00:09:52  <Trott>Kicked off a node-daily-master just so we can get a baseline of sorts. https://ci.nodejs.org/job/node-daily-master/904/
00:25:22  * node-ghjoined
00:25:23  * node-ghpart
00:36:57  * Fishrock123quit (Remote host closed the connection)
00:37:33  * Fishrock123joined
00:42:08  * Fishrock123quit (Ping timeout: 258 seconds)
00:44:52  * Fishrock123joined
00:58:14  <rvagg>mylesborins: figured out the smartos14 thing—there's a thing in Jenkins that removes the `release-only` bit for non-release builds (cause it screws everything up, see https://github.com/nodejs/node/pull/12958) but it doesn't work on smartos! `perl -pi -e 's/: release-only/:/g' Makefile` works on every other platform but smartos14, even `sed -i` doesn't do anything, I don't understand it but I'm going to chalk it up
00:58:14  <rvagg>to OS differences for now (which doesn't make a ton of sense tbh)
01:02:52  * joyeejoined
01:04:40  * Fishrock123quit (Remote host closed the connection)
01:14:37  * node-ghjoined
01:14:37  * node-ghpart
01:23:52  * Fishrock123joined
02:00:26  <joyee>Hi, anyone knows which target in the Makefile are the Raspberry Pis using to run tests?
02:00:31  <joyee>See https://github.com/nodejs/node/pull/16301
02:02:03  <refack>joyee: Trott asked about it a few hours ago. I think the Jenkins job was broken (it pulled `master` instead of the PR branch)
02:05:10  <refack>Anyway it's `make test-ci-js`
02:14:16  <joyee>refack: OK thanks
02:14:55  <joyee>refack: just got up so didn't see the previous conversations. (should definitely get a bounce server set up
02:15:22  <refack>http://logs.libuv.org/node-build/2017-10-18
02:15:42  <joyee>(thumb)
02:39:50  * Fishrock123quit (Remote host closed the connection)
02:39:58  <Trott>joyee: My attempted fixes have all run into various problems. Do you have a solution you're working on? My next step is to move `test-make-doc` to its own directory and add it that directory to `IGNORED_SUITES` in `test.py`. But if you're working on a simpler/better solution, I'm happy to stop what I'm doing. :-D
02:42:18  <Trott>Actually, I think I have a better solution than that...maybe...
02:43:44  <joaocgreis>About the arm-fanned job: the compile part runs on a machine that is configured with the scripts at the bottom of https://github.com/nodejs/build/pull/244/files , then it just uses make build-ci
02:44:49  <joaocgreis>then the test-binary uses test-ci-native on only one worker and test-ci-js on all the others dividing the tests.. the sum of this should always be exactly the same as test-ci
02:45:57  <joaocgreis>Windows should start using something like this when I have time to do it, didn't yet because the test matrix is already huge and I want to split the job
02:46:39  <joaocgreis>Trott, joyee, refack hope this helps!
02:47:37  <refack>👍
02:47:50  <Trott>Thanks joaocgreis. My current plan is to move the test to doctool, add doctool to IGNORED_SUITES, and add doctool explicitly to `make test` and `make test-ci` but leave it off of `make test-ci-js`.
02:47:55  <Trott>Testing locally now.
02:48:59  <refack>Is it the same issue as with the ESM test, that the build & test are coupled but run on different machines?
02:49:31  <refack>Call `make doc-only` from within the test
02:50:19  <refack>Alternatively skip on windows & arm. We just need it to be sane on one platform
02:51:54  <joyee>Trott: yeah that's what I thought about doing when I landed that PR
02:52:19  <joyee>(minus the IGNORED_SUITES part
02:52:42  <Trott>refack: As a last resort, yes. But if possible, I'd rather not skip on arm because it really should build there. Having it run depending on whether it's fanned or not is closer to feature detection IMO, so hopefully more robust.
02:53:21  <Trott>joyee: Your instincts were good. :-D
02:53:35  * Fishrock123joined
02:53:55  <joyee>Trott: hmm..any reason why we should build the docs on arm
02:54:05  <Trott>If anyone wants to review: PR is at https://github.com/nodejs/node/pull/16301
02:55:55  <Trott>joyee: On arm in general, sure. But on the Raspberry Pi devices specifically: We build on a separate machine than where the tests run. I *think* what we do now is just move over `out/Release`. If we run the test on those machines, we'll need to move over `out/doc` as well. I guess that's do-able, but I don't know how.
