00:00:00  * ircretaryquit (Remote host closed the connection)
00:00:07  * ircretaryjoined
00:08:35  * kazuponjoined
00:09:14  * jmar777quit (Remote host closed the connection)
00:09:53  * jmar777joined
00:11:55  <trevnorris>TooTallNate: You think it would be bad form to write the tests so they'd support both quiet and signaling NaNs?
00:12:59  * jmar777_joined
00:13:21  * kazuponquit (Ping timeout: 248 seconds)
00:13:29  * jmar777quit (Read error: Connection reset by peer)
00:15:29  * TheJHquit (Ping timeout: 248 seconds)
00:16:45  <TooTallNate>sblom: mathieu.cartoixa@free.fr? yes he's signed it
00:17:31  <sblom>TooTallNate: Awesome--thanks. Is there a good way for me to answer CLA questions for myself?
00:21:15  * kazuponjoined
00:21:40  <bnoordhuis>sblom: isaacs needs to grant you access to the cla. it's a google docs document
00:27:11  <TooTallNate>trevnorris: personally i'd think the function should always be consistent across platforms/processors‚Ķ but i don't really care about this signal vs. quiet NaN problem
00:27:52  <TooTallNate>trevnorris: that is, i've never needed one or the other specifically‚Ķ but will readFloat()/readDouble() read NaN properly back in from either type of encoded NaN?
00:37:43  <trevnorris>TooTallNate: iirc bnoordhuis had the most to say about it. can't fine the thread, but here was some discussion on it: http://git.io/ntK11w
00:38:18  * pooyaquit (Quit: pooya)
00:39:23  <trevnorris>but either bnoordhuis or pisc will have to remind me why it was important.
00:43:23  * kazuponquit (Remote host closed the connection)
00:49:46  * rumpjoined
00:50:17  * mikealquit (Quit: Leaving.)
00:51:14  <bnoordhuis>pquerna: https://github.com/joyent/node/pull/4647
00:55:37  <isaacs>sblom: yes, he has signed, it looks like
00:55:50  <isaacs>sblom: also, i shared the doc with you. usually searching by email address is most effective.
01:01:05  * kazuponjoined
01:01:23  * kristatejoined
01:02:59  * trevnorrisquit (Quit: Leaving)
01:04:42  * c4miloquit (Remote host closed the connection)
01:05:33  * c4milojoined
01:07:12  * hzquit
01:09:55  * c4miloquit (Ping timeout: 248 seconds)
01:11:42  <pquerna>bnoordhuis: looks reasonable.
01:17:28  <isaacs>bnoordhuis: i'm fine with it
01:17:36  <bnoordhuis>cool
01:18:11  <MI6>joyent/node: Ben Noordhuis master * 2cbf458 : http: close connection on 304 and chunked encoding Force the connection - http://git.io/ezRkOw
01:20:49  * loladirojoined
01:23:47  * c4milojoined
01:24:52  <isaacs>bnoordhuis: any progress on master's perf regression?
01:25:00  <isaacs>bnoordhuis: (i've been busy with other stuff myself)
01:25:14  <bnoordhuis>not really. i have / am busy with other things as well
01:25:44  <isaacs>do you think you'll get any time to look at it this week?
01:27:11  <bnoordhuis>isaacs: i hope so but...
01:27:52  <isaacs>at this point, the only thing holding 0.10 from going out is that our performance is < 0.8
01:28:22  <isaacs>and the udb clustering, but meh
01:28:24  * c4miloquit (Remote host closed the connection)
01:28:38  <isaacs>hardly anyone cares about that, just those folks bribing you and bert ;P
01:28:51  * c4milojoined
01:32:10  * stagasquit (Ping timeout: 246 seconds)
01:33:22  * c4miloquit (Ping timeout: 256 seconds)
01:35:51  * mikealjoined
01:42:22  <isaacs>indutny: ping. test/simple/test-child-process-fork-getconnections.js is failing
01:42:44  <isaacs>indutny: oh, is sporadic
01:42:49  <isaacs>indutny: not failing consistently
01:53:52  * rumpquit (Quit: rump)
01:56:13  * dapquit (Quit: Leaving.)
