*** elevenarms has joined #baserock | 00:22 | |
*** elevenarms has quit IRC | 01:15 | |
*** elevenarms has joined #baserock | 01:16 | |
*** elevenarms has quit IRC | 01:21 | |
*** elevenarms has joined #baserock | 01:43 | |
*** zoli__ has joined #baserock | 06:13 | |
*** zoli__ has quit IRC | 06:40 | |
*** sherm_ has joined #baserock | 07:32 | |
*** a1exhughe5 has joined #baserock | 08:05 | |
*** gfinney has joined #baserock | 08:06 | |
*** pacon has joined #baserock | 08:23 | |
*** mdizzle has joined #baserock | 08:39 | |
*** gfinney has quit IRC | 08:46 | |
*** gfinney has joined #baserock | 08:46 | |
*** gfinney has quit IRC | 08:50 | |
*** gfinney has joined #baserock | 08:51 | |
*** jonathanmaw has joined #baserock | 08:52 | |
*** bashrc has joined #baserock | 09:09 | |
*** bashrc has quit IRC | 09:10 | |
*** bashrc has joined #baserock | 09:10 | |
*** gary_perkins has joined #baserock | 09:18 | |
*** edcragg has joined #baserock | 09:32 | |
*** tiagogomes has joined #baserock | 09:35 | |
*** rdale has quit IRC | 09:44 | |
* Kinnison thinks a lot of our delta against busybox may no longer be needed | 09:49 | |
Kinnison | e.g. we no longer need the systemd units for networking thanks to networkd | 09:49 |
---|---|---|
*** pacon has quit IRC | 10:00 | |
*** ssam2 has joined #baserock | 10:06 | |
*** ChanServ sets mode: +v ssam2 | 10:06 | |
*** rdale has joined #baserock | 10:10 | |
mauricemoss_ | persia, the binutils patch got send in mid 2014 (not by me), but hasn't been applied yet. http://lists.gnu.org/archive/html/bug-binutils/2014-06/msg00017.html | 10:20 |
*** wschaller has joined #baserock | 10:23 | |
nowster | does anyone know why stage2-glibc might be doing this: | 11:03 |
nowster | drwxr-xr-x root/root 0 1991-08-24 23:00:00 lib/ | 11:03 |
nowster | lrwxrwxrwx root/root 0 1991-08-24 23:00:00 lib/lib -> /tools/lib/ | 11:03 |
richard_maw | lib/lib‽ it sounds like there's a `ln -s` somewhere, which hasn't checked that the target already exists and is a directory | 11:06 |
nowster | I can't see it in the definition... | 11:07 |
richard_maw | me neither, and I even had a poke at stage2-fhs-dirs :¬/ | 11:13 |
nowster | Got it... stage2-gcc | 11:13 |
nowster | install -d "$DESTDIR/lib" | 11:13 |
nowster | ln -s "$PREFIX/$libdir/libgcc_s.so" "$DESTDIR/lib/" | 11:13 |
nowster | ln -s "$PREFIX/$libdir/libgcc_s.so.1" "$DESTDIR/lib/" | 11:13 |
ssam2 | wow, good catch | 11:13 |
nowster | That probably should be going into tools/lib | 11:14 |
richard_maw | I saw that, I don't see how it's making a lib/lib -> /tools/lib there | 11:14 |
nowster | It's glibc that's doing that later on, I think, | 11:14 |
nowster | but not sure. | 11:15 |
nowster | Hmm. False positive? | 11:16 |
richard_maw | I recall we did need to make a symlink, but I don't recall where. | 11:16 |
*** tiagogomes has quit IRC | 11:25 | |
persia | mauricemoss_: Have you replied upstream reporting that it also works for you? | 11:31 |
mauricemoss_ | Not yet, will do. | 11:33 |
persia | If nothing else, it can get the conversation restarted. As with every project, patches against binutils sometimes don't get reviewed unless you review them :) | 11:36 |
nowster | I *think* it's in hardlink_all_files from stagingarea.py | 11:37 |
Kinnison | Speaking of patches -- persia: You had a patch for linux32 to support setting an armv7 personality on armv8 systems -- did that ever go anywhere? | 11:37 |
nowster | elif stat.S_ISLNK(mode): | 11:37 |
nowster | # Copy the symlink. | 11:37 |
nowster | if os.path.lexists(destpath): | 11:37 |
nowster | os.remove(destpath) | 11:37 |
nowster | os.symlink(os.readlink(srcpath), destpath) | 11:37 |
nowster | ...which will fail if there are files in the destpath | 11:37 |
persia | Kinnison: sadly no. I'll see if I can find where I put that and do something with it. | 11:38 |
persia | I hadn't realised anyone was using it. | 11:38 |
*** tiagogomes has joined #baserock | 11:38 | |
Kinnison | persia: I've a use-case which might benefit from it, so I'd love to take a look if you do find it | 11:38 |
nowster | ...and if /lib exists with files in it, it'll silently fail the os.remove() and make /lib/lib as a symlink. | 11:40 |
pedroalvarez | not tags if we use them as refs | 11:41 |
* pedroalvarez has answered a message from months ago | 11:42 | |
persia | Kinnison: Sure. | 11:42 |
* straycat didn't realise morph transformed cluster variables such as false to False | 11:42 | |
richard_maw | nowster: it won't silently fail the os.remove(), it'll raise an exception | 11:43 |
bashrc | is there a current devel system image download anywhere? It would save me a lot of messing around | 11:44 |
pedroalvarez | bashrc: http://download.baserock.org/baserock/ | 11:44 |
pedroalvarez | bashrc: http://wiki.baserock.org/download/ | 11:44 |
