00:00:01  * ircretaryquit (Remote host closed the connection)
00:00:09  * ircretaryjoined
00:02:39  * c4miloquit (Remote host closed the connection)
00:10:18  * abraxasjoined
00:11:54  * c4milojoined
00:20:59  <MI6>libuv-master: #43 UNSTABLE osx (1/183) linux (2/183) smartos (5/183) http://jenkins.nodejs.org/job/libuv-master/43/
00:29:48  * CoverSlidequit (*.net *.split)
00:30:25  * Raltquit (Ping timeout: 248 seconds)
00:31:04  * Raltjoined
00:35:13  * karupaneruraquit (Excess Flood)
00:37:58  <isaacs>cleaning up tests: https://github.com/isaacs/node/tree/windows-test-fixing <-- tjfontaine
00:38:04  * karupanerurajoined
00:39:10  * CoverSlidejoined
00:45:01  <MI6>joyent/node: Bert Belder msi-fixes * 77218d6 : win/msi: enable modify and repair (+13 more commits) - http://git.io/-LPTTA
00:55:30  <tjfontaine>isaacs: http://jenkins.nodejs.org/job/nodejs-oneoff/lastCompletedBuild/tapTestReport/
00:56:17  * qmx|awaychanged nick to qmx
01:04:19  * pooyajoined
01:06:50  * EhevuTovquit (Quit: This computer has gone to sleep)
01:06:55  * piscisaureus_joined
01:07:13  <piscisaureus_>isaacs: I'm a windows guy
01:07:20  <piscisaureus_>A sleepy one though
01:07:45  <isaacs>piscisaureus_: hey
01:07:52  <isaacs>piscisaureus_: something is weird with symbolic links on windows.
01:07:54  <isaacs>i don't get it.
01:08:02  <isaacs>stuff that i'm reasonably sure used to work is broken in master, AND in v0.8
01:08:12  <piscisaureus_>Ok - many weird things. But what exactly?
01:08:31  <isaacs>getting EPERMs when i link process.execPath to some random filename
01:08:31  <piscisaureus_>isaacs: ^?
01:08:48  <isaacs>is it against the rules to link a file that is in use?
01:08:51  <piscisaureus_>isaacs: are you running in privileged mode?
01:08:51  <isaacs>locking or some such?
01:08:56  <isaacs>um... dunno?
01:08:58  <isaacs>wht's that?
01:09:00  <piscisaureus_>isaacs: no that should be totally okay
01:09:02  <isaacs>oh, you mean, as admin
01:09:04  <isaacs>no
01:09:05  <piscisaureus_>yeah
01:09:10  <piscisaureus_>you have to run as admin
01:09:14  <isaacs>but i'm doing fs.symlink(process.execPath, symlinkPath, 'file')
01:09:17  <isaacs>and it still EPERMs
01:09:33  <isaacs>(sometimes)
01:09:48  <isaacs>other times, it doesn't EPERM, but then trying to spawn() the symlinkPath raises ENOENT
01:09:57  <isaacs>hence: http://jenkins.nodejs.org//job/nodejs-oneoff/lastCompletedBuild//tapTestReport/test.tap-25/
01:11:59  <isaacs>piscisaureus_: https://gist.github.com/5113436
01:12:02  <isaacs>the file is there, for sure
01:12:53  <isaacs>it looks executable:
01:12:53  <isaacs>c:\node>ls -lF test\tmp
01:12:53  <isaacs>total 2808
01:12:53  <isaacs>-rwxr-xr-x 1 Isaac Sc Administ 5749608 Mar 7 16:57 node-symlink*
01:12:56  <piscisaureus_>isaacs: I think you can't spawn a file that doesn't have .exe
01:13:00  <piscisaureus_>lemme try
01:13:02  <isaacs>orly?
01:13:19  <piscisaureus_>you can but you have to spawn 'test-symlink.' and not 'test-symlink'
01:13:22  <isaacs>HAHH!!
01:13:23  <isaacs>indeed
01:13:57  <isaacs>ok. so, i'm not getting the earlier errors (EPERM), so i must assume that it was some oddness in my earlier working dir
01:14:04  <isaacs>a fresh checkout seems to wokr now
01:14:07  <piscisaureus_>well EPERM shouldn't happen
01:14:17  <piscisaureus_>but they can happen if the file already existed I think
01:14:26  <piscisaureus_>maybe that's unlike unix so I
01:14:28  <piscisaureus_>er
01:14:34  <piscisaureus_>*so I'd have to fix that
01:14:48  <isaacs>i see
01:14:59  <isaacs>piscisaureus_: i have a pull req in ot libuv to do EISDIR on fs.read/write
01:15:09  <isaacs>fixes some more tests.
01:17:03  <piscisaureus_>isaacs: yeah the PR lgtm although your commit comment tipped the balance here :-)
01:17:15  <trevnorris>isaacs: well. i thought the reason for pooling memory was to reduce malloc's. little did I realize that persisting an object is like 100x's more expensive.
01:17:21  <piscisaureus_>otherwise I'd have told you to add the mapping to the read and write impl
01:18:37  <isaacs>piscisaureus_: hah. i was kinda on the fence myself.
01:18:44  <piscisaureus_>isaacs: we're also still lacking somewhat-reasonable thread pool scaling. I know ben has been working on that, it'd be nice to get that into 0.10 as to not regress too much over libeio
01:18:49  <isaacs>piscisaureus_: if you'd prefer putting it in read/write directly, then that's fine with me.
01:19:01  <piscisaureus_>isaacs: no. Your comment convinced me here.
01:19:24  <isaacs>piscisaureus_: wait, convinced which way?
01:19:30  <isaacs>piscisaureus_: to put in read/write, or to leave it where it is?
01:19:31  <trevnorris>isaacs: on vacation all next week with the in-laws
01:19:35  <trevnorris> i'll be on when I can (and usually to escape from everyone)
01:19:36  <isaacs>trevnorris: kewl.
01:19:37  <piscisaureus_>If the *only* causes of this error are outlined in your commit then I'm 100% fine with this. I'm a little scared that there might be other reasons though.
01:19:42  <piscisaureus_>But we'll find out the hard way
01:19:49  <isaacs>piscisaureus_: rigt.
01:19:49  <piscisaureus_>isaacs: leave it as-is
01:20:00  <isaacs>piscisaureus_: what'll happen is that someone will complain about getting an EISDIR from reading a usb stick or something
01:20:10  <isaacs>piscisaureus_: and we can just split it up then.
01:20:12  <piscisaureus_>well - not in that case probably
01:20:17  * trevnorrisquit (Quit: Leaving)
01:20:27  <isaacs>piscisaureus_: or whatever. some other random oddball case.
01:20:38  <piscisaureus_>more like for setting a socket option that a particular windows flavor doesn't support
01:20:43  <piscisaureus_>but no common cases anyway
01:20:49  <isaacs>right
01:20:54  <isaacs>except read() on a directory :)
01:20:57  <isaacs>which is super extra common
01:21:07  <isaacs>actually, i often use EISDIR to avoid a stat()
01:21:15  <piscisaureus_>ya
01:21:37  <isaacs>hm... changed test/simple/test-child-process-fork-exec-path.js to put a .exe on the symlink path, still getting ENOENT..
01:21:50  <piscisaureus_>hmm
01:22:05  <piscisaureus_>I can take a look tomorrow if you want
01:22:19  <piscisaureus_>I pretty much want to release 0.10 with a clean sheet
01:22:30  <piscisaureus_>that'd be much more preferrable over the 0.8 situation
01:22:31  <isaacs>yeah, that'd be great.
01:22:39  <piscisaureus_>esp. now that we're having CI
01:22:43  <isaacs>but i'd also like to actually release it
01:22:48  <piscisaureus_>yeah
01:22:57  <isaacs>if it's not API changes, we can fix bugs after the .0
01:23:04  <piscisaureus_>although in all honesty - we're really really almost there
01:23:09  <isaacs>yeah
01:23:13  <piscisaureus_>it's not like these 1 or 2 days are going to make a huge difference
01:23:16  <isaacs>there's a few dumb fixes, and i think 2 or three actual bugs
01:23:31  <piscisaureus_>and it would be nice to have 0.10.0 actually be good :)
01:23:44  <isaacs>piscisaureus_: https://github.com/isaacs/node/compare/windows-test-fixing
01:24:02  <piscisaureus_>I am totally fine delaying the .10 release a a bit (< week) for that
01:24:10  <isaacs>piscisaureus_: i'm not
01:24:14  <isaacs>it should go out monday
01:24:19  <isaacs>no one will install it for the first week anyway
01:24:24  <isaacs>but Monday has the biggest internet attention juice
01:25:40  <piscisaureus_>:s
01:25:52  <piscisaureus_>priority inversion
01:27:02  <isaacs>well, there are aspects of our quality that are definitely good right now.