02:56:08  <Trott>So, future enhancement for someone who wants to dig into it, maybe? :-D
02:58:45  <joyee>Trott: I guess we could do what refack suggested, run `make doc-only` in the test? (Similar to what my original PR did
02:59:00  <Trott>Two problems with that:
02:59:15  <joyee>I think we bumped into that -jn issue on freebsd because it's not `make doc-only`, but `make doc` so it builds NODE_EXE
03:00:03  <refack>a good case study for how fast bit-rot spreads. During the two weeks we didn't properly cover the PIs we got two similar failing edge cases
03:00:37  <Trott>Hmmm, never mind. Both of my "problems" may be non problems.
03:01:12  <refack>Night & good luck 🎩
03:01:48  <Trott>gn
03:02:34  <Trott>Maybe only run `make doc-only` if `out/doc` isn't found?
03:03:35  <Trott>joyee: I'm confused. Do you think it will work if we just change the `make doc` bits you added `Makefile` to instead invoke `make doc-only`? Or will that fail on freebsd?
03:04:12  <Trott>Because that sure *seems* like a straightforward solution....
03:07:36  <Trott>Trying that in https://github.com/nodejs/node/pull/16309.
03:18:52  <joyee>Trott: yep, that's what I think
03:19:41  <joyee>Oh, I read the new PR, I meant to invoke `make doc-only` in the test script
03:19:56  <joyee>as in, with child_process.spawn
03:25:04  * Fishrock123quit (Quit: Leaving...)
08:24:33  * seishunjoined
08:37:23  * joyeequit (Remote host closed the connection)
08:37:58  * joyeejoined
08:42:22  * joyeequit (Ping timeout: 264 seconds)
08:46:08  * joyeejoined
10:25:12  * mylesborinsquit (Quit: farewell for now)
10:25:42  * mylesborinsjoined
10:31:31  * joyeequit (Remote host closed the connection)
10:41:34  * joyeejoined
10:46:25  * joyeequit (Ping timeout: 248 seconds)
11:40:14  * node-ghjoined
11:40:14  * node-ghpart
12:50:25  * joyeejoined
12:57:27  * joyeequit (Ping timeout: 240 seconds)
13:26:11  * evanlucasjoined
14:02:14  * joyeejoined
14:08:33  * node-ghjoined
14:08:33  * node-ghpart
14:12:05  * joyeequit (Remote host closed the connection)
14:22:26  * joyeejoined
14:27:00  * joyeequit (Remote host closed the connection)
14:27:35  * joyeejoined
14:51:07  * node-ghjoined
14:51:07  * node-ghpart
14:54:24  * node-ghjoined
14:54:24  * node-ghpart
16:00:28  * node-ghjoined
16:00:28  * node-ghpart
16:00:57  * node-ghjoined
16:00:57  * node-ghpart
17:06:57  * node-ghjoined
17:06:57  * node-ghpart
17:07:29  * node-ghjoined
17:07:29  * node-ghpart
17:18:49  * Fishrock123joined
18:12:52  * node-ghjoined
18:12:52  * node-ghpart
18:37:27  * node-ghjoined
18:37:27  * node-ghpart
18:41:20  * node-ghjoined
18:41:20  * node-ghpart
18:55:51  * joyeequit (Remote host closed the connection)
19:01:53  * seishunquit (Ping timeout: 248 seconds)
19:06:09  * joyeejoined
19:10:24  * joyeequit (Ping timeout: 246 seconds)
19:54:42  * Fishrock123quit (Remote host closed the connection)
19:56:00  * Fishrock123joined
20:08:55  * seishunjoined
20:15:59  * joyeejoined
20:20:46  * joyeequit (Ping timeout: 264 seconds)
20:49:04  * node-ghjoined
20:49:04  * node-ghpart
20:51:32  * node-ghjoined
20:51:32  * node-ghpart
21:05:20  * Fishrock123quit (Remote host closed the connection)
21:17:10  * seishunquit (Ping timeout: 264 seconds)
21:27:53  <ofrobots>Hi All! I need to reproduce a windows only crash (https://ci.nodejs.org/job/node-test-binary-windows/12066/COMPILED_BY=vs2017,RUNNER=win2016,RUN_SUBSET=1/tapResults/)..