* pedroalvarez also points that devel-system artifacts are in cache.baserock.org | 11:46 | |
nowster | richard_maw: are you absolutely sure about that? | 11:46 |
petefoth | those appear to only be base images, with some work needed to make them into devel systems as described in http://wiki.baserock.org/quick-start/#index7h2 (which should be quick if the artifacts are cached) | 11:46 |
ssam2 | straycat: it's YAML doing so, I think | 11:47 |
bashrc | latest devel system appears to be 14.40 | 11:47 |
straycat | Is there a way I can do a kvm deploy of the rootfs to an lvm rather than a disk image on my file system? | 11:47 |
nowster | hmm | 11:47 |
richard_maw | nowster: 100%, unless it's been monkey-patched, which I don't believe we have done here | 11:47 |
straycat | sorry, i mean logical volume, not lvm | 11:47 |
ssam2 | bashrc: 15.02 is latest I think | 11:47 |
pedroalvarez | bashrc: we stopped releasing devel systems to release build systems | 11:48 |
straycat | ssam2, ahh | 11:48 |
ssam2 | oh yeah | 11:48 |
bashrc | right, but there's no devel system image for that | 11:48 |
straycat | i still don't understand why we stopped releasing devel systems | 11:48 |
ssam2 | limited resources | 11:48 |
straycat | ssam2, ahh | 11:48 |
pedroalvarez | the devel system artifact is almost 2G :/ | 11:48 |
bashrc | eek | 11:48 |
pedroalvarez | so yeah, better to release build systems | 11:49 |
bashrc | very well | 11:49 |
nowster | richard_maw: drat | 11:53 |
* straycat realises he still hasn't merged the kvm ext checks he added :/ | 11:56 | |
* Kinnison still has somewhere in the region of 400 mails in his baserock-dev mailbox "watched" for activity | 11:57 | |
nowster | the only place I can find something similar is in install-files.configure | 11:57 |
straycat | it might be cool to have a merge/bug squashing party | 11:58 |
Kinnison | They work for other projects, so why not? | 11:59 |
persia | We should probably schedule in advance, so people can find time. Maybe the 14th? | 12:04 |
pedroalvarez | hm.. genivi-common-api-runtime is failing to build because of a: configure: error: cannot find install-sh, install.sh, or shtool in "." "./.." "./../.." | 12:12 |
pedroalvarez | looks like autoreconf now generates a configure script that requires one of those files to build | 12:12 |
pedroalvarez | (I may be wrong) | 12:12 |
pedroalvarez | it also looks like creating an empty file called "install.sh" fixes the problem, but I'm not sure if this is the right way to solve it | 12:13 |
persia | It isn't | 12:15 |
persia | Generally that should be produced by `automake --add-missing` | 12:16 |
bashrc | just an observation based upon previous banter: keeping a current base and devel image for download on the baserock.org is probably more important than having a lot of old base images | 12:17 |
persia | If it doesn't work, then there's something funny about the autoconf installation. | 12:17 |
straycat | 14th works for me | 12:17 |
persia | bashrc: We don't have any mechanism for continuous push to download.baserock.org on commit yet. I believe that task is currently waiting on full implementation of pre-merge CI. | 12:18 |
*** jmalk has joined #baserock | 12:19 | |
bashrc | I was thinking in terms of disk space, because the devel image is such a monster some older base images could be culled to make room | 12:20 |
nowster | morph needs a -no-no-nanette option. ;) | 12:21 |
pedroalvarez | bashrc: afaik there are not disk-space problems | 12:21 |
pedroalvarez | build-systems are like devel systems but without things like ruby, or nodejs, which are needed for the baserock-import-tool | 12:22 |
bashrc | I expect in the majority of cases for visitors to the site, they're going to be wanting the devel image rather than the base | 12:24 |
persia | We used to just have devel, but people complained, which is why base was created. | 12:25 |
pedroalvarez | s/base/build/ | 12:25 |
bashrc | for a final deployment build makes sense, but if you want to do any amount of development... | 12:25 |
persia | Depends on the level of development you want to do. | 12:26 |
persia | Some folk use a build image, and only run git and morph, doing all their work in another environment. | 12:26 |
persia | Other folk like to have all their tools in a devel environment, and do the work locally. | 12:26 |
pedroalvarez | also, it's [well] documented how to upgrade your system to a devel-system | 12:27 |
pedroalvarez | In other news, `automake --add-missing` may need some patches in this component to work: http://paste.baserock.org/apafikimek | 12:27 |
* persia wonders if the process of downloaded cached chunks should be deserialised, for better use of bandwidth. | 12:27 | |
Zara | I'd also like devel systems to be downloadable, just to save myself time (I tend to work from a devel system), so if we have the resources for this in the future, it gets my vote. | 12:28 |
