**** BEGIN LOGGING AT Mon Apr 03 03:00:02 2017 Apr 03 08:11:54 hi. i'm putting some finishing touches on an embedded distro based on yocto. one thing i would want is to have the rootfs read-only with a second, writable partition with user data. Apr 03 08:13:45 what is the most common way to handle writable regions? my current idea is to mount the writable partition at /mnt/userdata and make all directories/files that must be writable links to that mount point Apr 03 08:14:35 ionte: That makes sense. You could also mount /var as writable and link any other writable directories under there Apr 03 08:14:52 paulbarker good idea Apr 03 08:15:27 also, where would you put that link creation? i mean in which recipe? Apr 03 08:15:59 probably the image recipe Apr 03 08:16:29 look at ROOTFS_POSTPROCESS_COMMAND if you haven't already Apr 03 08:17:33 nice, thanks! Apr 03 08:18:58 hm. if i decide to mount the writable partition as /var, how should i make sure the required subdirectories and files are created there? by default they are created in the root filesystem tarball... Apr 03 08:20:49 perhaps i should just put that creation outside of yocto, in my deploy script... Apr 03 08:53:07 Son_Goku: yes, and your point is? Apr 03 08:56:18 What is the relationship between SRCREV ?= "${AUTOREV}" and PV="...${SRCPV}". What significance has the AUTOREV and the SRCREV variable in this? Apr 03 08:56:48 The manual sais "We need to document AUTOREV and SRCREV_FORMAT here."... Apr 03 08:58:12 sveinse: which manual is this? Apr 03 08:59:11 jku: Bitbake user manual, section 4.4. version 2.2 Apr 03 09:00:02 sveinse: current dev-manual and reference manual do talk about autorev a bit Apr 03 09:01:44 e.g. http://www.yoctoproject.org/docs/latest/ref-manual/ref-manual.html#var-AUTOREV Apr 03 09:04:14 What follows what wrt SRCREV? If I have rev=.. in SRC_URI, does that imply that AUTOREV contains this rev, or is setting SRCREV overriding any rev=... statements in SRC_URI? Apr 03 09:06:35 Son_Goku: we'll update libdnf to latest version later, right now is a pre-release freeze Apr 03 09:08:28 Grepping oe and poky, it seems SRCREV is a specifier of which version to use from the SCM, where AUTOREV is some kind of "pick latest" identifier. So rev should not be used in SRC_URI if SRCREV is used, right? Apr 03 09:12:05 When do you use the "python pyfn () {" vs "def pyfn():" syntax? I see a mix of both types Apr 03 09:49:52 Is there a way to override this: ERROR: ParseError in ../sp/yocto/classes/../../VERSION: not a BitBake file from a include or require statement? The file contains a SP_VERSION="1.3.0" that I'd like read into the recipe. Otherwise I'd have to write a py function around it Apr 03 09:58:15 so i'm trying to run matchbox-keyboard + iceweasel on debian stable. but it seems like firefox reports password-fields incorrectly and so matchbox-keyboard doesn't pop up. its working fine in chromium. is there a way to work-around that inside matchbox-keyboard? Apr 03 09:59:01 d3xter: probably not, likely a bug in how firefox exposes the keyboard to the IM plugin. Apr 03 09:59:17 feel free to delve into the plugin though, maybe there's a case that isn't being handled Apr 03 10:00:03 rburton: ok thanks, will give it a try :) Apr 03 10:03:19 rburton: oh and another question. i've tried configuring matchbox-keyboard-0.1.1 on debian, but i get "./configure: line 13504: syntax error near unexpected token `FAKEKEY,' ./configure: line 13504: `PKG_CHECK_MODULES(FAKEKEY, libfakekey,,". what would be an easy fix for that? Apr 03 10:27:40 d3xter: do you have pkg-config installed? Apr 03 10:45:33 hm wic.filemap.ErrorNotSupp: FilemapFiemap: the FIEMAP ioctl is not supported by the file-system Apr 03 10:47:29 ed2: can fiemap be optional? Apr 03 10:47:43 my tmp is a tmpfs Apr 03 10:47:56 and that's not an uncommon environment Apr 03 10:51:35 rburton: Do you have this patch