01:27:05  <isaacs>much better than v0.8
01:27:29  <piscisaureus_>isaacs: windows-test-fixing lgtm/harmless
01:27:34  <piscisaureus_>although you should land your uv patch first
01:27:59  <isaacs>yeah
01:28:05  <isaacs>piscisaureus_: that lgty? i'll push it now
01:28:10  <piscisaureus_>isaacs: yea
01:28:24  <piscisaureus_>https://github.com/joyent/node/compare/msi-fixes <-- that needs to be reviewed by sblom and go into 0.10 as well
01:28:25  <isaacs>done
01:28:28  <MI6>joyent/libuv: isaacs master * b68ee40 : win: Map ERROR_INVALID_FUNCTION to EISDIR This error is raised when call - http://git.io/OsYE9A
01:28:40  <piscisaureus_>it fixes most of the problems I have with the MSI except for being unclear about install scope
01:28:53  <piscisaureus_>but I will push that after the release
01:29:17  <isaacs>kewl
01:30:38  <piscisaureus_>isaacs: which test had the symlink problem again?
01:30:45  <piscisaureus_>I can take a look when I'm wide awake again
01:31:27  <MI6>libuv-master: #44 UNSTABLE osx (10/183) linux (13/183) smartos (11/183) http://jenkins.nodejs.org/job/libuv-master/44/
01:31:44  <isaacs>piscisaureus_: test/simple/test-child-process-fork-exec-path.js
01:31:50  <piscisaureus_>ok, noted
01:31:56  <piscisaureus_>will check back tomorrow
01:31:58  <isaacs>piscisaureus_: i'm heading out as well.
01:32:13  <piscisaureus_>isaacs: wut already you lazy bastard
01:32:13  <piscisaureus_>:-p
01:32:17  <isaacs>piscisaureus_: we can fix all the windows tests tomorrow :)
01:32:30  <piscisaureus_>well that'd be great
01:32:41  <isaacs>g'nite :)
01:32:43  <piscisaureus_>it will work iff we don't find any real bugs
01:32:44  <piscisaureus_>yeah
01:32:49  <piscisaureus_>you too. Enjoy your evening
01:32:59  <piscisaureus_>Don't go drinking with ben tho - not good for a man :)
01:37:32  * piscisaureus_quit (Ping timeout: 255 seconds)
01:46:06  * mikealquit (Quit: Leaving.)
01:48:32  * c4miloquit (Remote host closed the connection)
01:49:56  * dapquit (Quit: Leaving.)
01:50:13  * dapjoined
01:50:13  * dapquit (Client Quit)
01:52:44  * pooyaquit (Quit: pooya)
02:01:29  * loladiroquit (Quit: loladiro)
02:02:09  * pooyajoined
02:12:21  * pooyaquit (Quit: pooya)
02:12:24  * mjr_quit (Quit: mjr_)
02:16:27  * mikealjoined
02:17:44  <MI6>joyent/node: isaacs master * 4b47bb4 : uv: Upgrade to b68ee40 - http://git.io/_5MiVg
02:25:26  * mikealquit (Ping timeout: 272 seconds)
02:32:35  <MI6>nodejs-master: #81 UNSTABLE windows-x64 (11/552) windows-ia32 (11/552) http://jenkins.nodejs.org/job/nodejs-master/81/
02:40:10  * kazuponjoined
02:49:56  * TooTallNatequit (Quit: Computer has gone to sleep.)
02:52:51  * mikealjoined
02:57:09  * mikealquit (Ping timeout: 240 seconds)
03:16:53  * c4milojoined
03:25:35  * kazuponquit (Remote host closed the connection)
03:26:20  * kazuponjoined
03:31:44  * brsonquit (Quit: leaving)
03:39:04  * jguerrerojoined
03:49:43  * TooTallNatejoined
03:56:40  * kazuponquit (Remote host closed the connection)
04:18:26  * philips_quit (Excess Flood)
04:18:56  * abraxasquit (Remote host closed the connection)
04:27:03  * kazuponjoined
04:27:34  * philips_joined
04:35:02  * TooTallNatequit (Quit: ["Textual IRC Client: www.textualapp.com"])
04:36:16  * kazuponquit (Ping timeout: 245 seconds)
04:45:34  * brsonjoined
04:46:00  * loladirojoined
04:50:06  * qmxchanged nick to qmx|away
04:53:29  * brsonquit (Quit: leaving)
04:53:41  * perezdjoined
05:00:04  * perezd_joined
05:02:17  * perezdquit (Ping timeout: 255 seconds)
05:02:18  * perezd_changed nick to perezd
05:05:07  * dominictarrjoined
05:06:57  * c4miloquit (Remote host closed the connection)
05:11:35  * abraxasjoined
05:16:55  * dominictarrquit (Quit: dominictarr)
05:32:58  * kazuponjoined
05:37:38  * kazuponquit (Ping timeout: 252 seconds)
05:51:36  * kazuponjoined
05:54:39  * kazuponquit (Read error: Connection reset by peer)
05:54:44  * kazupon_joined
06:00:12  * kazupon_quit (Ping timeout: 264 seconds)
06:05:38  * wolfeidauquit (Remote host closed the connection)
06:06:31  * bradleymeckjoined
06:21:39  * indexzerojoined
06:25:23  * kazuponjoined
06:25:44  * wolfeidaujoined
06:36:32  * perezdquit (Quit: perezd)
06:48:54  * mikealjoined
07:20:30  * benoitcquit (Excess Flood)
07:24:02  * pooyajoined
07:24:45  * benoitcjoined
07:26:46  * jguerreroquit (Quit: jguerrero)
07:28:30  * `3rdEdenjoined
07:40:45  * kazuponquit (Remote host closed the connection)
07:44:58  * bnoordhuisjoined
07:47:47  * pooyaquit (Quit: pooya)
07:51:20  * rendarjoined
08:06:35  * zot1joined
08:08:26  * zot1quit (Client Quit)
08:09:56  * kazuponjoined
08:09:56  * stagasquit (Quit: ChatZilla 0.9.90-rdmsoft [XULRunner])
08:15:02  * stagasjoined
08:18:44  * Chip_Zeroquit (Changing host)
08:18:44  * Chip_Zerojoined
09:09:41  * benoitcquit (Excess Flood)
09:10:13  * kazuponquit (Remote host closed the connection)
09:15:16  * benoitcjoined
09:43:05  * AvianFluquit (Remote host closed the connection)
09:55:27  * dominictarrjoined
09:59:52  * dominictarrquit (Client Quit)
10:20:42  * kazuponjoined
10:21:19  * kazuponquit (Read error: Connection reset by peer)
10:21:38  * kazuponjoined
10:24:20  <bnoordhuis>okay, i'm off the woods. call the police if i'm not back by tuesday
10:24:34  <bnoordhuis>'off to' rather :)
10:26:49  <rendar>lol :)
10:31:30  * bnoordhuisquit (Ping timeout: 260 seconds)
10:35:54  * piscisaureus_joined
10:49:49  * indexzeroquit (Quit: indexzero)
11:31:57  * abraxasquit (Remote host closed the connection)
11:34:05  * Raltquit (Remote host closed the connection)
11:37:54  * bradleymeckquit (Quit: bradleymeck)
11:43:15  * benoitcquit (Excess Flood)
11:44:12  <MI6>joyent/node: Bert Belder v0.8 * d5959c5 : Revert "build, windows: disable SEH" This is no longer necessary - the u (+2 more commits) - http://git.io/XIqJPQ
11:44:46  * benoitcjoined
11:55:11  <MI6>nodejs-v0.8: #32 FAILURE linux-ia32 (2/468) osx-ia32 (1/468) linux-x64 (2/468) osx-x64 (2/468) smartos-x64 (2/468) smartos-ia32 (2/468) http://jenkins.nodejs.org/job/nodejs-v0.8/32/
11:56:04  * abraxasjoined
12:00:16  <MI6>joyent/libuv: isaacs v0.8 * 1258d9e : win: map ERROR_INVALID_FUNCTION to EISDIR This error is raised when call (+4 more commits) - http://git.io/LPDoEA
12:02:07  <MI6>libuv-v0.8: #18 UNSTABLE osx (1/158) smartos (6/158) linux (1/158) http://jenkins.nodejs.org/job/libuv-v0.8/18/
12:06:42  * milanijoined
12:12:10  * abraxasquit (Remote host closed the connection)
12:14:32  <milani>how does test-embed work? I want to run chromium event loop inside libuv
12:24:58  <indutny>not that way
12:25:05  <indutny>you run chromium event loop as it is
12:25:13  <indutny>use uv_wait() in other thread
12:25:42  <indutny>and then send signals to chromium's event loop
12:25:47  <indutny>and running uv_run_once() in it
12:27:27  <dostoyevsky>Is it normal to stop/start event loops rapidly when using an asynchronous library in a syncronized fashion? I want to use a libuv-based library I wrote for a syncronous api...
