[03:33:27] lhodev: https://review.gerrithub.io/424973 already does that (binaries are installed) [05:25:31] *** Joins: darsto_ (~darsto@89-78-174-111.dynamic.chello.pl) [05:26:50] *** Quits: darsto (~darsto@89-78-174-111.dynamic.chello.pl) (Ping timeout: 272 seconds) [05:26:51] *** darsto_ is now known as darsto [06:05:30] *** Quits: alekseymmm (050811aa@gateway/web/freenode/ip.5.8.17.170) (Quit: Page closed) [09:34:49] *** Joins: travis-ci (~travis-ci@ec2-54-224-178-97.compute-1.amazonaws.com) [09:34:50] (spdk/master) nvme: Enable RNR Retry in RDMA transport (Ben Walker) [09:34:50] Diff URL: https://github.com/spdk/spdk/compare/3daa9e6f756f...438bae796682 [09:34:50] *** Parts: travis-ci (~travis-ci@ec2-54-224-178-97.compute-1.amazonaws.com) () [09:35:59] *** Joins: travis-ci (~travis-ci@ec2-54-160-200-139.compute-1.amazonaws.com) [09:36:00] (spdk/master) test/nvmf: Use create_transport instead of set_tgt_opts (Seth Howell) [09:36:00] Diff URL: https://github.com/spdk/spdk/compare/a9e484a66082...5fbbb6c05658 [09:36:00] *** Parts: travis-ci (~travis-ci@ec2-54-160-200-139.compute-1.amazonaws.com) () [09:41:14] *** Joins: travis-ci (~travis-ci@ec2-54-160-200-139.compute-1.amazonaws.com) [09:41:15] (spdk/master) memory: lock global mem_map only when necessary (Darek Stojaczyk) [09:41:15] Diff URL: https://github.com/spdk/spdk/compare/438bae796682...a9e484a66082 [09:41:15] *** Parts: travis-ci (~travis-ci@ec2-54-160-200-139.compute-1.amazonaws.com) () [10:44:29] *** Joins: travis-ci (~travis-ci@ec2-54-211-208-217.compute-1.amazonaws.com) [10:44:30] (spdk/master) lib/nvmf: fix a typo error (yidong0635) [10:44:30] Diff URL: https://github.com/spdk/spdk/compare/5fbbb6c05658...b7201f8e7f29 [10:44:30] *** Parts: travis-ci (~travis-ci@ec2-54-211-208-217.compute-1.amazonaws.com) () [13:44:43] jimharris: bwalker: sethhowe: Took note of your email to list (re: subject Patches for upcoming 10.18 release). The next phase of work to "complete" the shared-lib work on which Seth and I have been working, is to add inter-spdk lib dependencies. Been looking at that and the extra considerations related to handling alternate env implementations. Shall we aim to get that effort in for 18.10, or after? [13:45:09] are there patches out already? [13:46:16] Seth generated some cool graphs of our dependencies that we reviewed this morning [13:46:20] it would be great to get this in for 18.10, but if we don't have patches ready yet it may be hard to get them in for 18.10 [13:46:30] Not from me; i.e. dunno if sethhowe has done anything yet on that. [13:46:31] and he's working out some things based on that now [13:47:54] I'd like to see that (graphs) and see how they coincide with the script I had written as an aid to id'ing the dependencies myself. [13:49:08] *** Joins: johnmeneghini (~johnmeneg@pool-108-49-234-74.bstnma.fios.verizon.net) [13:49:27] Hello spdk land… [13:50:33] Is it just me, or is the mater branch busted? [13:50:34] Fedora:vagrant:spdk(master) > make [13:50:34] /home/vagrant/spdk_repo/spdk/mk/spdk.common.mk:164: /env.mk: No such file or directory [13:50:35] make: *** No rule to make target '/env.mk'. Stop. [13:50:49] johnmeneghini: run ./configure [13:50:51] do make clean, re-run configure, and make again [13:51:41] sethhowe: did you get any use from that script I shared with you for the inter-spdk dependencies, or possibly use something else? [13:51:46] So we have to run configure all the time now? It used to be optional [13:51:53] just one time [13:52:02] we changed the way it works internally [13:52:18] running the new configure once is required [13:52:24] after that, you just have to type make [13:52:57] configure generates that env.mk file [13:53:02] that you're missing in your error [13:53:06] Gotta love it. Seems like all you should need to do is git clone; git submodule —init —update; and make [13:53:37] running ./configure && make is the standard procedure for the vast majority of software packages. Anything using autotools (which we don't, but are mimicking) [13:54:11] Then I suggest we add a make rule in the top level Makefile that tests for env.mk and runs configure for you. :-) [13:54:45] we could at a minimum test for env.mk and give a better error message [13:55:03] maybe there is a default set of arguments to configure that are good enough for most people [13:55:08] Agreed. [13:55:23] ok - we'll just to clean that up [13:56:10] OK. Next problem. [13:56:12] Fedora:vagrant:spdk(master) > make clean [13:56:13] rm: cannot remove '/home/vagrant/spdk_repo/spdk/intel-ipsec-mb/obj/aes128_cbc_dec_by4_sse.o': Permission denied [13:56:13] rm: cannot remove '/home/vagrant/spdk_repo/spdk/intel-ipsec-mb/obj/aes128_cbc_dec_by8_avx.o': Permission denied [13:56:33] Fedora:vagrant:spdk(master) > sudo make clean [13:56:33] Fedora:vagrant:spdk(master) > [13:56:40] That kind of sucks… [13:56:46] it looks like you built it as root at one point [13:57:03] are you intentionally testing crypto? [13:57:14] I didn't do that… I think the vagrant scripts did. [13:57:33] ok - intel-ipsec-mb should definitely not get built as root [13:57:35] No, I'd like to ignore crypto [13:57:41] for now... [13:57:50] that's a bug. It also shouldn't be building crypto if it wasn't enabled [13:58:11] you're doing this inside the vagrant vm? [13:58:34] OK. I've opened a bug already. I'm sure Seth will love me. [13:59:08] Yes, I do everytihing inside of my vagrant/vm [14:00:07] Next question… what do you guys want to do about the SPDK community meeting next Tuesday? I think you will all be at the airport at the time it is supposed to start…. [14:00:21] yeah I think we're going to need to cancel [14:00:37] peluse can you cancel it [14:00:40] OK. Just wanted to check. [14:01:17] I'm looking through the vagrant scripts now to figure out how it gets built as root [14:02:18] bwalker: Confused about johnmeneghini env.mk build issue. *That* env.mk is not generated, but instead the one included from lib/env_dpdk. The *new* configure process Pawel is working will create a mk/config.mk, and to my understanding that's not in place yet. [14:03:30] Sorry for the confusion guys. Just dusting off my spdk repository so I can be ready for next week.... [14:04:27] Let me rephrase… I use my spdk repository every day… I just do keep up with the master branch… which is like a jungle [14:04:56] *just do not keep up with the master branch [14:05:10] https://review.gerrithub.io/#/c/spdk/spdk/+/426364/ is the patch work in progress for the mk/config.mk that will replace CONFIG.local. [14:05:17] yeah - I need to dive a little deeper into this build stuff anyway [14:05:41] ok, I was confused between the new config.mk and the old env.mk [14:06:02] Our build *is* complicated ;-) [14:06:18] > Lets remove this mess by producing final mk/config.mk file and get rid [14:06:18] of CONFIG.local [14:06:24] I love it! [14:06:42] yeah my head is already in the new strategy that pawel has out for review [14:07:04] now I can't explain why your build couldn't find env.mk though [14:07:36] lhodev: seth and i just chatted on the library dependency stuff - i'm going to start putting in the build infrastructure for it now [14:08:16] ohh, I know what happened [14:08:50] ok johnmeneghini - we changed the build just recently such that all possible CONFIG