persia | pedroalvarez: Ugh. Maybe there's another way to cause autoconf to provide those files. | 12:28 |
persia | Zara: How often do you need to download a *new* devel system? | 12:29 |
bashrc | I agree with Zara that it's mainly to save time when getting started. Otherwise it's a longwinded process, and that is likely to put off newcomers | 12:33 |
bashrc | make the barrier to doing something interesting as low as it can be | 12:34 |
Zara | persia: Not that frequently, but more often if I need to do things quickly and I hit weird errors that nobody has come across. In those situations it's quicker for me to make a new devel vm than to troubleshoot (and if I then get the same errors, it narrows the causes down), but it's fiddly. | 12:34 |
nowster | still nibbling away at the stage2-glibc symlink problem. | 12:35 |
nowster | *) | 12:35 |
nowster | loader=$(basename $(ls "$DESTDIR$PREFIX"/lib/ld-linux*)) | 12:35 |
nowster | ln -s "$PREFIX/lib/$loader" "$DESTDIR/lib/$loader" | 12:35 |
nowster | ...what if that "ls" doesn't find anything? | 12:35 |
nowster | because ld-linux is actually ld.so.1 | 12:36 |
persia | Zara: I generally just upgrade, or deploy a new devel from my local devel. To avoid causing issues, I try to only use defined software (no use-local-morph, no local installs), and try to keep a known-good devel image as one of my installed system versions for use with system-version-manager. | 12:36 |
Zara | persia: I think this works if you're sure that your vm itself isn't broken in some way (eg: if you have a sudden power outage; my laptop likes to suddenly turn off every so often). I usually do upgrade, and I try to keep a working vm on hand, but the issues come when the image you're upgrading from won't let you upgrade because it's broken in some way. | 12:40 |
* nowster tries http://paste.baserock.org/yezejohosi | 12:40 | |
Zara | bashrc: agree that it would lower barrier to newcomers. the only reason I haven't really pushed this before is because I think it serves a dual purpose as a 'building and deploying on baserock' tutorial, where lots of people know what the end result should look like. | 12:41 |
persia | Zara: You should be able to boot an alternate system without fear if your primary system is corrupt from power loss, assuming you have access to the console to control the boot target, but yes, I concede the point :) | 12:41 |
nowster | I suggest http://paste.baserock.org/ubovevixoy (not a real u-diff) | 12:44 |
tiagogomes | the reason why there is a initramfs on clusters/hardware-deployment.morph is it because it was need by the installer, or because it was needed to boot the deployed system in the hardware (e.g. drivers needed) | 12:50 |
Zara | persia: :) yeah, I'm sure there are ways around it. I'm just weighing in because I've wished there was a download before, and someone said the images were removed because they were unpopular. | 12:50 |
Zara | so now you know where I stand! :P | 12:50 |
nowster | possibly better: http://paste.baserock.org/uxupuqiquv | 12:50 |
persia | Zara: I think I'll adopt the position that it is a tutorial, and that morph doesn't parallelise cache downloads is a bug. That lets me cogently argue that not having devel images for download is to save the developer time. | 12:52 |
petefoth | And what care we for truth, when we can argue cogently? ;) | 12:52 |
* ssam2 still hopes that 'ostree pull' is good enough for artifact fetching | 12:53 | |
pedroalvarez | tiagogomes: the installer system doesn't have initramfs, is t he system to install that has initramfs | 12:53 |
* petefoth has cause to swear at StroyBoard's half-implmeneted UI :( | 12:54 | |
persia | ssam2: Does that parallelise? The main issue is that the download speed is limited by the latency for each packet, because of how TCP works, so if there are multiple connections with simultaneous traffic, the aggregate time-to-complete is lower. | 12:54 |
tiagogomes | pedroalvarez that's what I needed to know, ta | 12:55 |
*** wschaller has quit IRC | 12:56 | |
ssam2 | persia: it shares the content of files between the different artifacts | 12:57 |
ssam2 | i suppose that's a separate problem, though | 12:57 |
ssam2 | our artifacts don't really overlap in content... | 12:57 |
ssam2 | hopefully it also parallelises the transfer of those files, but I don't know | 12:58 |
pedroalvarez | persia: thanks for your help, Turns out that autoreconf was trying to generate that file but in build-aux/ | 13:03 |
pedroalvarez | and ./configure looks for them in . .. ../.. | 13:04 |
persia | Frustrating. I suspect something wrong with configure.ac or similar. | 13:04 |
richard_maw | nowster: ooh, that does sound plausible if ls is failing to signal failure with `set -e` set. | 13:05 |