12:29:04  * loladiroquit (Quit: loladiro)
12:46:17  <milani>indutny, what if I want to run both loops in a single thread? running them in two different threads makes many difficulties
12:46:34  <indutny>sorry, I didn't explained it fully
12:46:43  <indutny>you'll run all libuv callbacks in chromium's thread
12:46:49  <indutny>and wait for events to happen in another thread
12:46:52  <indutny>so uv_wait() in another thread
12:46:58  <indutny>uv_run_once() in chromium's thread
12:47:32  <milani>ok. I got it. so I run chromium thread and then in each iteration I call uv_run_once
12:48:08  * benoitcquit (Excess Flood)
12:49:19  * benoitcjoined
12:54:17  <indutny>yes and no
12:54:34  <indutny>you should call uv_run_once() only when uv_wait() returns in another thread
12:59:19  * loladirojoined
13:08:01  * loladiroquit (Quit: loladiro)
13:08:08  * qmx|awaychanged nick to qmx
13:09:11  * kazuponquit (Remote host closed the connection)
13:14:38  * sgallaghjoined
13:17:13  * loladirojoined
13:26:17  <milani>indutny, uv_sem_wait waits until uv's event queue become empty?
13:26:35  <milani>and then we can switch back to main thread and run another loop which blocks.
13:27:08  <milani>and we run uv_run_once in the other loop.
13:27:11  <milani>is it right?
13:31:30  * kazuponjoined
13:48:28  * benoitcquit (Excess Flood)
13:49:20  * benoitcjoined
13:56:09  * stagasquit (Ping timeout: 240 seconds)
13:59:39  <dostoyevsky>For me uv_run_once doesn't seem to work for executing events right after pushing them... somehow it doesn't do anything...
14:00:09  * stagasjoined
14:02:07  * zot1joined
14:02:23  * zot1part
14:05:06  * stagas_joined
14:07:31  * stagasquit (Ping timeout: 260 seconds)
14:07:46  * stagas_changed nick to stagas
14:10:54  * benoitcquit (Excess Flood)
14:14:00  * hzjoined
14:16:20  * benoitcjoined
14:17:55  * milaniquit (Ping timeout: 276 seconds)
14:33:39  * piscisaureus_quit (Quit: ~ Trillian Astra - www.trillian.im ~)
14:39:21  * hzquit (Disconnected by services)
14:39:25  * hzjoined
14:51:15  * kazuponquit (Remote host closed the connection)
14:54:52  * philips_quit (Excess Flood)
15:04:06  * philips_joined
15:07:49  * c4milojoined
15:11:28  * kazuponjoined
15:12:17  * jguerrerojoined
15:37:19  * mikealquit (Quit: Leaving.)
15:37:43  * mikealjoined
15:37:58  * mikealquit (Client Quit)
15:38:01  * piscisaureus_joined
15:41:10  * loladiroquit (Quit: loladiro)
16:00:20  <MI6>joyent/node: isaacs master * 9036877 : stream: Emit error on stream object, not global Apparently this function - http://git.io/77OCMA
16:03:58  <piscisaureus_>damm
16:04:59  <piscisaureus_>global module loading actually does not work on windows - because NPM installs to a path that's not on the default search path :-/
16:05:04  <piscisaureus_>another day
16:06:42  <isaacs>good morning
16:06:50  <isaacs>piscisaureus_: please read `npm faq`
16:06:56  <isaacs>piscisaureus_: do not install modules -g for require()
16:06:59  <isaacs>that's not what -g is for
16:07:03  <isaacs>-g is for cli programs
16:09:40  <isaacs>piscisaureus_: i can't figure out what is going on with that child-process-fork-exec test
16:10:01  <isaacs>piscisaureus_: becuase now it totally works for me
16:10:03  <isaacs>but it was failing a lot yesterday
16:10:20  <piscisaureus_>isaacs: hah!
16:10:22  <piscisaureus_>I don't know
16:10:35  <piscisaureus_>isaacs: but did you see my email? The test shouldn't pass
16:11:11  * mikealjoined
16:11:22  <piscisaureus_>isaacs: and really if you want it to pass I'm going to be really sad
16:15:44  <MI6>nodejs-master: #82 UNSTABLE linux-ia32 (1/552) osx-x64 (1/552) windows-x64 (13/552) windows-ia32 (11/552) http://jenkins.nodejs.org/job/nodejs-master/82/
16:19:09  <isaacs>piscisaureus_: oh, it shouldn't pass?
16:19:56  <isaacs>piscisaureus_: oh, i set the symlinkPath to '.exe'
16:20:48  <isaacs>piscisaureus_: ie, this makes it psas for me: https://gist.github.com/isaacs/5117638
16:21:08  <isaacs>piscisaureus_: also, no, we cannot change argv[0]/execPath/etc.
16:21:21  <isaacs>piscisaureus_: there are fiddly little things with homebrew, i'd rather just not open that can of worms
16:24:07  <isaacs>oh!!!
16:24:12  <isaacs>piscisaureus_: i'm running fork-exec-argv
16:24:15  <isaacs>piscisaureus_: not fork-exec-path
16:24:42  * `3rdEdenquit (Remote host closed the connection)
16:25:03  <isaacs>yeah, it fails
16:26:12  * bradleymeckjoined
16:26:18  <isaacs>i'm gonna just skip that test on windows.
16:26:33  <isaacs>or, we can replace the symlink with a copy
16:26:42  <isaacs>since it's not really supposed to be testing symlinking *anyway*
16:29:16  <isaacs>yep, that works.
16:30:19  <isaacs>AND THE WERE WERE SEVEN!
16:30:26  <isaacs>AND THERE WERE SEVEN!
16:30:35  <isaacs>^ exactly
16:31:14  <isaacs>LOUDBOT: whosaid
16:31:14  <LOUDBOT>tarheelcoxn in #trilug on freenode
16:39:22  * dapjoined
16:41:44  <isaacs>piscisaureus_: so, the major errors right now seem to be: 1) Assertion failed: handle->reqs_pending > 0, file src\win\tcp.c, line 1026 2) debugger if fuxxed 3) simple/test-http-curl-chunk-problem is weird, and 4) child proc stderr doesn't flush
16:50:45  <piscisaureus_>isaacs: hmm. the child stderr issue has been there for ages though
16:51:08  * kazuponquit (Remote host closed the connection)
16:51:16  <piscisaureus_>isaacs: 1) is very worrysome and my suspicion is that node is doing something BAAD with libuv
16:51:30  <piscisaureus_>(with all memory corruption / security risks that it brings)
16:51:43  <isaacs>right
16:51:49  <isaacs>piscisaureus_: do you have any time this afternoon to look at that?
16:52:23  <piscisaureus_>isaacs: the afternoon is already over for me. I'm about to head out and visit someone's birtday party
16:53:03  <isaacs>ah.
16:53:05  <isaacs>happy birthday :)
16:53:13  <isaacs>i'll ping scott
16:53:14  <piscisaureus_>isaacs: not my b'day :-)
16:53:17  <piscisaureus_>yeah
16:53:33  <isaacs>well, pass along the b'day well wishes, then
16:53:45  <isaacs>piscisaureus_: failing scott, i'll just learn windows libuv myself, i supose.
16:53:47  <isaacs>bout time
16:53:54  <piscisaureus_>isaacs: I can see if I have time this weekend
16:54:02  <piscisaureus_>isaacs: also +1 for learning libuv-win, good luck :-)
16:54:27  <piscisaureus_>isaacs: I'd probably have time monday morning as well (so that'd be sunday night for you)
16:55:02  <piscisaureus_>isaacs: i need scott to review msi-fixes for me :)
16:55:05  <isaacs>kewl
16:55:27  <isaacs>piscisaureus_: if he doesn't show up, i'll just make sure that htey work, and if they seem to DTRT, land them
16:55:36  <isaacs>had we can make whatever changes we need to later.
16:55:47  <isaacs>it's silly to hodl up a release for an installer that isn't 100% perfect.