01:56:33  * dapjoined
01:58:13  * mikealquit (Quit: Leaving.)
02:01:13  * dapquit (Client Quit)
02:01:28  * sblomquit
02:08:28  * kazuponquit (Remote host closed the connection)
02:09:40  * qmxchanged nick to qmx|away
02:23:32  * lohkeyquit (Ping timeout: 248 seconds)
02:25:31  * kazuponjoined
02:35:55  * mikealjoined
02:40:26  * mikealquit (Ping timeout: 255 seconds)
02:41:01  * c4milo_quit (Remote host closed the connection)
02:45:41  * kazuponquit (Remote host closed the connection)
02:57:48  * TooTallNatequit (Quit: Computer has gone to sleep.)
03:03:02  * jmar777_quit (Remote host closed the connection)
03:03:40  * jmar777joined
03:06:15  * mikealjoined
03:07:10  * felixgejoined
03:07:10  * felixgequit (Changing host)
03:07:10  * felixgejoined
03:07:45  * jmar777quit (Ping timeout: 248 seconds)
03:24:54  * brsonquit (Ping timeout: 276 seconds)
03:40:51  * cjdquit (Ping timeout: 248 seconds)
03:53:13  * kazuponjoined
03:58:13  * wavdedjoined
04:03:06  * lohkeyjoined
04:08:09  * brsonjoined
04:21:14  * bnoordhuisquit (Ping timeout: 255 seconds)
04:43:38  * TooTallNatejoined
04:43:40  * TooTallNatequit (Client Quit)
05:00:14  * perezdjoined
05:06:56  * wavdedquit (Quit: WeeChat 0.4.0)
05:28:17  * lohkeyquit (Quit: lohkey)
05:30:48  * TheJHjoined
05:37:27  * bradleymeckjoined
05:46:47  * lohkeyjoined
05:51:58  * wolfeidauquit (Remote host closed the connection)
05:57:49  * TheJHquit (Ping timeout: 246 seconds)
05:59:24  * mikealquit (Read error: Connection reset by peer)
06:00:21  * mikealjoined
06:16:08  * loladiroquit (Quit: loladiro)
06:33:22  * loladirojoined
06:50:47  * TheJHjoined
06:52:47  * loladiroquit (Quit: loladiro)
06:55:03  * brson_joined
06:55:08  * brsonquit (Ping timeout: 255 seconds)
06:56:04  * bradleymeckquit (Quit: bradleymeck)
07:00:17  * trevnorrisjoined
07:14:17  * loladirojoined
07:17:15  * wolfeidaujoined
07:28:22  * lohkeyquit (Quit: lohkey)
07:36:00  * `3rdEdenjoined
07:52:55  * rendarjoined
08:16:49  * loladiroquit (Quit: loladiro)
08:23:07  * brson_quit (Quit: leaving)
08:25:07  * stagasjoined
08:30:35  <indutny>isaacs: huh
08:30:38  <indutny>interesting
08:30:50  <indutny>well, it might be related to one bug that I've seen
08:31:29  <indutny>./test-child-process-fork-net2 is failing sporadically
08:31:37  <indutny>and it's not related to my patch
08:31:42  <indutny>isaacs: are you testing it on osx?
08:31:49  <indutny>if so - can you try reproducing it on ubuntu
08:31:53  <indutny>or smartos
08:36:46  * felixgequit (Quit: http://www.debuggable.com/)
08:43:08  * kazuponquit (Remote host closed the connection)
08:45:15  <trevnorris>indutny: you have mad cpp skills right?