nowster | busybox ls doesn't! | 13:06 |
persia | It's running in a subshell. I think I remember that not carrying the flags (-x, -e, etc.) | 13:07 |
nowster | ah right! | 13:07 |
nowster | of course. | 13:07 |
nowster | I was wrong about busybox | 13:07 |
nowster | it does DTRT | 13:08 |
nowster | yay! It works! | 13:20 |
nowster | drwxr-xr-x root/root 0 1991-08-24 23:00:00 lib/ | 13:20 |
nowster | lrwxrwxrwx root/root 0 1991-08-24 23:00:00 lib/ld.so.1 -> /tools/lib/ld.so.1 | 13:20 |
richard_maw | \o/ | 13:26 |
*** fay has joined #baserock | 13:32 | |
*** fay is now known as Guest75056 | 13:32 | |
*** fay_ has quit IRC | 13:33 | |
pedroalvarez | ok, weirdest error of the day: http://paste.baserock.org/jahelehixu | 13:39 |
pedroalvarez | At first glance I thought that `make` was doing weird things with $DESTDIR, so I decided to remove it before calling make | 13:40 |
pedroalvarez | but.. this is building libarchive, and the error says: grep: /attr.inst/usr/lib64/libattr.la: No such file or directory | 13:40 |
pedroalvarez | ok, I think I found the error: | 13:43 |
pedroalvarez | /usr/lib64/libattr.la:libdir='/attr.inst/usr/lib64' | 13:43 |
pedroalvarez | /usr/lib64/libacl.la:dependency_libs=' /attr.inst/usr/lib64/libattr.la' | 13:43 |
pedroalvarez | (running ` grep -r attr.inst /`) | 13:43 |
*** wschaller has joined #baserock | 13:44 | |
richard_maw | I've heard it recommended to just delete .la files. It was in the context of relocating them, but apparently it doesn't matter, since you can build fine without them. | 13:45 |
*** Guest75056 is now known as fay_ | 13:45 | |
pedroalvarez | richard_maw: do you mean I can just remove that file when installing attr? | 13:47 |
richard_maw | I've heard that recommended, since you don't need the .la files to find shared libraries. AIUI the libtool stuff is to provide compatibility of build-system between platforms where libraries work less well than on Linux | 13:49 |
*** sherm_ has quit IRC | 13:50 | |
persia | Another reason is for bundling libraries with applications for binary installation. | 13:51 |
* pedroalvarez removes the .la file at the end of the attr.morph definitions | 13:52 | |
pedroalvarez | -s | 13:52 |
pedroalvarez | oh, weird things are happening in attr | 13:58 |
pedroalvarez | is installing things in $DESTDIR$DESTDIR | 13:58 |
pedroalvarez | fixing that will fix this error :) | 13:58 |
pedroalvarez | but then I wonder why I don't have a /attr.inst folder in my system :/ | 13:59 |
pedroalvarez | fixed, will send a patch soon | 14:04 |
*** sherm_ has joined #baserock | 14:14 | |
*** fay_ has quit IRC | 14:33 | |
*** fay has joined #baserock | 14:34 | |
*** fay is now known as Guest30078 | 14:34 | |
*** tiagogomes has quit IRC | 14:41 | |
bashrc | any ideas what might be causing this: /tmp/morph_tmp requires 4000000000 bytes free, has 1048719360 | 14:43 |
persia | Check /etc/morph.conf to make sure you've told morph not to use /tmp as a temporary directory. | 14:44 |
persia | Or create a much larger /tmp | 14:44 |
Kinnison | Also suggest better wording for that message since you clearly didn't understand it immediately :( | 14:44 |
bashrc | there doesn't appear to be an /etc/morph.conf | 14:44 |
Kinnison | Then you've not yet finished configuring morph according to the recommended process | 14:45 |
Kinnison | the wiki has instructions | 14:45 |
persia | (but usually people reporting this don't have /etc/morph.conf) | 14:46 |
*** Guest30078 is now known as fay_ | 14:47 | |
persia | http://wiki.baserock.org/quick-start/ | 14:47 |
persia | http://wiki.baserock.org/quick-start/#index4h2 , specifically | 14:48 |
straycat | given how often we get this question, we should probably append the question we usually ask the reporter of this problem to the error message | 14:48 |
straycat | "Have you set tempdir in your morph.conf?" | 14:48 |
persia | +3 | 14:49 |
Kinnison | I'd rather "Is tempdir set appropriately in your morph.conf?" | 14:51 |
Kinnison | since morph can trivially answer the question "have you set it?" | 14:51 |
persia | If we're being nice, maybe suggest creating morph.conf if it is absent? | 14:51 |
pedroalvarez | bashrc: you had the same problem some time ago: http://irclogs.baserock.org/%23baserock.2014-11-19.log.html :) | 14:54 |
Zara | straycat: I'd go for 'Have you set tempdir in /src/morph.conf?'; the path *could* be different, but I think this normally happens to people who are trying to follow the wiki guide and accidentally miss a step | 14:54 |
Zara | so the path might be useful... but ymmv. | 14:55 |
*** tiagogomes has joined #baserock | 14:55 | |
*** wschaller has quit IRC | 14:59 | |
*** wschaller has joined #baserock | 15:07 | |
nowster | /tmp/morph_tmp is the default | 15:12 |