applied: http://lists.openembedded.org/pipermail/openembedded-core/2017-March/134764.html ? Apr 03 10:52:09 rburton: just guessing here if SEEK_HOLE is supported by tmpfs. Apr 03 10:53:17 rburton: If not - I'll make FIEMAP optional. If it's not supported then wic will just copy a file without preserving sparceness. Apr 03 10:54:29 yes Apr 03 10:54:54 rburton: looks like SEEK_HOLE/SEEK_DATA is supported. Please, revert that patch then. Apr 03 11:01:16 jku: silly me, thank you Apr 03 11:06:02 rburton: I'll send a patch. sorry for this. Apr 03 11:07:13 np Apr 03 11:07:26 obviously i didn't actually test those patches before they landed in next. Apr 03 11:07:33 maybe we need a tmpfs run on the AB somewhere Apr 03 11:08:02 rburton: I'll add test case to test both methods: FIEMAP and SEEK_HOLE Apr 03 11:08:33 rburton: I thought they're both supported everywhere and SEEK_HOLE is not used at all. Apr 03 11:09:26 hm a post from 2012 suggests that tmpfs has seek_hole in Apr 03 11:10:02 oh got it backwards Apr 03 11:57:27 kanavin: it's an informational fix Apr 03 11:57:31 it's already at 0.8.x Apr 03 12:15:52 Is it possible to (force) include a non-bitbake file into a recipe? What defines a bitbake file, btw? Its suffix? Apr 03 12:17:24 I'd like to read a VERSION file, which is formatted exactly as a bb file; # for comments and KEY="data" Apr 03 12:18:12 sveinse: I don't think so no, but you could read it in without too much trouble within an anonymous python function though Apr 03 12:18:38 bluelightning: yeah, sure, reading it from py is no problem Apr 03 12:18:49 bluelightning: why wouldn't an include/require work? Apr 03 12:18:57 er Apr 03 12:19:06 actually yes include/require should work Apr 03 12:19:13 I get ParseError in ../sp/yocto/classes/../../VERSION: not a BitBake file Apr 03 12:19:28 hmm Apr 03 12:19:41 Perhaps I need to add it as a layer file? Apr 03 12:20:47 ah, no... I see it now in bitbake/lib/bb/parse/parse_py/bbhandler.py Apr 03 12:21:12 I confirm adding it to BBFILES did not help Apr 03 12:21:17 it explicitly specifies .bb, .bbclass, .inc as the extensions it handles Apr 03 12:21:22 right Apr 03 12:42:04 What is the second argument to d.getVar() ? Apr 03 12:43:12 whether to expand or not Apr 03 12:43:22 ie do you get ${PN} or somerecipename Apr 03 12:43:32 got it, thanks Apr 03 12:48:26 I think I need help with fixed-distro-features-for-native ... RP: anyone you'd suggest I talk to? or do you want to give some pointers on what to try ? Apr 03 12:50:52 Is there a list or hash in py for all variables readable by d.getVar()? I'm trying to help myself to figure out if there are any variables that might hold the root path of the current layer Apr 03 12:53:23 Specifically, how to read '../../VERSION' from a py fn in a bbclass file. current directory is the build dir and the environment does not give any clues, so I must rely on a hint from a bb variable to the abs location of this file. Apr 03 12:55:12 for my first q: d.keys() Apr 03 13:00:12 sveinse: layer directory isn't available AFAIK. THISDIR should be the directory of the currently parsed recipe... that said, are you sure what you are doing is a good solution to whatever your problem is? Apr 03 13:00:44 Hi guys, Apr 03 13:01:20 I was wondering if anyone has a recipe that does the installation of .deb files? Apr 03 13:02:07 jku, well it depends :P, I simply want to read the VERSION file which is located in the root of the layer. You see PV depends on this file Apr 03 13:04:14 our layer is somewhat special, as it blends the actual application repo and its meta layer. And the application targets many other platform than Yocto, so I cannot make this "bitbakeized" Apr 03 13:05:37 targetting non-yocto platforms doesn't prevent most software from having sensible version numbers but I'll take your word for it Apr 03 13:06:52 This is how I ended up by building the sources