09:29:53  * loladirojoined
09:45:09  * loladiroquit (Quit: loladiro)
10:02:14  * TheJHquit (Ping timeout: 256 seconds)
10:09:49  * hzjoined
10:23:23  * trevnorrisquit (Quit: Leaving)
10:24:48  * c4milojoined
10:28:51  * c4miloquit (Ping timeout: 248 seconds)
10:28:57  * benoitcquit (Excess Flood)
10:29:41  * benoitcjoined
10:38:39  * abraxasquit (Remote host closed the connection)
10:51:21  * kristatequit (Remote host closed the connection)
11:18:46  * qmx|awaychanged nick to qmx
11:35:08  * Benviequit (Ping timeout: 252 seconds)
11:54:12  * qmxchanged nick to qmx|brb
11:54:15  <roxlu>hi, does node.js has a function to retrieve all files from a directory?
11:54:35  <indutny>no, use modules
11:54:39  <indutny>or write it yourself
12:00:52  * loladirojoined
12:06:06  <roxlu>ok
12:06:06  <roxlu>thanks
12:06:45  * qmx|brbchanged nick to qmx
12:07:10  * loladiroquit (Quit: loladiro)
12:10:12  * c4milojoined
12:14:28  * c4miloquit (Ping timeout: 248 seconds)
12:27:03  * bnoordhuisjoined
12:39:49  * abraxasjoined
12:56:11  * piscisaureus_joined
12:58:54  <MI6>joyent/node: Trevor Norris master * 3d286b6 : buffer: remove float write range checks Removed range checks when writin - http://git.io/5qoI4A
12:59:09  * sgallaghjoined
13:07:01  <piscisaureus_>bnoordhuis: +1 voor het landed daarvan
13:07:22  <bnoordhuis>piscisaureus_: landing it is the easy part :)
13:07:40  <piscisaureus_>yeah I know I already did the thinking :-p
13:07:55  <bnoordhuis>so how are you feeling, bertje?
13:08:28  <piscisaureus_>a little better
13:08:35  <piscisaureus_>actually now it's okay
13:08:43  <piscisaureus_>but when I woke up it wasn't great
13:09:22  <bnoordhuis>try a nice cup of mint tea, it helps according to my gf
13:14:07  <indutny>piscisaureus_: you can try following thing
13:14:17  * cjdjoined
13:14:23  <indutny>slice ginger in a thin sheets
13:14:34  <bnoordhuis>seen this? https://github.com/joyent/node/issues/4648#issuecomment-12589238
13:14:34  <indutny>add some honey
13:14:38  <bnoordhuis>i wonder what's up with that
13:14:40  <indutny>and fill it with boiled water
13:14:47  <indutny>helps a lot
13:14:49  <indutny>when sick
13:15:06  <indutny>bnoordhuis: reminding you about bug that I seen in child_process-fork-net2
13:15:18  <bnoordhuis>indutny: what about it?
13:15:18  <indutny>not sure if it's important, though
13:15:21  <indutny>ooh
13:15:27  <indutny>let me find gist
13:15:45  <indutny>https://gist.github.com/04a355abaf96457e4eb0
13:15:52  <indutny>bnoordhuis: can you run it a couple of times on ubuntu?
13:15:59  <indutny>in node.js directory
13:16:11  <indutny>it seems that sometimes sockets are emitting EOF
13:16:16  <bnoordhuis>maybe later today
13:16:22  <bnoordhuis>i'm trying hard not to get distracted
13:16:22  <indutny>ok
13:16:25  <indutny>:)
13:17:11  <piscisaureus_>bnoordhuis: what about "that" ?
13:17:33  <bnoordhuis>piscisaureus_: different nan types depending on the compiler/arch used?