bashrc | so I've been setting up morph and one thought which occurs is why isn't this stuff just a command. "morph setup", or something like that | 15:13 |
bashrc | possibly a lot of the boilerplate might be similarly consolidated into commands or command options | 15:14 |
bashrc | when trying to init the morph workspace I'm getting: No module named pylru | 15:19 |
straycat | you need pylru now | 15:19 |
SotK | Are you using latest morph or the one that came in your image? PyLRU should be in systems which have that dependency by default. | 15:20 |
bashrc | probably the quick start guide needs updating for that | 15:21 |
ssam2 | SotK: we haven't done a release of the 'build' system with PyLRU in it yet | 15:21 |
SotK | aha | 15:21 |
ssam2 | i tried a couple of times, but the yak kept growing more hairy, so to speak | 15:21 |
SotK | the release should still have an old enough morph in that this doesn't happen by default though right? | 15:21 |
bashrc | SotK: just following the "no frills" guide | 15:21 |
ssam2 | ...maybe pedroalvarez can do one this week :) | 15:21 |
pedroalvarez | mmmmmmkay | 15:22 |
pedroalvarez | :) | 15:22 |
SotK | bmottram: oh, that seems to duplicate the content of using-latest-morph | 15:23 |
SotK | I was unaware of that when I updated the using-latest-morph page to warn of this | 15:23 |
* petefoth sighs | 15:23 | |
* SotK wonders why we have four different "Getting started" guides | 15:24 | |
pedroalvarez | indeed, too many | 15:24 |
petefoth | Because we have no spec for Baserock docs, anybody can edit the wiki and nobody is tasked with reviwing any changes | 15:24 |
petefoth | So we will get 'bugs' | 15:25 |
SotK | two of them appear to be instructions on copy/pasting some things into a script and running it... should we not just provide scripts for people to run at this point? | 15:26 |
* petefoth walks away, muttering darkly... | 15:27 | |
pedroalvarez | hehe | 15:27 |
pedroalvarez | people should understand what are they doing when setting things up | 15:27 |
SotK | I agree, but providing a copy/paste page seems just as bad as a script in that regard (to me at least) | 15:28 |
petefoth | pedroalvarez: yes - and they can understand by reading w.b.o... oh wait! | 15:29 |
Kinnison | If we ever get to 'curl http://wiki..... | bash' I will explode | 15:29 |
bashrc | fixed the nofrills | 15:29 |
pedroalvarez | Kinnison: never | 15:29 |
SotK | bashrc: thanks :) | 15:29 |
petefoth | bashrc: ta! | 15:30 |
rjek | Kinnison: https://twitter.com/dotwaffle/status/572735066390048769 | 15:30 |
Kinnison | heh | 15:31 |
petefoth | Does anything need adding to other pages? | 15:31 |
SotK | the kvm script page looks ok | 15:32 |
petefoth | Will the steps in http://wiki.baserock.org/quick-start/ work? | 15:33 |
SotK | yep | 15:33 |
SotK | only things which duplicate the content of http://wiki.baserock.org/using-latest-morph/ will be broken | 15:33 |
SotK | (I didn't think anything did) | 15:34 |
petefoth | SotK: so no need to add anything there abotu pylru? | 15:34 |
petefoth | SotK: OK thanks | 15:34 |
SotK | nope, I added it to using-latest-morph when I merged the change to morph, and added pylru to morph's dependencies when I updated the morph sha in definitions | 15:35 |
*** gary_perkins has quit IRC | 15:35 | |
*** gary_perkins has joined #baserock | 15:37 | |
persia | Can we make using-latest-morph harder to find? I think it causes more problems than it solves, for most users. | 15:43 |
SotK | I think we should do releases more often if we do that | 15:45 |
straycat | We could have a script that periodically moves it to different urls, like the staircases in hogwarts. | 15:45 |
petefoth | persia: it's a wiki, you can do whatever you want :) | 15:45 |
Zara | hm, generally when I come on here with an error, the most common fix after 'you need to update your devel vm' is 'you need to update morph' | 15:46 |
persia | petefoth: Yes, but if I just remove the page (which is what I want), someone will surely put it back. | 15:46 |
persia | Zara: Surely that's an indicator that we're doing a bad job: if we need newer morph for something, and we recommend that something, we should be putting that newer morph in latest definitions. | 15:47 |
petefoth | persia: Maybe, but if you feel stringly and can justfiy your decision, then maybe not | 15:47 |
petefoth | s/stringly/strongly/ | 15:47 |
persia | That's why I ask here? | 15:47 |
persia | s/?/./ | 15:48 |
SotK | newer morph in definitions is useless if its nearly two months since the last images were released I feel | 15:48 |
petefoth | persia: I would have no obhjection | 15:48 |
persia | SotK: Why? One just upgrades one's devel to latest. | 15:48 |
SotK | What about people who don't have an existing devel machine? | 15:49 |
persia | Download a build machine, build the latest devel machine. | 15:49 |
