**** BEGIN LOGGING AT Tue Nov 27 03:00:01 2018 Nov 27 03:21:22 build #175 of malta/be is complete: Failure [failed sourceupload] Build details are at http://release-builds.openwrt.org/18.06/images/builders/malta%2Fbe/builds/175 blamelist: Freddy Leitner Nov 27 03:37:30 i just had a thought Nov 27 03:37:46 since cake is part of 4.19, will that mean it will become the default here? Nov 27 05:34:29 ping INAGAKI? Nov 27 06:53:48 hi, will linksys EA8100AH be added to the list of build targets? Nov 27 07:05:12 a look into the ToH would tell you about it Nov 27 07:06:46 that said, its mt7615n wlan cards aren't supported yet and probably won't be supported 'soon' either Nov 27 07:07:48 ok thx. thats what i needed to know Nov 27 10:27:31 Good morning Nov 27 10:27:54 https://twitter.com/vanhoefm/status/1067363695695671296/photo/1 Nov 27 10:29:12 Good morning. Sorry for repetition I'm on a new phone irc client Nov 27 10:29:56 I'm trying to do something unusual hopefully someone can help me or at least let me know if it is feasible Nov 27 10:31:35 I already have a toolchain for an esoteric platform I'm developing on. I'd like Openwrt to build a set of userspace packages for me. So I can install them using opkg. Nov 27 10:33:03 Can I use this toolchain? I tried to research openwrt external toolchains but it appears to use a toolchain built by openwrt Nov 27 10:33:16 stintel: nice and scary Nov 27 10:39:04 hm, I'm wondering how are the status LEDs implemented, in DTS I can see, that there are aliases for the LEDs like `aliases { led-boot = &power; }` but I couldn't find anything else then DTSes with `git grep led-boot` Nov 27 10:39:06 AndroUser: it's in the menuconfig if you want to use an external toolchain, Nov 27 10:40:38 Hi Karlp, have you tried doing this before? It seems to me that you need to use a toolchain built by openwrt. I need to use an arbitrary toolchain that i was given. Nov 27 10:43:40 AndroUser: I did this before and it works to some extent Nov 27 10:43:49 AndroUser: depends on how standard the toolchain is Nov 27 10:43:54 AndroUser: I've done it, it worked enough for my case. Nov 27 10:44:07 I had an already built buildroot, Nov 27 10:44:14 then switched the toolchain and rebuilt just some packages. Nov 27 10:44:26 you need to properly configure it, just pointing to the corss prefix is not enough, you'll also need to setup the file globbing patterns for the libc required libraries etc. Nov 27 10:44:29 I wass't trying to recreate the entire root filesystem Nov 27 10:44:59 the defaults should work weel enough for normal toolchains though Nov 27 10:45:01 *well Nov 27 10:45:22 Jow are there instructions on how to do this. The wiki suggested just modifying three fields. Nov 27 10:45:49 Karlp that is exactly what I'm trying to do Nov 27 10:45:50 should be all in menuconfig Nov 27 10:45:55 Only need a few packages Nov 27 10:46:12 unless you tell us what probelms you encountered exactly I am not sure how to help further Nov 27 10:46:25 Ok jow thanks I'll look into once I get to my computer. Thanks Nov 27 10:46:47 I just wanted to make sure it's feasible before starting. Nov 27 10:52:40 I once tested with a proprietary vendor toolchain and a linaro multilib toolchain Nov 27 10:52:45 both worked at that point in time Nov 27 10:53:03 Awesome Nov 27 10:53:25 yes, it will usually build, but it doesn't matter, that it will run (or run reliably) Nov 27 10:53:36 s/matter/mean/ Nov 27 10:53:42 ynezz: ? Nov 27 10:54:01 as you've said, it depends on the toolchain Nov 27 10:54:15 jow: What should the target be? I don't intend on building kernel so should the target even matter? Nov 27 10:54:16 and the packages you're trying to build Nov 27 10:54:44 AndroUser: I'd pick on of the virtual targets that are closest to the arch targeted by the toolchain Nov 27 10:55:03 AndroUser: e.g. if you have an arm toolchain, use the armvirt target if you have a mips toolchain, use the malta target Nov 27 10:55:41 Ok awesome thanks Nov 27 10:55:47 these virt targets have the elast amount of device specific image building code etc. Nov 27 10:55:52 *least Nov 27 10:56:02 jow ping Nov 27 10:56:07 Tapper: pong Nov 27 10:56:09 Ok cool Nov 27 10:56:17 Hi Nov 27 10:56:26 hi Nov 27 10:56:57 I just wanted to ask you you about the dialog that pops up when installing a package when using luci-app-opkg Nov 27 10:57:21 It does not seem to read to screen readers in firefox and crome Nov 27 10:57:44 It does work in ie tho. :-D Nov 27 10:58:59 I added aria-modal etc. stuff that should fix it, maybe try the latest version of luci-base and make sure that all browser caches are cleared Nov 27 10:59:07 All the rest works grate! It's just that one dialog that shos up at the bottom of the page. Nov 27 11:00:04 OK I will chek for any luci updates, but I am on a pretty new build Nov 27 11:00:33 the dialog is a div overlay Nov 27 11:00:52 with role="dialog" and aria-modal="true" Nov 27 11:01:06 this supposedly should tell screen readers to treat it like a dialog Nov 27 11:01:16 the label is the first heading elememnt withi nit Nov 27 11:01:21 *within it Nov 27 11:01:34 which in this case is a h4 element Nov 27 11:02:01 what reader software do you use? Nov 27 11:16:23 Sorry jow I have updated all my luci packages and used Ccleaner the dialog is still not working for me. Nov 27 11:16:52 I use NVDA Nov 27 11:18:04 The first time I tryed luci-app-opkg it just shoed up as text after the mane boddy of the page Nov 27 11:26:41 I don't know how to fix it then Nov 27 11:26:50 is this working for you? https://www.w3.org/TR/wai-aria-practices/examples/dialog-modal/dialog.html Nov 27 11:28:03 I will look now Nov 27 11:28:22 how do I put a negation in a package Makefile to _not_ build this package if 'PACKAGE_openvswitch [=y]' Nov 27 11:29:13 just !openvswitch ? Nov 27 11:29:53 Yes the Add Delivery Address test works for me. Nov 27 11:30:57 strange Nov 27 11:31:08 Strykar: DEPENDS:=!PACKAGE_openvswitch Nov 27 11:40:17 ty jow Nov 27 11:41:57 Good day (.*), I'm trying to port a pkg to OpenWrt (ptunnel-ng). When I run `make package/ptunnel-ng/compile`, it complains about missing deps for libc.so.6, libpcap.so.0.8, libpthread.so.0, libselinux.so.1 ... but when I run ldd on the created binary, it prints /lib/ld-musl-armhf.so.1, libgcc_s.so and libc.so. Where does the generated dependency graph gets its information about required library deps? Nov 27 11:42:38 lnslbrty: it finds ELF files in the build_dir which link host libraries Nov 27 11:43:10 this commonly happens when either the build tree is unclean (already ships with .o or executable files) or if it wasn't actualyl cross compiled (wrong gcc used during build) Nov 27 11:43:23 jow: ok, will check, thx Nov 27 11:43:56 libc.so.6 usually means that something ended up in PKG_BUILD_DIR (or PKG_INSTALL_DIR) that somehow links the host glibc Nov 27 11:44:20 