13:17:35  <bnoordhuis>kind of worrisome
13:17:43  <piscisaureus_>bnoordhuis: ah, that. Yes
13:17:56  <piscisaureus_>bnoordhuis: although if you ask me it could just be an llvm "bug"
13:18:17  <bnoordhuis>that sounds likely or at least possible
13:18:28  <bnoordhuis>but i'm checking it out anyway
13:21:12  <bnoordhuis>../src/node.cc:2235:50: warning: adding 'int' to a string does not append to the string [-Wstring-plus-int] String::New(OPENSSL_VERSION_TEXT + i, j - i));
13:21:19  <bnoordhuis>^ that clang
13:21:27  <indutny>hahaha
13:21:29  <indutny>omg
13:21:37  <indutny>it's amazing
13:21:52  <indutny>warning: your code not what you thought it will do
13:22:26  <piscisaureus_>bnoordhuis: good
13:25:45  <piscisaureus_>bnoordhuis: if it turns out to be a bug we could write a numeric_limits<float>::quiet_NaN ourselves
13:26:00  <piscisaureus_>No idea how you would do that in C btw (except hardcoding the representation)
13:26:53  <bnoordhuis>hrm. the test does indeed fail with ia32 clang but not ia32 gcc
13:27:24  <bnoordhuis>AssertionError: 127 == 255 at test (/home/bnoordhuis/src/nodejs/clang/test/simple/test-writedouble.js:171:10)
13:28:25  <piscisaureus_>bnoordhuis: maybe write a very simple test case that does
13:28:25  <piscisaureus_>float f = 0/0;
13:28:25  <piscisaureus_>printf("%x\n", * (int*) &f);
13:28:35  <piscisaureus_>and see if the representations are different in that case as well
13:28:55  <bnoordhuis>yeah, working on something like that as we speak^Wtype
13:34:07  <bnoordhuis>curious, clang and gcc write out the same double for 0/0
13:34:10  <bnoordhuis>but the float is different
13:34:25  <bnoordhuis>00 00 c0 7f vs 00 00 c0 ff
13:37:11  <bnoordhuis>same output with -m64
13:37:42  <bnoordhuis>but curiously enough the writedouble test passes when node is compiled with x64 clang
13:43:17  <bnoordhuis>something else that's curious is that clang generates SSE instructions with -m32
13:47:19  <bnoordhuis>interesting. a hard-coded 0/0 is translated to a sNaN, a programmatic divide-by-zero isn't
13:49:20  * jmar777joined
13:55:01  * c4milojoined
13:59:10  * c4miloquit (Ping timeout: 244 seconds)
14:02:42  * bradleymeckjoined
14:06:36  * qmxchanged nick to qmx|brb
14:07:08  * bradleymeckquit (Client Quit)
14:07:40  * felixgejoined
14:09:22  * kazuponjoined
14:12:16  * TheJHjoined
14:14:49  * kazuponquit (Remote host closed the connection)
14:26:01  * qmx|brbchanged nick to qmx
14:32:57  * c4milojoined
14:34:48  * sgallaghquit (Remote host closed the connection)
14:43:29  * indexzerojoined
14:50:54  * sgallaghjoined
15:00:55  * qmxchanged nick to qmx|mtg
15:01:03  * qmx|mtgchanged nick to qmx
15:02:05  * bradleymeckjoined
15:10:54  <piscisaureus_>bnoordhuis: so it's just random :-)
15:11:34  <piscisaureus_>bnoordhuis: so I'd say:
15:11:34  <piscisaureus_>#ifdef __CLANG__
15:11:34  <piscisaureus_>if (isnan(v))
15:11:34  <piscisaureus_> v = numeric_limits<float>::signaling_NaN()
15:11:47  <piscisaureus_>euh
15:11:55  <piscisaureus_>quiet_NaN() of course :-)
15:12:45  * felixgequit (Quit: http://www.debuggable.com/)
15:14:00  * indexzeroquit (Quit: indexzero)
15:15:34  * indexzerojoined
15:16:02  * kazuponjoined
15:22:49  * c4miloquit (Remote host closed the connection)
15:25:52  * piscisaureus_quit (Read error: Connection reset by peer)
15:26:37  * piscisaureus_joined
15:31:03  * rumpjoined
15:32:22  * AvianFlujoined