without using SRC_URI, btw, https://bpaste.net/show/8c1e7ff3bfc3 - it seems to be working. Anyone see anything obviously errors here? Apr 03 13:13:24 sveinse: are you aware that bitbake will have no idea about your sources when you do that? Apr 03 13:14:16 so can't figure out when to re-build Apr 03 13:15:20 the change of SRCPV might be enough Apr 03 13:15:22 *might* Apr 03 13:15:22 jku: How is this any different from SCM based rebuild? Is it hashing every file? Apr 03 13:16:01 sveinse: how will it know to rebuild if you have local changes? or is that not a concern Apr 03 13:16:34 sveinse: e.g. with git it has a git commit hash Apr 03 13:16:46 which is calculated for: On the build server, there is never any local edits of the sources, so it will get a new SRCPV when it changes, and this seems to work Apr 03 13:17:41 I've read the fetch2 py sources, and I do somewhat the same thing here. Because SRCPV will change due to hg's (hash)id will also change on source changes. Apr 03 13:17:42 innovative Apr 03 13:18:08 sveinse: are local modifications to the sources out of scope here? Apr 03 13:18:36 https://bugzilla.yoctoproject.org/show_bug.cgi?id=11288 Apr 03 13:18:37 Bug 11288: normal, Undecided, ---, richard.purdie, NEW , X Server fails to start on qemuarm64 Apr 03 13:18:55 partly. A limitation that is acceptable. Note that hg id will return a suffix of '+' if its locally edited, so it will work once for any local edits Apr 03 13:19:41 maxin: reassign to bruce? Apr 03 13:20:17 rburton: looks like kernel configs missing.. yes Apr 03 13:21:04 The reason for this scheme is that we want to have the recipe located in the same repo as the source code, a use case which is somewhat on the side of bitbake thinking. We used to have the metas in a separate repo, but it was very unpopular, as devs had to update two repos all the time. Apr 03 13:22:58 The repo is 800Mb, so doing SRC_URI and file:// leads to an unessesary copy of the sources into build/tmp -- which qmake never needs copied as it locates build output separate of the sources anyways. Apr 03 13:25:08 Is there a THISDIR that works for bbclass files? THISDIR reports the path to the calling bb recipe even for code inside a class file Apr 03 13:29:14 set a variable based on thisdir in the recipe that the class assumes has been set? Apr 03 13:30:00 rburton: yeah, good idea Apr 03 13:34:19 Has any of you guys here any corporate system/tools for product versioning that you need to integrate with yocto? E.g. how do you version the top-level product artefacts? Manual rename of the dated output images, automatic scripts, or just use them as they are? Apr 03 13:35:40 Because I need to build a scheme for handling output artefacts, and since we're building three different MACHINES, I can't do this from within a yocto recipe -- I need to make a post-build system on top of bitbake Apr 03 13:36:33 I'm not trying to leech anything, just curious to how other might do this Apr 03 13:37:33 sveinse: automatic build with jenkins, including automatic renaming and archiving the artefacts it is here. Apr 03 13:38:45 LetoThe2nd: Do you embed a product version into the image, or is a image just a collection/manifest of individual packages with their versions? Apr 03 13:39:46 sveinse: we also embed complete manifests into the images. Apr 03 15:04:23 Does yocto have some tool or setting to erase all sstate cache versions but the latest? Like a if rebuilding a recipe, then wipe the old ones first? Apr 03 15:05:06 sstate-cache-management or something in scripts Apr 03 15:05:18 or just remove anything that hasn't been touched in a week with a find call Apr 03 15:05:39 wiping old caches immediately would negate the point of being a cache Apr 03 15:08:15 yes, I can see that. Yet this recipe is the most volatile package (will be rebuilt in 98% of the builds, so the cache will be dominated by its artefacts). But