16:55:55  <piscisaureus_>yeah
16:56:00  <piscisaureus_>well heh :)
16:56:10  <isaacs>as long as it works, that's fine
16:56:16  <piscisaureus_>it's the so-called first impression
16:56:17  <isaacs>it's already an improvement over 0.8
16:56:19  <isaacs>meh
16:56:29  <isaacs>it took 5 stable releases before anyone started even trying 0.8
16:56:40  <isaacs>i'm totally cynical about the whole "first impression" thing
16:56:46  <isaacs>there's a first impression every day
16:56:49  <isaacs>.0 doesn't matter that much
16:56:55  <isaacs>what matters is that we are continually improving
16:57:34  <isaacs>piscisaureus_: [04:28|% 100|+ 534|- 4]: Done
16:57:37  <isaacs>piscisaureus_: windows ^
16:59:02  <piscisaureus_>good
16:59:09  <piscisaureus_>how are the other OS looking?
17:01:01  <MI6>joyent/node: Bert Belder msi-fixes * 3c22c42 : win/msi: enable modify and repair (+13 more commits) - http://git.io/EnwzVg
17:02:01  <tjfontaine>whihc one?
17:02:04  <tjfontaine>*which
17:02:28  <piscisaureus_>tjfontaine: os x, smartos, linux?
17:02:36  <piscisaureus_>are they all down to 0 failures?
17:02:44  <tjfontaine>have you met our jenkins? :)
17:03:06  <piscisaureus_>Where do I go?
17:03:09  <tjfontaine>http://jenkins.nodejs.org/job/nodejs-master/lastCompletedBuild/testReport/
17:03:12  <piscisaureus_>I can only find it when it posts stuff here
17:03:39  <piscisaureus_>isaacs: so your jenkins says there
17:03:45  <piscisaureus_>'s 11 and 13 failures
17:03:58  <piscisaureus_>tjfontaine: here it would be nice if jenkins could run all the tests more often
17:04:05  <piscisaureus_>instead of just "simple message"
17:04:32  <tjfontaine>daily isn't enough?
17:04:43  <piscisaureus_>hmm
17:04:44  <tjfontaine>http://jenkins.nodejs.org/job/nodejs-master-fnt/lastCompletedBuild/testReport/
17:04:45  <piscisaureus_>no
17:04:52  <tjfontaine>which tests do you care the most about?
17:05:00  <tjfontaine>or is your answer all of them
17:05:06  <piscisaureus_>my answer is all
17:05:17  <piscisaureus_>although internet tests are flaky because they require ipv6 connectivity etc
17:05:34  <tjfontaine>ya, I don't think the JPC has IPv6, at least not that I've seen yet
17:05:42  <piscisaureus_>so not necessarily flaky - but they are hard to get passing
17:05:48  <tjfontaine>I suppose I could spin a tunnel up on each slave
17:06:39  <tjfontaine>as far as jenkins windows showing 13, I don't think isaac has pushed all the fixes?
17:06:56  <piscisaureus_>it would be nice if jenkins had a convenient link to a "test failure report"
17:06:58  <piscisaureus_>of some sorts
17:07:04  <tjfontaine>which would say?
17:07:09  * bradleymeckquit (Quit: bradleymeck)
17:07:29  <piscisaureus_>terminal output of all failed tests
17:07:33  <piscisaureus_>on a single page?
17:07:58  <tjfontaine>I'm not sure that can happen, but the plugin is supposed to be showing failures for configurations on that testReport page
17:08:08  <tjfontaine>and then when you click the test you see the output
17:08:44  <tjfontaine>but I have to see if the bleeding edge of that plugin has it fixed
17:08:58  <piscisaureus_>yeah but the test name is just tap-xyz so it's hard to tell which one is which at first glance
17:09:25  <piscisaureus_>so I was looking at http://jenkins.nodejs.org/job/nodejs-master-fnt/lastCompletedBuild/DESTCPU=ia32,label=smartos/tapTestReport/ for example
17:09:50  <tjfontaine>ya, most of those are gc tests I think
17:10:02  <piscisaureus_>and to get the picture (e.g. somehow the weakref build failed, making all gc tests fail) I have to click all the links
17:10:06  <tjfontaine>all failing because the buildsystem didn't build weakref
17:10:26  <tjfontaine>http://jenkins.nodejs.org/job/nodejs-master-fnt/lastCompletedBuild/DESTCPU=ia32,label=smartos/console
17:10:32  <tjfontaine>you can see the output there as well
17:10:50  * trevnorrisjoined
17:12:49  * qmxchanged nick to qmx|lunch
17:14:29  * pooyajoined
17:15:04  * mikealquit (Quit: Leaving.)
17:21:32  * kazuponjoined
17:27:23  <tjfontaine>piscisaureus_: http://jenkins.nodejs.org/job/nodejs-oneoff/lastCompletedBuild/tapTestReport/ master with isaacs' fixes, 5 failing tests on windows
17:31:37  * kazuponquit (Ping timeout: 276 seconds)
17:35:26  <tjfontaine>piscisaureus_: I just hacked the plugin to display the description/testname instead of its number
17:44:35  <piscisaureus_>tjfontaine: nice!! +1
17:45:12  <isaacs>piscisaureus_: teh 4 failures are on my windows-test-fixing branch
17:45:34  <piscisaureus_>ahja
17:46:16  * hzquit
17:46:29  <tjfontaine>fixing the default testreport page is oddly non-obvious atm
17:46:34  <isaacs>piscisaureus_: as far as i'm concerned that branch is good to merge whenever
17:46:48  * hzjoined
17:47:03  <MI6>joyent/node: Bert Belder master * 3c22c42 : win/msi: enable modify and repair (+13 more commits) - http://git.io/G5Lf_A
17:48:13  <tjfontaine>piscisaureus_: in return maybe a peek at https://github.com/joyent/libuv/pull/735 :)
17:48:38  <isaacs>my suspicion is that evetnually, tjfontaine will either be a jenkins master, or will have written a jenkins competitor in node.
17:48:55  <isaacs>perhpas both
17:48:58  <tjfontaine>ya
17:49:24  <tjfontaine>get jenkins now, make something better later
17:53:01  <piscisaureus_>tjfontaine: My suspicion is that that code will truncate data
17:53:41  <tjfontaine>ok, at the end?
17:54:16  <piscisaureus_>tjfontaine: Suppose you read this
17:54:16  <piscisaureus_>chunk 1: "bla \nbla \n bla\n this part gets truncated"
17:54:16  <piscisaureus_>chunk 2: "foo \nbar \nbaz and this part actually gets printed"
17:54:26  * sblomjoined
17:54:41  <tjfontaine>hmm I suppose yes
17:54:43  <piscisaureus_>tjfontaine: the stuff between the last \n and the end of a chunk gets discarded
17:54:49  <piscisaureus_>unless it's the last chunk
17:55:00  <piscisaureus_>tjfontaine: to test, make `buf` very small
17:55:01  <tjfontaine>right
17:55:02  <piscisaureus_>and you'll see it
17:55:12  <tjfontaine>I get your point, I can add the flush
17:57:40  * c4miloquit (Remote host closed the connection)
18:00:40  <tjfontaine>piscisaureus_: also
18:00:44  <tjfontaine>whoops
18:00:52  <piscisaureus_>tjfontaine: whoop whoop?
18:00:59  <tjfontaine>piscisaureus_: also I will fix the extra # it adds ever chunk :)
18:01:05  <tjfontaine>*every
18:01:07  <piscisaureus_>nice :)
18:01:11  <tjfontaine>today is not going to be a good day.
18:01:26  <piscisaureus_>tjfontaine: yesterday I had the algorithm almost flushed out
18:01:27  <tjfontaine>started this morning by trying to put my glasses on after I had already put my contacts in
18:02:15  * c4milojoined
18:02:22  <MI6>nodejs-master: #83 UNSTABLE windows-x64 (11/552) windows-ia32 (11/552) http://jenkins.nodejs.org/job/nodejs-master/83/
18:03:06  <tjfontaine>oh btw piscisaureus_ there is this view as well: http://jenkins.nodejs.org/job/nodejs-master/83/DESTCPU=ia32,label=windows/tapResults/?
18:03:52  <tjfontaine>not quite only failures and stdout
18:04:09  <tjfontaine>coudl probably do a quick userscript/greasemonkey script to filter them though
18:05:33  * `3rdEdenjoined
18:06:22  <trevnorris>isaacs: to start i've replicated current method of buffer pooling in cc. ~3x's faster.
18:07:07  * hzquit
18:08:15  * hzjoined
18:08:36  * piscisaureus_quit (Ping timeout: 264 seconds)
18:09:22  <isaacs>trevnorris: sounds interesting.