15:40:33  * indexzeroquit (Quit: indexzero)
15:41:32  * rumpquit (Quit: rump)
15:52:35  * dapjoined
16:00:41  * c4milojoined
16:02:18  * c4miloquit (Remote host closed the connection)
16:02:44  * c4milojoined
16:03:02  * stagasquit (Quit: ChatZilla 0.9.89-rdmsoft [XULRunner 1.9.0.17/2009122204])
16:07:29  * c4miloquit (Ping timeout: 248 seconds)
16:09:31  * kazuponquit (Remote host closed the connection)
16:11:30  * `3rdEdenchanged nick to `3rdEden|Dinner
16:17:07  * indexzerojoined
16:18:41  * qmxchanged nick to qmx|lunch
16:24:36  * c4milojoined
16:26:10  * `3rdEden|Dinnerquit (Remote host closed the connection)
16:27:47  * kazuponjoined
16:35:41  * pooyajoined
16:36:30  * pooyaquit (Client Quit)
16:39:48  * indexzeroquit (Ping timeout: 256 seconds)
16:47:18  * qmx|lunchchanged nick to qmx
17:05:14  * bradleymeckquit (Quit: bradleymeck)
17:05:57  * bradleymeckjoined
17:06:46  * bradleymeckquit (Client Quit)
17:15:40  * sgallaghquit (Remote host closed the connection)
17:18:59  <bnoordhuis>piscisaureus_: strange fact, clang32 returns the same value for signaling_NaN() and quiet_Nan(), 00 00 c0 ff
17:19:24  <bnoordhuis>but __builtin_nanf and __builtin_nansf return different values
17:21:22  * AvianFluquit (Remote host closed the connection)
17:22:58  * indexzerojoined
17:23:35  * perezdquit (Quit: perezd)
17:25:40  * mikealquit (Quit: Leaving.)
17:29:31  * AvianFlujoined
17:40:54  * piscisaureus_quit (Ping timeout: 264 seconds)
17:44:49  * bradleymeckjoined
17:49:14  * perezdjoined
17:52:49  * philips_quit (Excess Flood)
17:53:27  * loladirojoined
17:53:47  * perezdquit (Client Quit)
17:54:08  * sgallaghjoined
17:58:13  * pooyajoined
18:00:28  * mikealjoined
18:08:43  * philips_joined
18:17:26  * mikealquit (Quit: Leaving.)
18:19:47  * mikealjoined
18:25:28  * stagasjoined
18:34:47  * c4miloquit (Remote host closed the connection)
18:35:20  * `3rdEdenjoined
18:36:44  * TooTallNatejoined
18:40:08  * trevnorrisjoined
18:46:18  <isaacs>indutny: yeah, os x
18:49:27  <trevnorris>i'm trying to figure out what's happening to a variable between passing it to a cc module, and how it arrives as an Argument& variable. any ideas where to look?
18:49:51  * c4milojoined
18:50:13  <indutny>huh?
18:50:16  <indutny>aaah
18:52:18  * V1joined
18:53:17  * `3rdEdenquit (Disconnected by services)
18:53:22  * V1changed nick to `3rdEden
18:55:36  * AvianFluquit (Remote host closed the connection)
18:56:22  * AvianFlujoined
18:59:16  <MI6>joyent/node: Scott Blomquist master * 0774798 : build: Add some gyp artifacts to .gitignore - http://git.io/NtgwJw
19:03:07  * lohkeyjoined
19:03:53  * V1joined
19:06:49  * brsonjoined
19:07:33  * `3rdEdenquit (Disconnected by services)
19:07:39  * V1changed nick to `3rdEden
19:12:57  * AvianFluquit (Remote host closed the connection)
19:13:42  * AvianFlujoined
19:18:58  * kazuponquit (Remote host closed the connection)
19:20:54  * indutny_joined
19:22:52  <trevnorris>isaacs: you authored benchmark/client_latency.js. have a brief explanation of what that out file shows?
19:23:02  <isaacs>hm... looking
19:24:01  * indutnyquit (Read error: Operation timed out)
19:24:08  <isaacs>trevnorris: i think i was trying to track down some kind of issue where the client's latency was stepping up in the wrong way
19:24:25  <isaacs>like, you'd expect to get maxSockets requests going in parallel, right?