SotK | Then people will question why some old bug is in the new machine they just downloaded though maybe? | 15:51 |
tlsa | a few weeks ago I found it was impossible to build a baserock system from master definitions with the morph supplied in the system available for download on baserock.org | 15:51 |
tlsa | because that morph didn't understand armv8be or somesuch | 15:51 |
persia | tlsa: I thought we were solving that with the definitions version thing, and that we never expected that to happen again. | 15:51 |
tlsa | perhaps | 15:52 |
tlsa | you would need to make a new release though | 15:52 |
persia | SotK: We can update the build machine occasionally. Running *any* software that is not from definitions strikes me as not in line with the point of Baserock. | 15:52 |
tlsa | whcih contains a morph that checks version | 15:52 |
persia | tlsa: Yes. One more release, and we can be done. | 15:52 |
pedroalvarez | ha! | 15:53 |
SotK | persia: One more release works I guess. And another release each time we change the definitions format? | 15:53 |
persia | Why? | 15:53 |
persia | Surely once we have versioned defintions, we'll be able to have a migration path. | 15:53 |
SotK | Do we intend for old morph to be able to build new definitions? | 15:53 |
persia | Well, no. | 15:54 |
persia | hrm. | 15:54 |
tlsa | persia: that will improve the error message (your morph is too old for these defintitions, instead of something obscure) | 15:54 |
tlsa | but not fix the need for updated morph | 15:54 |
persia | So, yes, one release per definitions version ought do it. | 15:54 |
* SotK wouldn't like to find Baserock in a year or two's time, download the build system and immediately have to go through a really long migration path before its up to date | 15:54 | |
tlsa | SotK: snap | 15:54 |
pedroalvarez | persia: +1 | 15:55 |
SotK | persia: +1 | 15:55 |
pedroalvarez | adding a new architecture makes our definitions change its version? | 15:55 |
* persia credits SotK with the suggestion | 15:55 | |
persia | pedroalvarez: No, but does that require a release? | 15:56 |
SotK | Can we not separate architecture from morph? | 15:56 |
persia | I'd like morph not to understand architectures | 15:56 |
pedroalvarez | persia: well, if we add a new architecture, the current release won't be able to build those systems | 15:56 |
pedroalvarez | erm | 15:57 |
pedroalvarez | yeah, that would solve my problem | 15:57 |
persia | pedroalvarez: Only because we've hardcoded yet another architecture naming scheme in morph. | 15:57 |
pedroalvarez | indeed | 15:57 |
persia | We can't do it for version 0, but for definitions version 1, I'd hope we could declare architectures in definitions. | 15:58 |
* pedroalvarez is imagining a definition file with architecture information, making things easier when adding new architectures | 15:58 | |
persia | When you finish imagining, please send to the list, to help guide the folk who eventually define definitions version 1 | 16:00 |
kejiahu | I am trying to net boot baserock on moonshot nodes, but I got the error message while booting http://paste.baserock.org/ujiqihuyiw and it only boot into emergency mode. does anybody have a idea what may cause these happens | 16:08 |
SotK | My OSTree work breaks a whole bunch of the cmdtests since they do things like look for chunk artifacts in the artifact cache directory, how should I solve this given that the chunk name is no longer in the artifact "filename" (a ref in an OSTree repo)? | 16:12 |
Kinnison | I guess you get to convert the useful ones to yarns :) | 16:13 |
SotK | lucky me! :) | 16:13 |
* SotK wonders how to interrogate the cache from outside morph without knowing the cache key | 16:14 | |
persia | You also get to purge the useess ones :) | 16:14 |
pedroalvarez | kejiahu: last time I saw that, the system was trying to mount a disk that didn't exist | 16:17 |
pedroalvarez | kejiahu: could you check your /etc/fstab ? | 16:17 |
kejiahu | pedroalvarez, that's what I thought at first, but the /etc/fstab seems ok. | 16:19 |
kejiahu | pedroalvarez, I can't show it now, as the machine is un-reachablt for me now | 16:20 |
ssam2 | SotK: perhaps the trick is to add a command to morph that calculates a cache key | 16:25 |
ssam2 | list-artifacts already does that, but maybe gives output in an unhelpful format | 16:26 |
pedroalvarez | kejiahu: well, we can figure out. Did you set in the cluster morphology of that system any "FSTAB_*" option? | 16:27 |
kejiahu | pedroalvarez, yes, it was set in the cluster file, /dev/sda2 /srv/distbuild auto defaults,rw,noatime 0 2 | 16:31 |
kejiahu | pedroalvarez, /dev/sda2 is a local disk with pre-installed system which I can choose to boot from. but I boot into baserock through net booting | 16:32 |