21:28:12  <ofrobots>It doesn't happen on the windows machine I have access to
21:28:33  <ofrobots>Is there a way for me to get access to the build machine?
21:28:49  <ofrobots>That's the only environment where this test 'crashes'
21:29:04  <refack>We have a "process" for delegating acess
21:29:08  <ofrobots>Side question: do we generate dumps on crashes?
21:29:23  <refack>https://github.com/nodejs/build/projects/2
21:29:31  <ofrobots>It would be good if I (the person who launched the build) could extract dumps would needing access to the machine
21:29:46  <refack>RE: dumps, only on smartOS 🤷‍♂️
21:30:01  <refack>But I can set that up, or "be your proxy"
21:30:39  <ofrobots>the stack printed is quite useless.. 'oh no!'
21:30:53  <refack>Did you try building with VS2017?
21:31:09  <ofrobots>refack: trying to figure out how to get that onto my machine..
21:31:10  <refack>We've seen crashes only with VS2017 generated binaries
21:31:46  <refack>https://www.visualstudio.com/downloads/#build-tools-for-visual-studio-2017
21:32:05  <refack>in the orphaned list at the bottom: Build Tools for Visual Studio 2017
21:32:42  <refack>*that is crashes that happen only with the VS2017 generated binary
21:32:51  * Fishrock123joined
21:34:03  <ofrobots>getting that setup now
21:34:44  <refack>Looking at your PR I believe it's that same issue - https://github.com/nodejs/node/issues/15558#issuecomment-333278380
21:34:54  <refack>the VS2017 has a very high fail rate
21:35:34  <refack>I've been narrowing it down, I think it's a double free in a race condition when the session is closed
21:37:09  <ofrobots>refack: thanks for the pointer, that does seem likely related
21:37:48  <ofrobots>for my test it failed in 2/2 runs.
21:38:00  <ofrobots>not sure what the flakiness rate is for the original issue
21:38:07  <refack>about that
21:38:23  <refack>all the yellows you might have seen recently in CI
21:40:26  * Fishrock123quit (Remote host closed the connection)
21:43:46  * Fishrock123joined
21:45:09  * evanlucasquit (Remote host closed the connection)
21:51:16  * Fishrock123quit (Remote host closed the connection)
22:02:49  * node-ghjoined
22:02:49  * node-ghpart
22:04:04  * Fishrock123joined
22:07:18  <ofrobots>refack: I can reproduce the intermittent failure with VS2017
22:07:24  <ofrobots>fails about 50% of the time
22:07:35  <ofrobots>I can confirm that it is related to session closing
22:08:45  <refack>Ok, if you fix that we're getting two birds with one stone
22:11:11  <ofrobots>I was hoping to get #16308 landed in time for LTS
22:11:18  <ofrobots>but that's looking iffy now
22:11:52  <refack>We've been living with 15558 for a while. I won't consider your fails a blocker
22:12:01  <refack>the version is explicitly built on VS2015
22:12:23  <ofrobots>I want to look a bit further until I am certain that it is 15558..
22:13:16  <refack>I'll stress your PR on a VS2015 vs VS2017 machines
22:35:14  <rvagg>joaocgreis: I've just approved the DigiCert changes, I'd expect to see a certificate drop today or tomorrow after that
22:36:56  * joyeejoined
22:41:05  * joyeequit (Ping timeout: 248 seconds)
22:41:37  * Fishrock123quit (Remote host closed the connection)
22:42:21  * node-ghjoined
22:42:21  * node-ghpart
22:42:55  * Fishrock123joined
22:48:27  <rvagg>mylesborins: 6.12.0-rc2 went through smartos14 just fine this time. That problem with release-only was sufficiently weird that I decided to reboot the machine and clear the workspace and that seems to have done the trick
23:00:08  * Fishrock123quit (Remote host closed the connection)
23:17:51  * node-ghjoined
23:17:51  * node-ghpart
23:20:28  * node-ghjoined
23:20:28  * node-ghpart