**** BEGIN LOGGING AT Mon Sep 25 03:00:00 2017 Sep 25 07:54:47 JaMa, hi. Yesterday I got error/too few inodes on /tmpfs building core-image-base. Something new... Sep 25 07:56:54 same setup as usual, rm_work(). Previous build from scratch (flawlwss) was around mid september iirc. Sep 25 12:18:34 Hello all, I am looking at recipe available for curl with OE, http://layers.openembedded.org/layerindex/recipe/5765/ and in the recipe flags it explicitly disable the nghttp2(http2) support. I have a recipe which I can use to install nghttp from source available here, https://github.com/nghttp2/nghttp2/releases but still not able to get curl to compile with it. Any pointers on how can I get this sorted out? Sep 25 12:19:53 To be more clear, instead of using official recipe for curl I did create a custom one against source available here: https://curl.haxx.se/download.html Sep 25 12:20:06 using devtool command Sep 25 12:32:21 baali: read the docs on packageconfig so you can turn that --without-nghttp2 into a PACKAGECONFIG[nghttp2] that you can then turn on Sep 25 12:33:09 rburton: got it, thanks for the pointer. Sep 25 13:01:55 rburton, hey. My usual build of core-image-base failed because of inode's shortage. Strange, I build in tmpfs and didn't change anything, Sep 25 13:06:29 ant_work: didn't you run out of inodes in tmpfs? Sep 25 13:06:49 ant_work: I have to mount it with significantly bigger amount of inodes for big builds Sep 25 13:06:56 low, they were around 90.000 Sep 25 13:07:11 I did build the usual core-image-base Sep 25 13:07:41 something must have changed since beg. september, when I did a previous build from scratch Sep 25 13:08:33 none 28783068 1166696 27616372 5% /home/jenkins/oe/world/shr-core/tmp-glibc Sep 25 13:08:36 maybe rm_work has been postponed/skipped? My bad I didn't check it Sep 25 13:08:46 Filesystem Inodes IUsed IFree IUse% Mounted on Sep 25 13:09:57 JaMa, right, for big stuff Sep 25 13:11:06 I'll retry later Sep 25 13:11:54 fwiw on the same setup I do remember failure (low space) doing bitbake -c do_populate_sdk core-image-base Sep 25 13:12:35 first I though about races, then I have seen the space was getting low. Sep 25 13:12:40 Time to investigate as well Sep 25 13:14:11 ant_work: never seen that, but if you use wic images then there's another hardlink copy made briefly which could explain Sep 25 13:15:08 yes...what puzzles me now is the normal build of the small console-base-image Sep 25 13:16:02 i build in a tmpfs all the time and never run out of inodes (space, yes) Sep 25 13:16:16 right, is a first for me Sep 25 13:17:59 and btw kernel must have been already built because after wiping /tmp the build took just a few seconds... Sep 25 13:18:20 so I suspect rm_work might misbehave **** BEGIN LOGGING AT Mon Sep 25 16:44:51 2017 Sep 25 16:48:51 thanks rburton, with few more edits, that did the trick. I think I can document this nghttp2 related dependency for curl **** BEGIN LOGGING AT Mon Sep 25 19:21:18 2017 Sep 25 19:21:26 rburton: Done. Sep 25 19:22:37 baali: and queued for staging thanks Sep 25 19:24:09 :) Sep 25 19:29:02 for nghttp2 recipe, would meta-networking/recipes-protocols/ be the right place? Sep 25 19:34:07 yeah probably Sep 25 19:34:11 (not my layer) Sep 25 19:35:27 thanks rburton, I will dig more tomorrow and commit it, hopefully at correct location. Sep 25 21:20:37 Hey, is there anyone here that's made use of the way that OE names symlinks for dtbs based on the Linux kernel type(s), I'm just trying to figure out what the uses are for folks for something that is the same regardless of the kernel image type you're building **** ENDING LOGGING AT Tue Sep 26 03:00:02 2017