I'll find a solution, thanks. Apr 03 15:18:57 anyone knows how to fix "Warning when extracting archive entry: Can't set UID=0" - > opkg-0.3.0 and libarchive >= 3.2.1 Apr 03 15:28:13 Have I understood this right: if I have a compile time setting 'EXTRA_QMAKEVARS_PRE = "-r VERSION=${ANY}"', I can add EXTRA_QMAKEVARS_PRE[vardepsexclude] = "ANY" to prevent that any updates to ANY will trigger a rebuild, right? Apr 03 15:28:54 sveinse: correct, yes. Apr 03 15:29:09 kergoth: great, thanks Apr 03 15:30:48 And append ANY to BB_HASHBASE_WHITELIST to remove it from the taskhash? Apr 03 15:34:04 sveinse: vardepsexclude removes it from the checksums when pulled in via that variable already Apr 03 15:34:20 the whitelist is global, to remove it from *everywhere*, it's brute force, and usually unnecessary Apr 03 15:35:12 hashbase_whitelist is vardepsexclude for every variable, really :) Apr 03 15:35:15 afk Apr 03 15:37:45 I'm doing something like ANY:="${@get_myversion(d)}", but if it is pulled from the recipe when I use it with vardepsexclude, then I get what I need Apr 03 15:46:55 kergoth: any idea? Warning when extracting archive entry: Can't set UID=0 Apr 03 15:47:42 Not a clue, sorry. when exactly are you seeing this? Apr 03 15:49:24 when installing ipks into offline sysroot Apr 03 15:50:43 what exactly are you trying to accomplish here? Apr 03 15:51:08 obviously manually installing packages as a user will result in files with wrong permissions, since you aren't root, which is likely exactly what that warning is telling you Apr 03 15:51:22 s/installing packages/installing packages on the host rather than target/ Apr 03 15:55:17 kergoth: i try to install nativesdk ipks manually to sysroot using opkg-cl and -o option. I am wondering why libarchive (or opkg-0.3.0) (poky-2.0) is trying to install with UID 0, i never had this with libarchive from fido Apr 03 15:55:35 ddora: there are files owned by uid 0 in the tarball in the ipk Apr 03 15:55:45 presumably, anyway Apr 03 15:56:10 seems pretty straightforward to me. just ignore the warning and move on if it's not a concern Apr 03 15:56:11 control files? Apr 03 15:56:20 if you're curious, examine the ownership of the files in the ipk Apr 03 15:56:29 ar x foo.ipk; tar -tzf data.tar.gz; tar -tzf control.tar.gz Apr 03 15:56:46 thanks! Apr 03 15:57:07 np Apr 03 15:58:57 ddora: https://github.com/kergoth/dotfiles/blob/master/scripts/ipkcontents, https://github.com/kergoth/dotfiles/blob/master/scripts/ipkinfo, https://github.com/kergoth/dotfiles/blob/master/scripts/ipkunpack Apr 03 15:59:06 might be of interest, for convenience Apr 03 15:59:40 thank you very much for sharing! Apr 03 15:59:44 does bitbake have an option for printing what variable change that triggered rebuild? Apr 03 16:02:04 sveinse: bitbake -S printdiff will print "why sstate isn't used" basically. bitbake-whatchanged used to be of interest, but i don't think it's maintained anymore. bitbake-diffsigs -t is likely also of interest, you can bitbake foo; bitbake foo; bitbake-diffsigs -t foo build -> compare the previous two builds of foo Apr 03 16:02:20 ddora: np Apr 03 16:20:01 Must a recipe always result in packages? Can a recipe be a service, like executing a build step, e.g. upload to a staging server? Apr 03 16:21:48 sveinse: if you want to change what tasks run, you can. Apr 03 16:21:51 see deltask and addtask Apr 03 16:22:04 base.bbclass and package.bbclass & friends define the defaults, that's all Apr 03 16:24:21 Basically I have two choices when it comes to our post build deployment: Either write it completely outside yocto, or make a recipe for it and run it with -c deploy or something. Apr 03 16:24:31 yep Apr 03 16:24:50 I know it's slightly outside the scope for the recipes, but then again it is convenient to have all in the same environment Apr 03 16:29:43 heh, Cooker is so busy doing the bitbake -S printdiff