18:09:42  <isaacs>sblom: g'morning!
18:10:19  <trevnorris>isaacs: the basic idea is to replace the entire slowbuffer/fastbuffer thing and just make a single call for allocation. the rest is handled internally.
18:11:14  <isaacs>trevnorris: sounds neat.
18:11:32  <isaacs>trevnorris: i think that this will be a good thing to do in 0.12
18:12:20  <sblom>isaacs: howdy
18:13:33  <MI6>joyent/node: isaacs created branch windows-test-fixing - http://git.io/tYdJmA
18:13:35  <isaacs>sblom: ^
18:14:10  <isaacs>sblom: the sticky wickets right now are that assertion error, the debugger client, and the fact that stdio is async on windows.
18:14:39  <isaacs>sblom: the stdio problem has been with us since 0.6, so if it's to hard to fix, we can just say screw it, and skip those tests on windows
18:15:58  * benoitcquit (Excess Flood)
18:18:29  <sblom>isaacs: Okay. I'll stay focused on the test-cluster-net-send assertion.
18:19:35  * mikealjoined
18:19:46  <isaacs>sblom: awesome. lmk when you find anything
18:23:20  * benoitcjoined
18:25:04  * qmx|lunchchanged nick to qmx
18:26:59  * AvianFlujoined
18:35:47  * c4miloquit (Remote host closed the connection)
18:36:36  * kazuponjoined
18:40:20  * TooTallNatejoined
18:41:12  * kazuponquit (Ping timeout: 252 seconds)
18:43:58  * AvianFluquit
18:44:45  * AvianFlujoined
18:45:54  <indutny>hoya
18:45:56  <indutny>I'm on windows
18:45:58  <indutny>finally....
18:46:16  <indutny>and I fell your sorrow
18:46:29  * tjfontaineconsoles indutny
18:46:54  * brsonjoined
18:47:30  <TooTallNate>indutny: i'm with you
18:47:30  <trevnorris>anyone on the cc side mind looking at this and letting me know if anything screams out "you're going to die!" https://gist.github.com/trevnorris/5118651
18:47:40  <TooTallNate>well not at the moment, but later today, planning on it :p
18:48:18  <indutny>persistent init
18:48:33  <indutny>better do it in function
18:48:41  <indutny>if (x.IsEmpty()) { init x }
18:49:12  * mikealquit (Quit: Leaving.)
18:49:52  <indutny>otherwise looks good
18:50:16  <trevnorris>indutny: awesome. thanks.
18:52:45  <indutny>the only one thing I want to say about windows so far
18:52:48  <indutny>fonts are terrible
18:52:55  <indutny>and their render is even more terrible
18:56:36  * bradleymeckjoined
19:07:00  <indutny>whoa
19:07:02  <indutny>back to osx
19:11:59  <trevnorris>how do I add something to process.binding?
19:12:21  <indutny>dig :)
19:12:26  <indutny>and you'll find out
19:12:37  <indutny>start by searching for some properties that come from C++
19:20:47  <trevnorris>;-)
19:20:51  * benoitcquit (Excess Flood)
19:20:54  * piscisaureus_joined
19:21:32  * mikealjoined
19:21:51  * benoitcjoined
19:25:42  * piscisaureus_quit (Read error: Connection reset by peer)
19:26:23  * c4milojoined
19:26:47  * piscisaureus_joined
19:30:04  * piscisaureus__joined
19:30:06  * mikealquit (Ping timeout: 250 seconds)
19:33:08  * piscisaureus_quit (Ping timeout: 240 seconds)
19:33:55  * AvianFluquit (Remote host closed the connection)
19:35:23  * AvianFlujoined
19:36:35  <isaacs>internet came back, hooray!
19:36:40  <isaacs>just in time to head out to yoga class.
19:36:42  <tjfontaine>welcome home!
19:36:52  <isaacs>trevnorris: you're probably the best person to review this at this point: https://github.com/joyent/node/pull/4963
19:37:27  * kazuponjoined
19:40:12  <trevnorris>isaacs: wtf? i never considered nested domains.
19:40:43  <indutny>haha
19:40:56  <indutny>isaacs: you missed all fun
19:41:01  * EhevuTovjoined
19:41:02  <indutny>now I've windows on my mac
19:41:08  <isaacs>indutny: awesome!
19:41:14  <indutny>and I won't ever fix this stupid unix issues
19:41:23  <isaacs>indutny: push some fixes to the windows-test-fixing branch!
19:41:27  <indutny>that was a tough decision, but I don't rerget it
19:41:33  <indutny>haha :)
19:41:35  <indutny>ok
19:41:35  <isaacs>indutny: it's the enterprise choice.
19:41:42  <indutny>I'll look into it
19:41:49  <isaacs>awesome!
19:41:51  <indutny>right now I'm too tired of terrible fonts to continue running it
19:42:00  <indutny>so I'll try finding some time on weekend
19:42:06  <isaacs>0.10 is going out monday regardless. if we fix windows, we can live withotu the sour stain on our pride, though.
19:42:08  * kazuponquit (Ping timeout: 240 seconds)
19:42:11  <isaacs>:)
19:42:17  <indutny>pride
19:42:17  * isaacstrying to guilt piscisaureus__ into helping..
19:42:21  <indutny>like lion's pride?
19:48:10  <trevnorris>isaacs: lgtm
19:49:18  * isaacs&
19:56:35  <sblom>First discovery on the test-cluster-net-send investigation: I'm seeing what looks like an uninitialized req in uv_process_reqs. (pretty much everything == 0xfeeefeee)
19:57:33  <sblom>I'll add some logging to the test case to see at what point this arises.
19:58:06  * mikealjoined
19:58:51  <sblom>Seems like a bit of a race--connecting the debugger reduces it to about 1 in 5 times.
19:59:07  <sblom>(Versus essentially 100% repro without the debugger attached.)
19:59:46  <tjfontaine>interesting
20:00:03  * dap1joined
20:00:04  * dapquit (Read error: Connection reset by peer)
20:00:25  * EhevuTovquit (Quit: This computer has gone to sleep)
20:01:25  <sblom>This is looking like it's going to be one of those bugs that makes me want to go raise goats in the Alps.
20:01:30  * EhevuTovjoined
20:01:43  <tjfontaine>that sounds like fun
20:01:48  <tjfontaine>the goats I mean
20:02:04  <sblom>Only problem is that I'd have to learn to yodel...
20:02:28  <tjfontaine>nah, I presume that's one of those things you get organically by almost falling off a ledge
20:02:57  * benoitcquit (Excess Flood)
20:08:44  * piscisaureus__quit (Ping timeout: 245 seconds)
20:10:08  * Kakerajoined
20:10:35  <Kakera>are there any Windows IDEs besides VS that can browse and compile libuv?
20:10:52  <Kakera>code::blocks can open sln files, but the compilation fails because it references .lib files
20:10:52  * benoitcjoined
20:10:55  <Kakera>which GCC doesn't support
20:11:23  <tjfontaine>you would still need to use cl and link in your buildsystem
20:11:31  <tjfontaine>but that being said
20:12:23  <tjfontaine>it does gyp -> vsprojfiles -> msbuild, all you really need to install are the compilers and msbuild
20:12:39  <tjfontaine>then use an IDE you want, and have it regenerate the files, and trigger msbuild
20:14:18  <Kakera>my project uses many C++11 features which VS has poor support for so I don't want to use msbuild at all
20:15:02  <trevnorris>have you tried using clang on windows?
20:15:10  <tjfontaine>afaik there's not a compiler toolchain that's not from msft that allows you to link against system built dlls
20:15:28  <tjfontaine>trevnorris: that's just starting to allow linking against dlls, nowhere near production
20:16:12  <trevnorris>heh, haven't touched windows in years.
20:16:12  <tjfontaine>I dunno maybe mingw can
20:16:41  <tjfontaine>Kakera: there is some mingw support in the makefile
20:17:45  * `3rdEdenquit (Remote host closed the connection)
20:18:00  <Kakera>I'm not aware of an IDE that can open a Makefile like an .sln file, displaying separate projects and all
20:19:13  <indutny>I'm afraid its impossible
20:19:28  <indutny>though, it should be almost possible with gyp file
20:19:43  <tjfontaine>so lets get this right, you want a compiler with c++11 features, an ide that can interpret the layout of libuv, and all at the same time not using VS/msbuild etc
20:20:19  <Kakera>yea pretty much
20:20:28  <tjfontaine>you could try gyp -> cmake and QtCreator, but you're pretty much SoL on winders there guy
20:21:03  <tjfontaine>pick your favorite ide and teach gyp how to render project files for it
20:21:49  <indutny>btw, is visual studio express free?
