and got no errors and no test failures.

[jbergstroem]

@indutny I haven't seen it since; don't take it into consideration until I can give reliable info.