for me that it does not respons to SIGINT or SIGTERM... Apr 03 16:35:28 sveinse: image recipes don't result in packages, so yes they can do anything. another example is the package-index recipe. Apr 03 16:38:14 rburton: perfect, thanks Apr 03 16:39:35 I notice it uses do_fetch[noexec] = "1" instead of deltask do_fetch Apr 03 16:44:35 it's easier to undo a noexec if need be than to undo a deltask, as deltask removes task dependencies in both directions, to re-add would require re-adding the deps too Apr 03 16:44:39 case by case thing Apr 03 16:49:49 right Apr 03 16:51:01 btw, I've already fallen in my own trap of requiring a path variable to be set before calling a function in a bbclass :( I wish there were someway I could get the retrieve for the current bbclass file or directory. Apr 03 16:53:13 ...some way I could retrieve the current bbclass file or directory... (I can't write today) Apr 03 16:55:35 sveinse: hard to do directly, normally FILE is the currently parsed file, but it doesn't apply to bbclasses. if you don't mind hardcoding the class relative path, you could easily use inline python to find the absolute path by searching BBPATH for it. Apr 03 16:56:40 kergoth: how do I know the name for the bbclass file? Hardcoding? Apr 03 16:57:33 it may be possible to dig into bitbake internal variables to get it, as is used in bitbake to implement EXPORT_FUNCTIONS, but that wouldn't be ideal either Apr 03 16:59:45 right. an odd limitation, iirc, as moving functions to bbclass is encourages, but the system don't scale along with it. I can get the path to the caller's bb-file, but I don't know my own path relationship to it. Apr 03 16:59:54 s/iirc/imho/ Apr 03 17:04:12 sveinse: iirc it was slightly more complex conceptually just because bbclass parsing involves a cache, to avoid reparsing a class that has multipl e'inherit' lines around, but it just means we should add a new variable for the current bbclass. i'd suggest opening a bug to add that feature to bitbake. it'd be trivial to add Apr 03 17:06:53 kergoth: ok, noted. thanks Apr 03 17:24:10 I get a taskhash mismatch which I'm trying to debug. When I call bitbake-diffsigs -t recipe task I get nothing Apr 03 17:24:44 Ah, perhaps due to the the taskhash mismatch, which is an error, then the data isn't updated fully Apr 03 17:30:29 what is a good apprach for debugging taskhash mismatches? Apr 03 17:39:10 I have this generic recipe whose purpose is to run do_increase_version every time. I added do_increase_version[nostamp] = "1", but *that* causes the recipe to die in task hash mismatch every time. Any clues to why? Apr 03 17:39:18 Here's the full recipe: https://bpaste.net/show/68b2ecc1f649 Apr 03 17:49:13 hmm, nostamp seems to be toxic Apr 03 17:52:38 anything depending on a nostamp task will have its checksums change every time, it can't do otherwise, afaik Apr 03 17:53:09 are you sure you want to do that in a task rather than at parse time? Apr 03 17:54:57 kergoth: yes I think so, because this task must be run explicitly. We build run bitbake over three MACHINE= iterations, and the version must not change between them Apr 03 17:55:37 likely don't want anything depending on that task, then Apr 03 17:55:45 Likewise, the release deployment tasks after build must also be run separately since krogoth can't bundle these three MACHINE builds into one invocation Apr 03 17:55:55 Absolutely nothing shall depend on them, no Apr 03 17:58:20 Think I found it with your hint, kergoth: I have a recipe and a class with the same basename. Apparently that created some dependency linking. Renaming the recipe seems to fix the issue Apr 03 17:58:38 huh, that's interesting Apr 03 17:58:52 if you can repro that reliably, report a bug, that should not be the case Apr 03 17:59:00 ideally produce a minimal test case, if possible Apr 03 17:59:14 obviously you have higher priority stuff to