20:21:55  <Kakera>yes
20:22:00  <indutny>ok at least
20:22:07  <indutny>so why not using it?
20:22:14  <indutny>ah
20:22:21  <indutny>you want to use libuv with your project
20:22:23  <tjfontaine>apparently he wants c++11 features that cl doesn't support yet
20:23:23  <sblom>test-cluster-net-send investigation continues: we or v8 are stomping the heap at some point https://gist.github.com/anonymous/a01f678f3945838d133d
20:23:33  <sblom>(I don't know why that's anonymous instead of sblom. Weird.)
20:24:02  * dap1quit (Read error: Connection reset by peer)
20:24:09  <sblom>I assume "we or v8 are stomping the heap" is more usually "we" than "v8"?
20:24:12  <tjfontaine>I'm sure it's us shitting on the heap
20:24:15  * dapjoined
20:24:32  <tjfontaine>but it's not impossible that v8 has gone wrong
20:25:21  <tjfontaine>brb lunch
20:25:29  <sblom>Yeah. Alright. The one good thing about this bug, at least, is that it's pretty reliably reproducible.
20:25:36  <tjfontaine>:)
20:25:41  <sblom>That's going to be the only thing that keeps me out of the Alps.
20:28:48  <indutny>let me know if I can help
20:32:43  * dominictarrjoined
20:38:13  * kazuponjoined
20:42:28  * bradleymeckquit (Quit: bradleymeck)
20:42:38  <Kakera>but are there IDEs that can browse libuv on Linux?
20:43:07  <trevnorris>Kakera: what do you mean by "browse"?
20:43:39  * kazuponquit (Ping timeout: 276 seconds)
20:43:46  <indutny>define "ide"
20:43:49  <indutny>vim can browse libuv
20:44:01  <indutny>pretty easily
20:44:56  <Kakera>I mean work similarly to opening the sln file in VS
20:45:47  <Kakera>http://puu.sh/2egY3
20:45:58  <mraleph1>trevnorris: I don't know if you still interested. but don't do like this -> https://gist.github.com/trevnorris/5111477
20:47:45  <trevnorris>mraleph1: yeah. what i was trying to do was keep a reference to a handle w/o needing to persist it, so in the future I could check if v8 had gc'd it or not.
20:48:32  <mraleph1>trevnorris: but you have to persist :-)
20:48:59  <mraleph1>trevnorris: handles are reused.
20:49:30  <trevnorris>mraleph1: yeah. that's the sucktitude. persisting is very expensive. I don't need all the fluff of a persistent handle. just a simple .HasBeenGCd().
20:49:58  <trevnorris>basically like a WeakMap, but on the cc side.
20:50:58  * bradleymeckjoined
20:51:00  <mraleph1>trevnorris: it is not possible, even internally to implement this without handles, it follows from how GC works. GC does not visit dead objects.
20:51:53  <trevnorris>mraleph1: well, WeakMaps allow to set a key as an object, then if it's the only reference to object it will still get gc'd. so later if I check for it, it's just undefiend.
20:52:06  <mraleph1>though internally it would be possible to have a cheaper implementation than a weak handle.
20:52:16  <mraleph1>*than a weak persistent handle*
20:52:16  <trevnorris>anything would help.
20:52:47  <mraleph1>yeah, but externally you are stuck with persistent handles
20:53:02  * bradleymeckquit (Client Quit)
20:54:02  * sgallaghquit (Remote host closed the connection)
20:54:14  <trevnorris>hm. so looks like i'm stuck from the api side. have an idea what more could be done internally?
20:55:36  <trevnorris>mraleph1: ultimately i'm trying to create a memory pool and hand out chunks to js object. i don't mind persisting the whole, but i'd like to know if the js object using a chunk has been gc'd so it can be reused.
20:56:03  <trevnorris>w/o needing to persist it and record that in the weakcallback.
20:58:10  * bradleymeckjoined
20:59:20  * c4miloquit (Read error: Connection reset by peer)
20:59:31  * c4milojoined
21:01:08  * c4miloquit (Read error: Connection reset by peer)
21:01:12  * c4milo_joined
21:02:55  <trevnorris>mraleph1: oh, and much thanks for irhydra. made figuring out my deopts a lot easier.
21:03:07  <mraleph1>\o/
21:03:13  <mraleph1>you are welcome
21:03:29  <mraleph1>I still have to push out some things, but I did not have time.
21:07:51  * c4milo_quit (Read error: Connection reset by peer)
21:08:09  * c4milojoined
21:12:43  <trevnorris>mraleph1: so you think the basics of what i'm looking for could be done internally?
21:12:54  <trevnorris>don't know how long it'd take me, but i'll even try writing a patch.
21:15:07  <_ry>Kakera: you can write a new GYP backend to generate project files for you IDE
21:15:23  <_ry>Kakera: not sure how hard that is... but maybe it's not terrible. :/
21:19:50  <tjfontaine>I think that's exactly what I said before lunch
21:25:12  <_ry>just checking out the new "make bench" pretty sweet
21:25:29  <_ry>does it give stable numbers?
21:26:41  <mraleph1>trevnorris: yes, it can be done internally: you keep the handle around and then zap it with undefined if after GC you discover that object is going to die. it's roughly the same as with weak handles minus weak callback and preserving an object until next GC.
21:27:37  * mmaleckichanged nick to mmalecki[zzz]
21:27:51  <mraleph1>trevnorris: you'll still need a "global" (persistent) handle though because local handles do not work.
21:28:04  <mraleph1>trevnorris: also I don't think such patch will be accepted :-)
21:28:12  <mraleph1>but who knows.
21:28:45  <_ry>guys wrk should be built with gyp and go in dep/
21:29:12  <TooTallNate>_ry: not a bad idea
21:29:13  <_ry>also wrk's license isn't mentioned in Node's license <-- isaacs
21:29:24  <tjfontaine>_ry: bench does give mostly stable numbers, depending on your environment of course
21:30:10  <_ry>send the upstream gyp build to the wrk guy
21:30:16  <trevnorris>haven't looked at make bench lately. does it still only run each test once?
21:30:33  <_ry>seems like it
21:30:46  <tjfontaine>I'm not sure if wrk should go in deps, it's not strictly a dependency for node to actually run
21:31:06  <TooTallNate>_ry: meh on moving to dep/ though, we don't do that for clousure_linter or gyp for example
21:31:11  <_ry>word
21:31:14  * c4miloquit (Remote host closed the connection)
21:31:16  <trevnorris>yeah, only a single run isn't going to give very useful numbers for tests like fs.
21:31:47  <TooTallNate>oh wow, it's just a Makefile :D
21:31:52  <_ry>ircretary: tell isaacs to mention wrk in node's license file
21:31:52  <TooTallNate>this'll be eay
21:31:52  <ircretary>_ry: I'll be sure to tell isaacs
21:31:53  <tjfontaine>it's simple
21:31:54  <TooTallNate>easy
21:32:21  <_ry>would be nice to be able to run that bench under msvs
21:32:32  <_ry>seems like it's all makefile driven at the moment
21:33:02  <_ry>anyway. /me disappears again :P
21:33:05  * _ryquit (Quit: leaving)
21:33:31  * rendarquit
21:33:48  <tjfontaine>bench will be coming to msvs, have no fear
21:34:22  <trevnorris>mraleph1: hm. yeah. that could go all types of wrong. maybe i'm going at this from the wrong way.
21:36:06  <trevnorris>mraleph1: can you think of an inexpensive way to determine when external memory that was chunked and handed off could be reused?
21:36:13  * isaacsfg
21:36:49  <mraleph1>trevnorris: with a regular API your only friend is a persistent handle :-)
21:37:51  <trevnorris>mraleph1: heh, ok. maybe i'll just open a bug or some such and see what the v8 team would consider.
21:38:07  <mraleph1>that would also work.
21:39:01  * kazuponjoined
21:41:05  <trevnorris>mraleph1: thanks for the insight. would have burned out trying to figure out how to do that.
21:42:19  * c4milojoined
21:43:13  <TooTallNate>https://github.com/TooTallNate/node/commit/wrk-gyp
21:43:23  <TooTallNate>oh wow, right when ry leaves :p
21:43:32  <tjfontaine>ya he buggered right off :P
21:44:08  * kazuponquit (Ping timeout: 240 seconds)
21:44:27  * c4milo_joined
21:44:36  <isaacs>TooTallNate: yah, i saw his message.
