**** BEGIN LOGGING AT Thu Mar 29 03:00:03 2018 Mar 29 07:56:05 hrw: you made lwn: https://lwn.net/SubscriberLink/750238/26055a95b7808283/ Mar 29 08:16:56 koen: yes, I did. Mar 29 08:39:12 Hi. How come that if I do "bitbake -c compile virtual/kernel", and then in a devshell for virtual/kernel do "cd $KBUILD_OUTPUT && make bzImage", all C files are recompiled ? Mar 29 09:01:31 Hi. lately our CI machines started failing in random places during GCC compilation with "broken pipe". the CI machines have lots of CPU's (around 40), basically that's the whole difference between our normal machines (8 cores). we don't get the error on our private machines. Mar 29 09:01:45 we're building Rocko images. Mar 29 09:01:55 did anyone experience that before ? Mar 29 11:59:43 hello Mar 29 12:05:55 what is the common way for the rootfs on an embedded device that can be powered off at any time? Mar 29 12:06:10 we use an slc flash with ext4 Mar 29 12:15:28 sn00bie: common was, as it? Mar 29 12:15:33 *as in Mar 29 13:50:25 LetoThe2nd, what did you mean sry i didnt understand xou Mar 29 13:50:26 you Mar 29 13:52:39 sn00bie: you asked for "the common way". like, what do you expect? Mar 29 13:54:50 whe have at the prototyping and development phase an corrupted file system with ext4 and rw mounting Mar 29 13:55:25 well the obvious thing is a readonly FS :-) Mar 29 13:55:48 mhhhh Mar 29 13:56:22 and that is the well known practice, because the filesystem cant get corrupted, only the user data/settings right? Mar 29 13:57:19 a read only rootfs is usually the very first step, because it is rather easy to implement. Mar 29 13:58:18 i tried that, and yes it is easy, but we ve some user settings, so only 1 partition isnt enough... Mar 29 13:58:27 other bits and pieces are dynamically remounting things only when actually needed, caching in battery backed memory, early powerfail detection in hardware... Mar 29 13:58:43 LetoThe2nd, sry batterie is no option :-( Mar 29 13:58:51 yes, obviously you need more than one partition then. Mar 29 13:59:56 i change the uboot script, and yes mounting rootfs as ro is very easy, but after that we ve some more work to get the rest of the system work... Mar 29 13:59:59 user data etc... Mar 29 14:00:42 sounds like at the moment you're finding out why a *real* product is something completely different from a prototype or proof of concept. :) Mar 29 14:01:47 there is an IMAGE_FEATURE read-only-rootfs, by the way that does most of the needed tricks Mar 29 14:02:09 where can i get this feature? Mar 29 14:02:31 its an IMAGE_FEATURE, like any other IMAGE_FEATURE in OE Mar 29 14:02:57 okay, thanks for that info Mar 29 14:03:03 i check that Mar 29 14:03:04 :-) Mar 29 14:05:35 usually a proper powerfail strategy consists of several pieces that are partially hardware, partially application specific. i can share a few pointers, but most companies tend to keep that as core know how and not give out for free Mar 29 14:07:43 oh, okay. if you can give me some hints, that would be great Mar 29 14:10:05 1) keep rw to a minimum, in both time and "space" 2) have ways to detect corruptions in the rw-able parts 3) have sane defaults 4) make hw capable of detecting powerfails.... thats a few rather common ones .) Mar 29 14:45:38 LetoThe2nd, thx a lot! Mar 29 14:52:15 sn00bie: have fun Mar 29 15:31:03 At this point glibc is *the* Linux API - such statements are just so anti open source but I am not surprised Mar 29 15:32:14 systemd's philosophy is to standardize base OS and I am understand the desire Mar 29 15:32:31 s/am/can Mar 29 15:35:37 Well... eventually systemd will contain the C library given the current trajectory ;) Mar 29 15:42:41 I am sure if a small system like procd can leave its openwrt affinity aside, many smaller distros will switch to it there is s6 out there Mar 29 15:43:02 there is a reason why docker switched to musl Mar 29 15:43:12 for default template Mar 29 15:44:04 I find linux kernel folks more reasonable where they accept that linux cant become a rtos for light bulb Mar 29 15:45:18 at some point you have to draw the line of sanity Mar 29 19:52:40 rly quiet here **** ENDING LOGGING AT Fri Mar 30 03:00:01 2018