Kinnison | kejiahu: can you paste the full boot including all the kernel messages into the pastebin? | 16:34 |
edcragg | pedroalvarez kejiahu Kinnison this is a more complete log of the situation, http://paste.baserock.org/yehojakiha | 16:34 |
Kinnison | that looks different | 16:35 |
kejiahu | ed's log may different from mine, as I was trying to boot a distbuild system, and he is booting a devel baserock system | 16:35 |
Kinnison | your log is a failure to find a disk | 16:36 |
Kinnison | ed's log seems to contain a segfaulting journald | 16:36 |
Kinnison | which is quite worrisome in a different way | 16:36 |
kejiahu | Kinnison, mine contains similar systemd failures as ed, I'll get the full log soon | 16:39 |
edcragg | this is simultaneously affecting any systems booted from separate rootfs's on the same nfs server | 16:40 |
pedroalvarez | could this problem be related to what we tried to fix in nfsboot.configure? | 16:43 |
*** sherm_ has quit IRC | 16:43 | |
rdale | i'm getting a syntax error from busy box on a configure script - i'm not sure what to do about that | 16:43 |
persia | rdale: Could you paste the error? | 16:44 |
kejiahu | http://paste.baserock.org/fozabesota here is the full log with kernel messages | 16:46 |
pedroalvarez | kejiahu: that one looks like "[ TIME ] Timed out waiting for device dev-sda2.device." is the error | 16:48 |
pedroalvarez | kejiahu: could you try to remove /dev/sda2 from /etc/fstab and see if that fixes it? and after that check why is not mounting /dev/sda2? | 16:49 |
CTtpollard | that seemed to fix my sda2.device timeout error the other day | 16:50 |
rdale | persia: this the part of the shell script with the error: http://paste.baserock.org/wagohosina | 16:51 |
kejiahu | pedroalvarez, yes, I am going to do that, it taks a while... | 16:52 |
pedroalvarez | kejiahu: why? | 16:52 |
persia | rdale: Which line in the paste is line 13062? | 16:52 |
rdale | it is before the snippet i pasted, but 'Checking lzma' is appearing in the build log and that is after the line with the error | 16:53 |
kejiahu | pedroalvarez, it's within moonshot chassis, and all access can only be done through iLO and virtual serial port, and today the nodes response quite slow for some reason. | 16:54 |
rdale | persia: if i remove all the code in the snippet i pasted, i don't get an error - so it is in there somewhere | 16:56 |
pedroalvarez | kejiahu: wasn't the system nfsbooting? can you change the /etc/fstab file in the nfs server and reboot? | 16:56 |
kejiahu | pedroalvarez, yes the nfs server was in the same chassis, and I have to access it through the some routine | 16:58 |
pedroalvarez | argh! :) | 16:58 |
*** a1exhughe5 has quit IRC | 16:59 | |
kejiahu | pedroalvarez, http://paste.baserock.org/nowesikahu it still gets those faults, excpet the one for /dev/sda2 | 17:01 |
*** sherm_ has joined #baserock | 17:02 | |
rdale | persia: busybox doesn't like PKG_CHECK_MODULES() in the configure script | 17:03 |
*** gfinney has quit IRC | 17:03 | |
pedroalvarez | kejiahu: hm... "Failed to set file attributes: Inappropriate ioctl for device"? | 17:04 |
kejiahu | pedroalvarez, I feel it's more like a problem of netboot server rather than local hardware, as ed got similar problem on all the rootfs which used to be working before today, but corrupted today | 17:06 |
pedroalvarez | kejiahu: looks like a known bug, but I don't know if that is the root cause | 17:06 |
pedroalvarez | maybe you were using before systemd 218 and now you are using systemd 219? | 17:07 |
pedroalvarez | if that's the case, then this may help: https://www.mail-archive.com/search?l=systemd-devel@lists.freedesktop.org&q=subject:%22%5Bsystemd-devel%5D+%5BPATCH%5D+journal%3A+fix+Inappropriate+ioctl+for+device%09on+ext4%22&o=newest&f=1 | 17:07 |
kejiahu | pedroalvarez, but I didn't make any upgarding to my devel system.. | 17:09 |
pedroalvarez | kejiahu: oh, I thought it was a new deployment. So same system was working and now it doesn't? | 17:10 |
kejiahu | pedroalvarez, so this issue happens on both new deployed system, and old deployed ones. and they wera all on same nfs server | 17:13 |
pedroalvarez | I'm out of ideas then :/ | 17:16 |
kejiahu | pedroalvarez, thanks for your help anyway. seems tiago got the idea | 17:17 |
tlsa | root | 17:21 |
edcragg | pedroalvarez: a trove deployed in the moonshot was interfering with the existing netboot server | 17:22 |
pedroalvarez | hm.. true.. troves have a nfs server | 17:23 |
persia | Hrm? Doesn't the cluster declaration of kernel command line identify the nfs server from which to boot? | 17:26 |
*** mdizzle has quit IRC | 17:36 | |
*** ssam2 has quit IRC | 17:43 | |
*** sherm_ has quit IRC | 17:49 | |
*** bashrc has quit IRC | 17:55 | |
*** elevenarms has quit IRC | 18:00 | |
*** elevenarms has joined #baserock | 18:02 | |
*** wschaller_ has joined #baserock | 18:02 | |
*** jonathanmaw has quit IRC | 18:04 | |
*** wschaller has quit IRC | 18:05 | |
pedroalvarez | jjardon: is your +1 only for libarchive? or also for persistence-* lorries? | 18:05 |