19:24:34  <isaacs>and it was not doing that properly
19:24:38  * indutny_quit (Client Quit)
19:25:25  * AvianFluquit (Remote host closed the connection)
19:25:25  <trevnorris>interesting, ok. thanks.
19:25:34  * indutnyjoined
19:26:45  <isaacs>i dn't remember exactly how the numbers work, though
19:26:52  <isaacs>gotta run, i'll be back in a few hours
19:31:40  <trevnorris>son of a b!@#. spent two hours troubleshooting why llvm wouldn't build. forgot I needed to checkout llvm AND clang at the release version.
19:34:27  <bnoordhuis>trevnorris: are you investigating that nan thing?
19:34:53  <trevnorris>bnoordhuis: yeah. clang has a bunch of NaN tests, so I'm going to build it for ia32 and see if they pass.
19:35:15  <bnoordhuis>trevnorris: i kind of know what the issue is (looked into it this afternoon)
19:36:01  <bnoordhuis>but i think we're looking at a compiler implementation detail here
19:36:10  <bnoordhuis>which is not a good thing
19:36:56  <TooTallNate>what are the dangers with renaming the node binary again?
19:37:25  <trevnorris>ok. how'd you determine that? been working on this for hours and feel like I haven't gotten anywhere.
19:37:55  <bnoordhuis>trevnorris: i checked the disassembly of a few functions, traced a test script
19:37:59  <bnoordhuis>wait, let me gist something
19:38:56  <bnoordhuis>trevnorris: https://gist.github.com/0c340629e451ff31ef66
19:39:00  <tjfontaine>clang still tries where possible to emulate gcc, they'd probably want a PR for this if they haven't discussed it before
19:39:22  * indexzeroquit (Quit: indexzero)
19:39:50  <bnoordhuis>the fun thing is that the output of __builtin_nans() and friends is the same in gcc and clang
19:40:01  <bnoordhuis>but not that of std::numeric_limits<T>
19:40:17  * indexzerojoined
19:40:38  <tjfontaine>I presume you're using libc++ vs libstdc++?
19:42:40  <bnoordhuis>no, it's using the system headers and libs
19:43:30  <bnoordhuis>so that suggests the system headers export different std::numeric_limits templates, depending on whether it's gcc or clang parsing them
19:43:46  <bnoordhuis>but that's a side bug really :)
19:44:43  * indexzeroquit (Client Quit)
19:47:05  <tjfontaine>heh
19:47:39  <TooTallNate>bnoordhuis: are you aware of any dangers renaming ./node ?
19:47:51  <TooTallNate>i seem to recall something wrt native addons
19:48:02  <trevnorris>um... ok, still just absorbing that bit of code.
19:48:04  <bnoordhuis>TooTallNate: yes, on windows
19:48:28  <bnoordhuis>but i forgot the details
19:48:44  <TooTallNate>ok that's what i seem to remember as well
19:48:54  <bnoordhuis>.lib files were involved. bert should know more
19:49:43  <trevnorris>bnoordhuis: if that's the case, what's a possible next step?
19:50:04  <bnoordhuis>trevnorris: well... reverting to a js-only solution
19:50:34  <trevnorris>bnoordhuis: so always writing a signaling NaN is better than only writing one with clang for ia32?
19:51:03  <bnoordhuis>trevnorris: it's more subtle than that
19:51:24  <bnoordhuis>gcc always seems to clear the sign bit, clang doesn't
19:51:33  <bnoordhuis>that's what's making the node test fail
19:51:52  <bnoordhuis>that's something you can paper over by checking if it's a nan and masking off the sign bit
19:52:19  <bnoordhuis>but there are probably more subtle incompatibilities
19:52:20  <trevnorris>yeah. so is there the possibility of injecting code on that specific build to do so?
19:53:52  * bradleymeckquit (Quit: bradleymeck)
19:54:02  <trevnorris>bnoordhuis: can you tell me what deps/v8/src/ia32/code-stubs-ia32.cc is for?