address now, but later maybe, if you would :) Apr 03 17:59:54 no sorry, now it's gone Apr 03 18:00:30 I can't get it to fail anymore Apr 03 18:01:54 and for the record I will... if I get it to fail consistently again Apr 03 18:02:20 np :) Apr 03 18:22:21 When is the DEPENDS items fulfilled in a recipe? E.g. if I add a task that has no dependencies and no connection to do_build, does bb always satisfy DEPENDS first? Apr 03 18:25:43 sveinse: see deptask flag in base.bbclass Apr 03 18:27:40 hi Apr 03 18:29:16 tenho o simatic2000 da simemens. quero saber como instalar o postgresql. Alguem tem ideia ? Apr 03 18:29:27 I have simemens simatic2000. I want to know how to install postgresql. Anyone have an idea? Apr 03 18:29:31 :) **** BEGIN LOGGING AT Mon Apr 03 18:41:26 2017 Apr 03 18:44:03 I want switch stable poky to master, so I checkout the master branch and meet this https://paste.fedoraproject.org/paste/I6qtPd6~-8B~lVBsn0LGQl5M1UNdIGYhyRLivL9gydE= Apr 03 18:44:19 what should I do to clean up the cache without clean up the download cache? Apr 03 18:48:08 ayaka: you just need to delete whatever you set TMPDIR to, usually tmp/ under the build directory. Apr 03 18:48:16 ayaka: (also it says your networking is broken) Apr 03 18:48:30 maybe that error should tell you *what* tmpdir is… Apr 03 18:48:47 rburton, I see thank you Apr 03 18:49:10 btw, how to select the C libraries in yocto system? Apr 03 18:49:25 TCLIBC Apr 03 18:49:28 defaults to glibc Apr 03 18:49:32 you can also set it to musl Apr 03 18:50:27 rburton, those thing doesn't cover in the documents Apr 03 18:51:03 rburton, sorry it is Apr 03 19:07:26 rburton, I remove it but it still not work for me Apr 03 19:07:38 I didn't set any TMPDIR at my local.conf Apr 03 19:08:31 ayaka: start with a new build folder Apr 03 19:09:04 lsandov, doesn't use oe-init-build-env? Apr 03 19:09:28 ayaka: source oe-init-build-env build-krohogth, for example Apr 03 19:09:53 where build-krogoth is the new build folder, basically empty except for conf Apr 03 19:14:03 lsandov, no it doesn't work either Apr 03 19:14:39 ayaka: same error? Apr 03 19:14:46 lsandov, yes Apr 03 19:15:45 the networking bit? that's unrelated to tmpdir, and means your proxies are not configured Apr 03 19:16:12 * fray notes 19 bitbake builds on the same machine doesn't work will if the inotify is limited to 8k files.. Apr 03 19:16:25 but I did get the load on the machine over 2400 before it stopped.. ;) Apr 03 19:21:30 lsandov, I add that variable and run bitbake Apr 03 19:21:37 then remove it, it seems work now Apr 03 19:21:57 ayaka: TCLIBC? Apr 03 19:22:22 lsandov, BB_NO_NETWORK = "1" Apr 03 19:23:01 ayaka: ok, so apparently you have all in your DL_DIR Apr 03 19:23:46 fray: using an AB? Apr 03 19:23:54 no, just on the machine I've got Apr 03 19:23:58 lsandov, not actually I am downloading those new source code since I upgrade to master from morty Apr 03 19:24:22 running 9 builds at once (more reasonable number), my load is consistently around 200-400 range.. Apr 03 19:24:48 so machine is a bit overloaded, but not too bad.. but it's not using swap yet -- so that is keeping the performance up Apr 03 19:25:17 (I'm running some test builds against the YP 2.2.1 release.. nothing too big) Apr 03 19:25:24 fray: look at these plots for a simple build (core-image-minimal) https://wiki.yoctoproject.org/wiki/Vmstat Apr 03 19:25:48 fray: what you have is that x 9, too much Apr 03 19:26:44 fray: are you doing some kind of performance test? Apr 03 19:27:09 I've got 144 threads and 256 GB of ram.. (and a whole lotta raid) Apr 03 19:27:17 so a load of 200 isn't too bad.. Apr 03 19:27:31 400 is a bit much, but if most of that is I/O bound then it's actually not all that bad either Apr 03 19:27:34 fray: that is something :) Apr 03 19:28:05 model name : Intel(R) Xeon(R) CPU E7-8890 v3 @ 2.50GHz Apr 03 19:28:07 * 4 Apr 03 19:28:19 fray: you