options have to show up in the CONFIG.local file [14:08:57] whereas previously, if you used the default, they wouldn't be present [14:09:03] jimharris: Can you expand on what you mean by "start putting in the build infrastructure for it now"? [14:09:16] Does that mean we *are* a 'go' for getting this into 18.10? [14:09:17] so if you have an old CONFIG.local file generated by configure [14:09:22] it's missing some things [14:09:30] so that we can add rules into lib/Makefile that it needs to build subdirectory X before subdirectory Y [14:09:31] in particular, CONFIG_ENV [14:09:38] which is the path to env.mk [14:09:47] jimharris: Gotcha on that piece (ordering of building libs) [14:09:48] so re-running configure fixes that problem [14:10:02] this is all in preparation for eliminating the CONFIG files and just generating a makefile fragment from configure [14:10:13] then we can instrument each library's makefile on which other spdk libraries it depends on [14:10:31] ok, now looking at how it built the intel-ipsec-mb library [14:11:11] jimharris: I was hoping to have a discussion about the dependency on libspdk_env_dpdk.so with Seth et. al. [14:11:49] my intel-ipsec-mb library was created by the root account. This was done my the setup.sh script (or some vagrant script) [14:12:04] yeah - it's probably running git submodule update --init as root [14:12:07] which it should not [14:12:55] lhodev: i'd like to get enough of the work in place at least so that if it doesn't get into 18.10, that it could safely go into a 18.10.1 [14:13:15] what's the discussion you need on libspdk_env_dpdk.so? [14:13:21] johnmeneghini: I think the problem is in pkgdep.sh [14:13:30] when the VM comes up, that gets run, right? [14:13:51] Agreed. [14:14:08] and it's run as root [14:14:15] or at least under sudo [14:14:37] jimharris: Regarding your spdk lib dependency graph discussions, are y'all thinking that you will or will NOT create dependencies to libspdk_env_dpdk.so? [14:14:54] Karlo and Seth have been busy improving the vagrant scripts. I'm just finding some of the bugs. [14:15:07] lhodev: i understand your question - don't have an answer yet [14:15:11] this was actually just introduced recently when crypto was checked in [14:17:31] jimharris: I want to aim for supporting alt env's as cleanly as possible and so how (or if) we specify a dependency for spdk lib's on the env is of concern to me. [14:17:55] comm meeting - yes will cancel [14:19:30] lhodev: ack [14:21:31] lhodev: Sorry, not very good at talking and typing. Your script has been very helpful. I will send you an e-mail with some of the graphs I have been able to generate from it. [14:25:00] I wonder if pkgdep.sh should consult the output from configure [14:25:59] sethhowe: Thanks! I'll def keep an eye out for your email. [14:26:03] Here's another bug I found in vm_setup.sh [14:26:05] Fedora:vagrant:~ > spdk_repo/spdk/test/common/config/vm_setup.sh -i -t librxe,fio,vpp [14:26:05] Last metadata expiration check: 1:21:02 ago on Fri 12 Oct 2018 01:04:12 PM PDT. [14:26:05] Package git-2.14.4-1.fc27.x86_64 is already installed, skipping. [14:26:05] Dependencies resolved. [14:26:05] Nothing to do. [14:26:06] Complete! [14:26:06] spdk source already present, not cloning [14:26:07] sudo: ./scripts/pkgdep.sh: command not found [14:26:42] Why do we have hardcode paths in this script. I worked to remove them all — at one point. :-) [14:27:21] I think this is a fix for the problem you hit with 'make clean' [14:27:22] https://review.gerrithub.io/#/c/spdk/spdk/+/429282/ [14:27:45] johnmeneghini: So that you wouldn't run out of things to do (running for