19:54:33  <bnoordhuis>trevnorris: yes, quite easily. if bit 51 is clear and bits 52-62 are set, it's a snan
19:54:50  <bnoordhuis>you turn it into a qnan by flipping bit 51
19:55:11  <bnoordhuis>caveat emptor: i may be a bit off (pardon the pun)
19:55:44  <bnoordhuis>wait, let me write a quick patch to show what i mean
19:56:21  <trevnorris>bnoordhuis: not to screw my own patch, but `->NumberValue()` is returning a snan. so does it do any good by that point to flip the bit afterwards?
19:58:15  * bradleymeckjoined
19:58:28  <bnoordhuis>yes/no/maybe
19:58:37  <bnoordhuis>it's harmless if fpu exceptions are disabled
19:58:57  <trevnorris>ok. so flipping the bit would help it pass the test. got it.
19:58:58  <bnoordhuis>but like i said, this qnan/snan thing may not be the only difference
20:06:46  * qmxchanged nick to qmx|coffee
20:08:20  <bnoordhuis>okay, gotta run
20:08:51  <bnoordhuis>trevnorris: i'll probably revert the buffer float optimizations for now
20:09:06  <bnoordhuis>we can maybe reapply them later but not in v0.10
20:09:35  <trevnorris>bnoordhuis: so bit flipping isn't an option?
20:09:52  <bnoordhuis>that just fixes this one thing
20:10:07  <bnoordhuis>but there may be more obscure bugs lurking in there
20:10:16  <bnoordhuis>it needs a little more time to mature, it looks like
20:10:31  <trevnorris>but woudn't that affect all the code, not just this one change?
20:11:08  <bnoordhuis>not until you start serializing floats
20:11:28  <bnoordhuis>(and fpu exceptions are disabled)
20:11:52  <bnoordhuis>okay, i'm off. have a good night everyone
20:14:45  * bnoordhuisquit (Read error: Operation timed out)
20:18:49  * bradleymeckquit (Quit: bradleymeck)
20:19:28  * qmx|coffeechanged nick to qmx
20:20:34  <trevnorris>isaacs: can I use ircretary to leave bnoordhuis a message?
20:23:12  <tjfontaine>trevnorris: ircretary tell <user> <msg>
20:23:22  <tjfontaine>iirc
20:23:53  <trevnorris>tjfontaine: thanks.
20:24:40  <trevnorris>this snan thing is making my eyes bleed.
20:29:43  * kazuponjoined
20:33:19  <TooTallNate>hahaha
20:34:25  <TooTallNate>trevnorris: what if we defined a static buffer with the correct NaN in memory
20:34:44  <TooTallNate>trevnorris: and then when NumberValue() returns and type of NaN, memcpy the correct NaN buffer
20:34:54  * kazuponquit (Ping timeout: 264 seconds)
20:34:54  <trevnorris>ircretary tell bnoordhuis please check out GH-4648 before reverting those changes.
20:34:55  <ircretary>trevnorris: I'll be sure to tell bnoordhuis
20:35:12  <trevnorris>TooTallNate: it's not how any other implementation works. checkout GH-4648
20:36:53  <trevnorris>TooTallNate: I mean, I'm totally cool with that solution. mainly I think this one problem isn't worth reverting the changes.
20:38:59  <trevnorris>TooTallNate: actually that's a really slick solution.
20:39:18  <trevnorris>just don't totally understand when he said about this affecting other code, "not until you start serializing floats (and fpu exceptions are disabled)"
20:39:47  <TooTallNate>trevnorris: i'd imagine memcpy() is slower than *ptr = NaN though
20:40:02  <TooTallNate>trevnorris: i'm not totally sure what he means by that either
20:41:49  <trevnorris>TooTallNate: i'll admit my bias after having worked ~30hrs on that patch, but i'd have expected this to be a known issue had typed arrays been properly checked.
20:42:02  <trevnorris>that's why I used this way of doing it in the first place.
20:42:26  <TooTallNate>trevnorris: i mean your current impl is definitely "correct" code
20:42:39  <TooTallNate>trevnorris: we're just working around compiler bugs at this point, right?