jjardon | pedroalvarez: it was for libarchive, but +1 for the other as well (you already noticed about the indentation) | 18:09 |
pedroalvarez | jjardon: yay! thanks | 18:10 |
*** tiagogomes has quit IRC | 18:17 | |
*** edcragg has quit IRC | 18:18 | |
*** edcragg has joined #baserock | 18:19 | |
*** edcragg has quit IRC | 18:24 | |
*** wschaller_ has quit IRC | 18:34 | |
straycat | AttributeError: 'dict' object has no attribute 'startswith' | 18:47 |
straycat | oh ansible, it was all going so well... | 18:47 |
SotK | :( | 18:47 |
straycat | earlier i got an error message detailing exactly what i'd done wrong and exactly what i should do to fix it, it was very good | 18:48 |
straycat | oh i suck | 18:50 |
SotK | accidental yaml dict? | 18:51 |
straycat | including a var definition under vars_files | 18:51 |
straycat | *included | 18:51 |
SotK | :) | 18:51 |
* SotK spams the mailing list and heads home | 18:51 | |
franred | ummm, I break morph....http://paste.baserock.org/ovogutizec -- compiling baserock/openstack-v4 | 19:04 |
persia | Oh wow. | 19:06 |
persia | It is admittedly a complex system, but still, we should be able to deal with that. | 19:07 |
franred | I've tried to increase the recursion limit but it does not fix it, so I wonder if we create a loop in some part :S | 19:08 |
persia | The thing that confuses me is that it looks like only seven layers, which doesn't seem particularly deep | 19:08 |
franred | persia, I cut the log, it is much bigger, sorry to not make the "<snip>" more visible | 19:09 |
franred | s/much// | 19:09 |
persia | Ah, I see line 59 now. | 19:09 |
persia | Do you know how many layers it has? | 19:10 |
franred | 168, if I haven't checked it wrong | 19:15 |
persia | That seems excessive. | 19:19 |
persia | Might be useful to have a loop detector, as I suspect that is the problem. | 19:19 |
persia | Maybe add a print statement that reports *which* component is being analysed along the way? That way there is a handy list of the candidates for the loop (and maybe we can see the loop). | 19:20 |
franred | persia, Im doing that :) | 19:21 |
franred | I need to come back home, I will continue debugging this tomorrow | 19:29 |
franred | persia, looks like it get annoyed because 2 strata includes the same chunk - for example I had pycrypto in openstack-common and in openstack-services and openstack-services includes openstack-common | 19:35 |
franred | when removed pycrypto from openstack-services it gets annoyed for the next common chunk | 19:36 |
persia | Ah, yes, of course. | 19:36 |
persia | Because you have pycrypto build-depending on openstack-common, which build-depends on pycrypto, etc. | 19:36 |
persia | Generally speaking, you want to have each chunk in only one place. | 19:37 |
persia | If you need them to be different for some reason, then identify them differently using @ | 19:37 |
persia | (grep @ in definitions for some examples) | 19:37 |
franred | yep, but the error is pretty awful - we should say something like you have duplicated your chunk in A and B | 19:37 |
persia | But if you do have them be different, try really hard not to include both in a system. | 19:37 |
persia | Indeed. I like the report ybd produces about duplicates. | 19:38 |
persia | (don't use ybd to build: it doesn't yet, but it does produce nice reports about definitions) | 19:38 |
franred | persia, I don't need them duplicated, I moved out of openstack-services because was common with others strata | 19:38 |
persia | Oh, excellent. So it's just cleanup to remove the duplicates. | 19:39 |
franred | yep | 19:39 |
persia | Cool. | 19:39 |
straycat | iirc there's a TODO to add cycle detection, we've just been busy with everything else morph has to do | 19:47 |
*** rdale_ has joined #baserock | 19:48 | |
straycat | come to the bug squashing party and add cycle detection to morph! | 19:48 |
*** rdale has quit IRC | 19:49 | |
*** rdale_ has quit IRC | 19:50 | |
franred | ok, looks like I managed to clean all the duplicated chunks so the cycle has disappeared | 19:51 |
*** rdale has joined #baserock | 19:54 | |
persia | And it builds now? | 19:56 |
*** gary_perkins has quit IRC | 20:06 | |
pedroalvarez | Given that he has sent a patch, I'd say 'yes' | 20:30 |
*** rdale_ has joined #baserock | 20:41 | |
*** rdale has quit IRC | 20:44 | |
*** chiru has joined #baserock | 20:48 | |
*** zoli__ has joined #baserock | 21:57 | |
*** chiru has quit IRC | 22:01 | |
*** zoli__ has quit IRC | 22:18 | |
*** zoli__ has joined #baserock | 23:05 | |
*** zoli__ has quit IRC | 23:28 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!