are basically creating fork bombs Apr 03 19:28:35 load of 2410 was a fork bomb.. this is just a 'busy system' Apr 03 19:29:21 (interactive wise, the system is responsive for such a high load -- currently around 220) Apr 03 19:29:43 I'd say each build is probably running at about 75-80% based on just watching the screen.. Apr 03 19:29:50 so that isn't bad performance.. Apr 03 19:30:00 I just have to make sure these 9 sets of builds complete Apr 03 19:32:00 load right now is in the 130 range... says I'm currently about 23% idle as well.... Apr 03 19:32:17 (not sure I trust Linux process accounting) Apr 03 19:36:22 fake accounting Apr 03 19:36:37 'Make the server great again...' Apr 03 19:37:12 "Make MN have a brown-out again." Apr 03 19:37:37 * fray should check the temps on that machine.... Apr 03 19:38:27 room.. 'incoming air' to the sevrer is 86F.. air at my shoulder height is 91F.. Apr 03 19:39:14 CPU1-4 temps, 62C, 60C, 57C and 59C Apr 03 19:39:29 with the electric bill and the thermal signature of the building, they'll be busting your door down thinking you are a grow-op. Apr 03 19:39:44 Ohh I should check power usage.. ;) Apr 03 19:39:58 power consumption.... Apr 03 19:40:11 machine 'all-time high' is right now.. 1220 Watts.. Apr 03 19:40:36 average for the last hour is 1000 watts.. average for past 7 days is 562 watts Apr 03 19:41:55 AC input power is just under 300 W (* 4 power supplied).. output power is around 260W (* 4 power supplies) Apr 03 19:42:04 that is more lossy then I would have expected Apr 03 19:42:45 13% power loss seems 'excessive' for supplies that are '90 Platinum rated' Apr 03 19:45:50 * fray wonders how loud the office is right now.. (probably unbearable) Apr 03 19:47:13 fray: i just got a power monitor thingy on my domestic mains and now wonder what sort of idle tuning xeons have as my build machine appears to idle at about 150W Apr 03 19:47:38 if i just run laptop-mode will that do the right thing i wonder Apr 03 19:47:55 ya, I wish this only idled at 150W.. when I'm not using it, I often power it down.. but even then (with the BMC only) it draws about 120 watts) Apr 03 19:48:08 ouch Apr 03 19:48:26 thats a hell of a lot for a bmc Apr 03 19:48:26 luckily I have a remote power switch on the box, so I can actually power it off Apr 03 19:48:57 ya.. I don't think the BMC really takes 120 watts.. but I suspect between the 4 power supplies, all of the fans and the BMC controller that accounts for it.. Apr 03 19:49:10 I almost wonder if I yanked 2 of the 4 supplies if the power efficiency would go up Apr 03 19:49:34 the machine is capable of running off one supply, but there is a warning that 1 supply may not have enough power if the chassis is loaded full of drives and ram Apr 03 19:49:43 (I'm nowhere near loaded on either...) Apr 03 19:50:54 Intel defines the TDP for each CPU at 165W... Apr 03 19:51:18 so that is 660 of the 1200 right there.. then add the cooling, drives, ram, raid card, etc.. 1200 isn't 'unreasonable' Apr 03 19:52:23 * fray should be able to shut down two of the supplies (remotely) and check.... Apr 03 19:55:27 ya, even with 2 of the supplied down -- it's showing input (AC) vs output (DC) power is still a 13% loss.. Apr 03 23:30:07 bluelightning: nice Apr 03 23:30:22 bluelightning: one issue, i think you accidentally depend on a newer python than bitbake acxtually requires right now: Apr 03 23:30:26 File "/data/kergoth/mel/elm/test/poky/bitbake/lib/bb/siggen.py", line 426 Apr 03 23:30:26 return formatstr.format(**colors, **values) Apr 03 23:30:26 ^ Apr 03 23:30:26 SyntaxError: invalid syntax Apr 03 23:42:28 kergoth: argh, really? Apr 03 23:42:40 damn Apr 03 23:43:52 well, fairly easy to work around, at least with the way I structured it I only need to do it in one place... Apr 03 23:49:04 kergoth: fix pushed Apr 03 23:49:08 thanks **** ENDING LOGGING AT Tue Apr 04 03:00:01 2017