20:43:08  <trevnorris>TooTallNate: yeah. it's a compiler bug. he wrote a test here: https://gist.github.com/0c340629e451ff31ef66
20:43:14  <TooTallNate>right
20:43:21  <TooTallNate>sweet test case :)
20:43:43  <TooTallNate>i love learning his little C wise-man tricks
20:45:03  <trevnorris>seriously. what did he do in a previous life to get that knowledge?
20:45:16  <TooTallNate>trevnorris: i mean we could even only do the memcpy() only when #if __clang__ && __ia32__
20:45:35  <TooTallNate>(those aren't the correct define names :p)
20:46:09  <TooTallNate>trevnorris: well when you've been doing linux kernel programming for xx years, that kind of knowledge builds up
20:46:13  <trevnorris>TooTallNate: heh, know what you mean. but that's what I was thinking.
20:46:48  <trevnorris>really? he does linux kernel programming? intense.
20:47:13  <TooTallNate>not sure about the programming, but i know he frequents the LKML
20:47:19  <TooTallNate>or at least used to
20:49:12  * loladiroquit (Quit: loladiro)
20:54:21  <trevnorris>TooTallNate: and fwiw, llvm's own `make check` for ia32 builds fails all over the place. guess it doesn't get as much love as the x64 side.
20:55:18  <tjfontaine>trevnorris: #llvm on oftc probably has guidance on that front, and with ben's test case you should get feedback
20:55:38  <trevnorris>tjfontaine: good point. I'll hop in there.
20:57:22  * loladirojoined
21:00:07  * sgallaghquit (Remote host closed the connection)
21:16:06  <trevnorris>well, hope bnoordhuis doesn't mind me using his code. had me file a bug report.
21:36:59  <trevnorris>TooTallNate: think I should have left those perf changes out on the slowbuffer copy issue? (in my warped mind getting laterally par performance was an "essential" change)
21:38:51  <TooTallNate>trevnorris: probably should have been separate commits but not a huge deal
21:38:59  <trevnorris>cool
21:39:08  <TooTallNate>trevnorris: i mean it made me have to question it in github comment, so if that's any indication l)
21:39:09  <TooTallNate>;)
21:39:28  <trevnorris>heh, good point.
21:39:47  * jmar777quit (Remote host closed the connection)
21:39:55  <trevnorris>but it was 3x's slower w/o that change! =)
21:40:21  <TooTallNate>it's a good change
21:40:48  <TooTallNate>but another commit message explaining it would answer questions to somebody looking back through the commit log
21:41:31  <trevnorris>ah, yeah. totally forgot to mention that change in the commit log. I'll fix that.
21:59:11  * brson_joined
21:59:15  * wolfeidauquit (Remote host closed the connection)
22:00:01  * brsonquit (Ping timeout: 245 seconds)
22:02:13  <trevnorris>TooTallNate: ok. included note in commit msg. thanks for the reminder.
22:04:34  * rendarquit
22:08:55  * AvianFlujoined
22:09:02  * qmxchanged nick to qmx|away
22:16:52  * TheJHquit (Ping timeout: 248 seconds)
22:23:00  <indutny>ttyl
22:23:03  * indutny&
22:23:04  <LOUDBOT>COMING SOON TO VIDEOCASETTE: NYM WARS 4, TWEETERS REVENGE
22:23:32  <CoverSlide>.
22:37:19  <trevnorris>TooTallNate: ok, for real this time.
22:46:53  * brson_quit (Ping timeout: 255 seconds)
22:47:30  * brsonjoined
22:57:27  * AvianFluquit (Remote host closed the connection)
23:02:10  * c4miloquit (Remote host closed the connection)
23:02:28  * jmar777joined
23:08:46  * chiltsquit (Ping timeout: 256 seconds)
23:09:40  * chiltsjoined
23:09:49  * wolfeidaujoined
23:29:11  * `3rdEdenquit (Remote host closed the connection)
23:29:33  * rumpjoined
23:41:08  * piscisaureus_joined
23:48:26  * hzquit