21:44:38  * c4miloquit (Read error: No route to host)
21:44:42  <isaacs>wrk should stay in tools, it's not a dep
21:44:46  <isaacs>but the license should probably be mentioned.
21:44:57  <TooTallNate>yup, agree'd
21:45:58  <TooTallNate>isaacs: gyp file works on OS X it seems
21:46:28  <TooTallNate>isaacs: does wrk work on windows? if not, with the gyp file it might!
21:47:48  <MI6>joyent/node: isaacs master * ea30ca9 : doc: Add wrk's license to LICENSE file - http://git.io/JAYrtg
21:48:32  * c4milo_quit (Read error: No route to host)
21:48:37  <isaacs>TooTallNate: i would be very surprised if wrk works on windows
21:48:38  * c4milojoined
21:48:50  * hzquit
21:48:51  <tjfontaine>by works we mean builds?
21:48:55  <tjfontaine>:)
21:49:06  <isaacs>well, builds, and also runs
21:49:09  <TooTallNate>isaacs: ahhah, ya i guess it's not using libuv or anything ;)
21:49:17  <isaacs>it doens't have a IOCP backend
21:49:23  <TooTallNate>right
21:49:53  <TooTallNate>in any case… ry got his gyp file… not sure the overall benefit but it's there :p
21:49:59  <isaacs>sure.
21:54:12  * qmxchanged nick to qmx|away
22:02:32  * c4miloquit (Read error: No route to host)
22:02:39  * c4milo_joined
22:02:57  <MI6>nodejs-master: #84 UNSTABLE osx-ia32 (1/552) windows-x64 (11/552) windows-ia32 (11/552) linux-ia32 (1/552) http://jenkins.nodejs.org/job/nodejs-master/84/
22:13:12  <wolfeidau>TooTallNate: Heya, just had a question regarding that smartos ld issue, do you know of any other node-gyp projects which build some shared libraries?
22:13:55  <TooTallNate>wolfeidau: you could try almost any of mine :D
22:13:56  <TooTallNate>node-ffi
22:13:59  <TooTallNate>node-lame
22:14:00  <TooTallNate>node-ogg
22:14:03  <TooTallNate>node-vorbis
22:14:10  <wolfeidau>TooTallNate: Sweet thanks
22:14:12  <TooTallNate>they've all been gyp-ified
22:14:19  <sblom>test-cluster-net-send investigation continues: Looks like we might be freeing a uv_timer_s prematurely.
22:14:42  * mikealquit (Quit: Leaving.)
22:14:42  <tjfontaine>hm
22:15:19  <MI6>joyent/node: [email protected] master * 9f4c3b0 : http: check if incoming parser has already been freed Fix #4948 This ad - http://git.io/i4EKVw
22:15:23  <sblom>Can someone who's smarter than me spend some time speculating about how that might cause the assert that we're seeing?
22:15:55  <sblom>Meanwhile, I'm working on finding where the responsible free is happening.
22:16:19  * stagasquit (Ping timeout: 256 seconds)
22:16:48  <TooTallNate>sblom: good luck soldier
22:16:56  <TooTallNate>sblom: http.js is a minefield :D
22:18:04  <trevnorris>my brain's starting to go so i'm having a hard time seeing why changing args.This() to args[0]->ToObject() would cause a seg fault: https://gist.github.com/trevnorris/5118651
22:18:09  * c4milo_quit (Remote host closed the connection)
22:18:50  <tjfontaine>trevnorris: what does gdb say?
22:20:26  * c4milojoined
22:22:37  <indutny>bt
22:26:02  <trevnorris>https://gist.github.com/trevnorris/5120418
22:26:25  <trevnorris>guess i could have left out the last 54 lines. =P
22:27:24  <sblom>One thing my debug instrumentation has uncovered is that we're not setting a uv_req's type field in at least one case. https://github.com/joyent/libuv/blob/master/src/win/pipe.c#L89
22:27:29  <sblom>I can't prove that's bad.
22:27:38  <sblom>And it doesn't look like it's causing this particular test case assert.
22:27:41  <sblom>But isn't that weird?
22:29:05  <MI6>nodejs-master: #85 UNSTABLE windows-ia32 (12/553) smartos-x64 (1/553) linux-x64 (14/553) osx-x64 (5/553) windows-x64 (13/553) smartos-ia32 (1/553) linux-ia32 (5/553) osx-ia32 (2/553) http://jenkins.nodejs.org/job/nodejs-master/85/
22:29:15  <tjfontaine>that does seem weird
22:30:23  <tjfontaine>isaacs: that last commit broke things
22:32:25  <isaacs>yeah.....
22:32:26  <isaacs>seems that way
22:32:36  <isaacs>gonna roll back, i think
22:33:21  <wolfeidau>TooTallNate: node-ffi builds fine and also uses the smartos linker also checked the .a file with ar -t and it is fine
22:33:39  <isaacs>oh, wtf.
22:33:41  <isaacs>yeah
22:35:17  <isaacs>sorry, folks
22:35:29  <MI6>joyent/node: isaacs master * 632b7d8 : Revert "http: check if incoming parser has already been freed" This reve - http://git.io/OFPXSA
22:35:30  * AvianFluquit (Remote host closed the connection)
22:36:45  <isaacs>looks like it's just the test that's fubar. the actual chagne to http.js is fine
22:40:59  * kazuponjoined
22:42:24  <TooTallNate>wolfeidau: well so then it's something in your guys' court, right?
22:42:40  <trevnorris>indutny: um... no idea what's going on now. went back to a revision that worked and that's seg faulting too. well, this will be fun.
22:42:58  * mikealjoined
22:43:25  <wolfeidau>TooTallNate: Yeah looks that way
22:43:48  <indutny>happens sometimes
22:45:32  * kazuponquit (Ping timeout: 250 seconds)
22:45:48  <MI6>joyent/node: Ben Noordhuis master * e325ace : buffer: speed up ascii character scanning Speed up ASCII character scann (+1 more commits) - http://git.io/undodw
22:47:08  <MI6>joyent/node: isaacs master * 29cd0f2 : domains: Handle errors thrown in nested error handlers If a domain error - http://git.io/5elRzg
22:47:35  <isaacs>anyone wanna review a zlib bug? https://github.com/joyent/node/pull/4958
22:47:42  <isaacs>TooTallNate indutny ^
22:48:49  <isaacs>indutny: https://github.com/joyent/node/pull/961 <-- is that still a thing?
22:48:51  <MI6>nodejs-master: #86 UNSTABLE windows-ia32 (11/552) windows-x64 (11/552) http://jenkins.nodejs.org/job/nodejs-master/86/
22:49:52  <indutny>isaacs: meh
22:50:05  <indutny>I think streams2 will already do this
22:50:14  <indutny>also
22:50:21  <indutny>PR itself might be buggy.
22:50:31  <indutny>but please keep it open :)
22:50:42  <indutny>I'll return back to it eventually
22:51:59  <wolfeidau>TooTallNate: This interesting the .a files in node-leveldown are different https://gist.github.com/wolfeidau/5120604 ever seen that with node-gyp before?
22:52:28  <isaacs>indutny: sure.
22:52:35  * c4miloquit (Remote host closed the connection)
22:52:38  <TooTallNate>wolfeidau: try cat'ing yours… i have a feeling they're text files of some sort
22:52:43  <isaacs>//github.com/joyent/node/pull/4958
22:52:45  <TooTallNate>wolfeidau: but no, i haven't seen that before
22:52:53  <isaacs>hrm. pastefail
22:54:30  <wolfeidau>TooTallNate: Yeah it is text, updated https://gist.github.com/wolfeidau/5120604 with output, very odd
22:55:31  <TooTallNate>wolfeidau: looks binary to me… but in any case, give this a shot: https://github.com/rbranson/node-ffi/commit/649eecfef79517dd65482b0df6efa9c572095d26
22:55:48  <TooTallNate>wolfeidau: it would be good if you could verify if, say, node-speaker compiles
22:56:09  <MI6>joyent/node: isaacs windows-test-fixing * 7dfe074 : test: Kill zombies when debugger-client fails on windows (+7 more commits) - http://git.io/TDJudQ
22:58:14  <isaacs>huh. this is weird.
22:58:23  <isaacs>the buffer-ascii change is different on windows than on unix
22:58:35  <isaacs>AssertionError: "h├⌐rit├⌐" == "hC)ritC)"
22:58:41  <indutny>oh gosh
22:58:43  <isaacs>the left is what windows gets, the right is what unix gets
22:58:55  <indutny>wait
22:59:05  <indutny>left is ASCII
22:59:14  <indutny>I know this symbols
22:59:54  <wolfeidau>TooTallNate: Your a wizard, that worked!