cover). [14:28:26] johnmeneghini: That one is my fault. I didn't think about the implications at the time. [14:30:59] I just ran that script and it created a directory two levels above my spdk repo [14:31:06] no where near where I ran it from [14:31:30] oh, it does cd ~ [14:31:41] so I assume that's what it is supposed to be doing [14:32:09] bwalker: Yes. It places all of the dependencies in the users home directory. [14:32:27] so is the fix to change that line to sudo spdk_repo/spdk/scripts/pkgdep.sh? [14:32:29] I don't recommend hardcoded paths in any script. cd ~ is a hardcoded path. [14:33:24] yeah I'm not sure why it couldn't just use whatever directory you're currently in [14:35:10] See my comments at: https://review.gerrithub.io/#/c/spdk/spdk/+/429282/ [14:35:10] sethhowe: Could you create the graph with the nodes in 3D space that we could rotate? Maybe use an Oculus Rift to explore? ;-) [14:35:44] lhodev: Seth is no longer in the office. . . [14:36:16] Neither is johnmeneghini [14:37:49] I need to quit for the day. Got to go an kiss the wife and have dinner. I'll see you all next week. We have 36 confirmed attendees for the SPDK Meetup. It's going to be much bigger than last year [14:40:17] johnmeneghini, thanks John - AWESOME!! [15:21:01] lhodev, sethhowe: https://review.gerrithub.io/#/c/spdk/spdk/+/429287/ [15:21:11] framework for defining inter-library dependencies [15:32:10] *** Quits: johnmeneghini (~johnmeneg@pool-108-49-234-74.bstnma.fios.verizon.net) (Quit: Leaving.) [15:35:36] there is makefile wizardy in there [15:35:50] define depdirs_rule [15:35:50] $(DEPDIRS-$(1)): [15:35:50] $(1): | $(DEPDIRS-$(1)) [15:35:50] endef [15:36:07] shamelessly stolen from dpdk [15:36:45] I see what it is doing, but I don't know how long it would have taken me to figure out how to write that [15:37:14] is this how DPDK specifies the dependencies? [15:38:14] basically - although i'm trying to reduce duplication by having the upper level makefile pass down the dependency dir list to the lower level makefile [15:38:30] and then the lower level makefile uses that to link against the correct libraries when building the shared library [15:38:44] but of course i just realized this approach doesn't work for lib/bdev and its subdirs [15:40:35] bwalker: can you go through some of the +2 patches? i think we can get a few of those in - and then will have fewer to get in next week [15:40:41] ok [15:49:11] *** Joins: travis-ci (~travis-ci@ec2-54-197-118-7.compute-1.amazonaws.com) [15:49:12] (spdk/master) lib/nvmf_tgt:remove the g_active_poll_groups variable (GangCao) [15:49:12] Diff URL: https://github.com/spdk/spdk/compare/b7201f8e7f29...e6819a3c52f4 [15:49:12] *** Parts: travis-ci (~travis-ci@ec2-54-197-118-7.compute-1.amazonaws.com) () [15:52:39] oh - it's even worse for the event/subsystems/* libraries [15:54:08] *** Joins: travis-ci (~travis-ci@ec2-54-163-81-59.compute-1.amazonaws.com) [15:54:09] (spdk/master) lib/bdev: reset bdev internal properties (shahar salzman) [15:54:09] Diff URL: https://github.com/spdk/spdk/compare/d94846c7f08d...1ca4b472524f [15:54:09] *** Parts: travis-ci (~travis-ci@ec2-54-163-81-59.compute-1.amazonaws.com) () [15:57:00] yeah those are tricky [16:38:37] *** Joins: travis-ci (~travis-ci@ec2-54-159-9-164.compute-1.amazonaws.com) [16:38:38] (spdk/master) util: make spdk_parse_ip_addr return -EINVAL instead -1 (Pawel Wodkowski) [16:38:38] Diff URL: https://github.com/spdk/spdk/compare/1ca4b472524f...7818939c6da7 [16:38:38] *** Parts: travis-ci (~travis-ci@ec2-54-159-9-164.compute-1.amazonaws.com) () [19:12:27] *** Quits: bwalker (~bwalker@134.134.139.72) (ZNC - http://znc.in)