23:00:13  <TooTallNate>wolfeidau: sweet!
23:00:18  <isaacs>oh, yeah, it's busted.
23:00:20  <wolfeidau>TooTallNate: Added that to both the deps gyp files and building now
23:00:21  <isaacs>indutny:
23:00:22  <isaacs>> new Buffer(new Buffer('hérité').toString('ascii'))
23:00:22  <isaacs><Buffer 68 43 29 72 69 74 43 29>
23:00:25  <isaacs>unix ^
23:00:34  <isaacs>> new Buffer(new Buffer('hérité').toString('ascii'))
23:00:34  <isaacs><Buffer 68 c3 a9 72 69 74 c3 a9>
23:00:36  <isaacs>windows ^
23:00:41  <wolfeidau>TooTallNate: I need to know why! lol
23:01:13  <wolfeidau>TooTallNate: I will comment on that issue and get @rvagg to close it if it all works :)
23:01:23  <TooTallNate>wolfeidau: i guess thin archives are a new-ish feature to linkers… not totally compatible everywhere
23:02:16  <wolfeidau>TooTallNate: Yeah, I haven't really delved into this stuff in years, you can lost in all the nuances of different platforms/tools
23:02:23  <MI6>nodejs-master: #87 UNSTABLE windows-ia32 (11/554) osx-x64 (1/554) windows-x64 (11/554) osx-ia32 (1/554) http://jenkins.nodejs.org/job/nodejs-master/87/
23:02:46  <TooTallNate>wolfeidau: i'm just diving into it, so i'm still learning all those nuances
23:02:58  <TooTallNate>wolfeidau: but glad that arcane gyp flag works for you :p
23:03:16  <isaacs>indutny: ohhh... wait. forgot to rebuild. i'm a doof.
23:03:56  <wolfeidau>TooTallNate: Those arcane flags are the fairy dust that makes this stuff work some times :)
23:04:56  <wolfeidau>TooTallNate: Either way I have a dtrace log of every command run, going to rerun it now with that flag and see what the diff is
23:05:10  <wolfeidau>TooTallNate: Thanks again
23:05:19  <TooTallNate>wolfeidau: well it's -T on the linker, i'm pretty sure
23:05:28  <isaacs>indutny: yeah, it works fine. ignore me.
23:05:31  <TooTallNate>wolfeidau: setting that flag to 1 removes -T
23:06:31  <wolfeidau>TooTallNate: Ahh yeah read something from 2007 about some issues with that on an older version of solaris, that was the only reference to that error I could find
23:07:13  <TooTallNate>wolfeidau: can you let me know if "speaker" compiles for you or not?
23:07:14  * asdf12joined
23:07:24  <isaacs>asdf12: hi!
23:07:28  <isaacs>:)
23:07:42  <isaacs>asdf12: so, in that PR... the code is good, but the test is problematic.
23:07:58  <wolfeidau>TooTallNate: Will try it now
23:08:14  <asdf12>regarding #4948, how do you advise the confusing requests in a request ? originally it was a external http request to google.com
23:08:27  <asdf12>to show a real use case
23:08:29  <isaacs>asdf12: i think you could probably just set a flag or something.
23:08:51  <isaacs>asdf12: if (gotFirstRequest) { req.end('hello'); } else { gotFirstRequest = true; http.request(...) }
23:08:57  * bradleymeckquit (Quit: bradleymeck)
23:08:59  <isaacs>asdf12: so you dont' recurse to forever.
23:09:03  <wolfeidau>TooTallNate: Same error as leveldown mate
23:09:12  <TooTallNate>wolfeidau: i was afraid of that…
23:09:30  <wolfeidau>TooTallNate: Want to see the whole output?
23:09:32  <TooTallNate>wolfeidau: so in that case, i think i might want to add this flag to node-gyp itself, so it's applied to all "static_library" targets…
23:09:38  <TooTallNate>wolfeidau: no i believe you :)
23:09:42  <TooTallNate>wolfeidau: cause I get it was well
23:10:25  <wolfeidau>TooTallNate: Yeah understood, does node-ffi have the same flag enabled?
23:10:41  <TooTallNate>wolfeidau: yes, that's the commit i liked to you
23:10:43  <isaacs>asdf12: does that make sense?
23:10:45  <TooTallNate>linked
23:10:47  <asdf12>isaacs: yea sorry, i checked the tests when it was suppose to fail, never re-ran it after applying the fix, since it was a one-liner :X
23:11:06  <wolfeidau>TooTallNate: aha derp yeah just looked it it in more detail lol
23:11:09  <asdf12>so i didn't realize it would trickle on
23:11:12  <sblom>test-cluster-net-send investigation: okay--really getting somewhere now. Looks like the uv_tcp_endgame has already run when uv_poll gets a new request for an overlapped whose corresponding req is already cleaned up.
23:12:24  <isaacs>asdf12: yeah, luckily we do run `make test` quite often, so it didn't get far :)
23:12:32  <isaacs>asdf12: my bad for letting it slip out in the first place.
23:12:51  <wolfeidau>TooTallNate: Need to read more about these thin .a files
23:12:52  <sblom>Do we have reason to believe piscisaureus_ is still awake?
23:13:05  <sblom>(or, more precisely, going to be back here today?
23:13:06  <sblom>)
23:13:07  <isaacs>sblom: my guess is that he is not around.
23:13:10  <TooTallNate>wolfeidau: right… i mean gyp enables them by default…
23:13:15  <isaacs>sblom: he said he was heading to a birthday party
23:13:21  <sblom>isaacs: Okay.
23:13:25  <TooTallNate>wolfeidau: but google is also in charge closely of their build tools
23:13:36  <TooTallNate>wolfeidau: so they don't have to worry about cross-OS compat as much as we do
23:13:48  <wolfeidau>TooTallNate: Aha
23:13:54  <isaacs>sblom: but he said he'd be back saturday/sunday (ie, our late-this-evenign)
23:16:31  <isaacs>heading home. this cafe is closing. will be back online in a bit. going to dig into simple/test-cluster-bind-twice-v2, possibly our most complicated test, which of course is one of the windows failers :)
23:16:40  <isaacs>(throwing einval)
23:19:38  <wolfeidau>TooTallNate: As this is at the top level couldn't this impact building on other platforms? I note you applied this for centos, which version had this issue?
23:20:27  <TooTallNate>wolfeidau: centos 5 was the reported version
23:36:17  <TooTallNate>wolfeidau: node-gyp v0.9.0 defaults to non-thin archives
23:36:26  <TooTallNate>wolfeidau: you can pass --thin to do the original behavior
23:37:58  <wolfeidau>TooTallNate: Did you just release that? Would be interested to see if that avoids making changes to leveldown
23:38:10  <TooTallNate>wolfeidau: yes, just now...
23:38:31  <trevnorris>tjfontaine: son of a bitch. i had forgotten to initialize the memory before it was being used. took me an hour to figure that out...
23:38:43  <tjfontaine>trevnorris: that would do it yup :)
23:38:48  <wolfeidau>TooTallNate: Well i will give it a test :)
23:41:37  * kazuponjoined
23:43:52  <trevnorris>isaacs: love the note about not going into 0.10. isn't that going out monday?
23:45:14  <TooTallNate>trevnorris: which note?
23:45:28  <trevnorris>TooTallNate: https://github.com/joyent/node/pull/4964
23:45:57  <TooTallNate>trevnorris: nice. and ya, monday is the plan
23:46:39  * kazuponquit (Ping timeout: 260 seconds)
23:47:00  <trevnorris>heh, yeah. no possibility of going out that soon. don't see this change taking any less than a month.
23:52:41  <isaacs>trevnorris: yes :)
23:53:08  <isaacs>trevnorris: just forestalling the "backport to stable" requests :)
23:54:10  <trevnorris>isaacs: ah, that's what you're referring to. yeah, there are so many ways that could go wrong.
23:54:42  <asdf12>isaacs: can i just do process.exit() to exit a test cleanly? that would be better than adding too much code to close hanging sockets
23:54:59  <asdf12>well not too much, but process.exit is a one liner
23:56:04  <TooTallNate>asdf12: preferably not…
23:56:11  <tjfontaine>man this jenkins tap plugin really has some fun quirks -- http://jenkins.nodejs.org/job/nodejs-master/lastCompletedBuild/DESTCPU=x64,label=windows/testReport/http://jenkins.nodejs.org//job/nodejs-master/DESTCPU=x64,label=windows/87//tapTestReport/test.tap-24/
23:58:21  * Kakeraquit (Ping timeout: 245 seconds)