could be something simple like an xxx-config executable which isn't actually meant to be packaged Nov 27 11:44:36 or some artifact created during the cross compilation phase Nov 27 11:45:02 jow: so I have to check the PKG_(BUILD|INSTALL)_DIR ? Nov 27 11:46:32 I'd start with a `find build_dir/target-*/pkgname-*/ipkg-install/` Nov 27 11:46:37 see what is in there Nov 27 11:47:08 identify each *.so or ELF executable and verify its architecture with e.g. the "file" command Nov 27 11:48:43 file ipkg-arm_cortex-a9_vfpv3/ptunnel-ng/usr/sbin/ptunnel-ng ipkg-install/usr/bin/ptunnel-ng .pkgdir/ptunnel-ng/usr/sbin/ptunnel-ng Nov 27 11:48:43 ipkg-arm_cortex-a9_vfpv3/ptunnel-ng/usr/sbin/ptunnel-ng: ELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, for GNU/Linux 2.6.32, BuildID[sha1]=1b6acdb7418e8478c08e32045b063fa74b46e1d1, stripped Nov 27 11:48:43 ipkg-install/usr/bin/ptunnel-ng: ELF 32-bit LSB shared object, ARM, EABI5 version 1 (SYSV), dynamically linked, interpreter /lib/ld-musl-armhf.so.1, not stripped Nov 27 11:48:43 .pkgdir/ptunnel-ng/usr/sbin/ptunnel-ng: ELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, for GNU/Linux 2.6.32, BuildID[sha1]=1b6acdb7418e8478c08e32045b063fa74b46e1d1, stripped Nov 27 11:48:46 wtf? Nov 27 11:48:54 How could that happen? Nov 27 11:50:16 It does not xcompile. :( Nov 27 11:50:39 can you share your OpenWrt Makefile used so far? Nov 27 11:50:44 y Nov 27 11:50:48 1mom Nov 27 11:51:40 *yawn* Nov 27 11:51:51 i was awake and working till 4am Nov 27 11:52:00 then got woken up by a hungry kid at 6:15 Nov 27 11:52:12 jow: https://pastebin.com/012SZxvk Nov 27 11:52:16 and now at 1pm i feel that no coffee could be strong enough Nov 27 11:52:33 blogic: amphetamins <3 Nov 27 11:53:48 did you use the shipped autogen.sh ? Nov 27 11:54:10 can you share the complete log of "make pacakge/ptunnel-ng/{clean,compile} V=s" as well?= Nov 27 11:56:10 lnslbrty: can't spot anything obviously wrong, going to try it locally... Nov 27 11:56:11 blogic I feel your pane. Kids are grate at that stuff. Nov 27 11:56:40 jow: https://pastebin.com/mRkdM3dY Nov 27 11:57:08 jow:sorry, it's mixed with a german locale ;) Nov 27 11:58:05 lnslbrty: found the issue Nov 27 11:58:17 lnslbrty: $(INSTALL_BIN) $(PKT_INSTALL_DIR)/usr/bin/$(PKG_NAME) $(1)/usr/sbin/$(PKG_NAME) Nov 27 11:58:30 lnslbrty: you have a typo there (PKT_INSTALL_DIR vs. PKG_INSTALL_DIR) Nov 27 11:58:45 so you copied your host /usr/bin/ptunnel-ng into the pacakge Nov 27 11:59:18 noticed it here since I don't have any ptunnel-ng on my system so it errored out with "install: cannot stat '/usr/bin/ptunnel-ng': No such file or directory" Nov 27 12:00:09 jow: damn, I was unable see that typ0 Nov 27 12:00:15 jow: thx for your help Nov 27 12:01:04 jow latest thing I can find on custom modal dialogs Nov 27 12:01:09 https://developer.paciellogroup.com/blog/2018/06/the-current-state-of-modal-dialog-accessibility/ Nov 27 12:01:17 jow: work now, ty very much <3 Nov 27 12:01:22 Sorry I cant be more help Nov 27 12:01:57 Tapper: the w3c example toggles display:none with display:block while LuCI uses visibility:hidden vs. visibility:visible + offscreen rendering Nov 27 12:02:05 Tapper: will try nvda later this week Nov 27 12:02:28 Thank you jow Nov 27 13:03:22 jow, installing feeds now: 'WARNING: No feed for package '!PACKAGE_openvswitch' found' Nov 27 13:04:47 ugh, nm Nov 27 13:07:11 Hi jow, I was AndroUser earlier on. I'm playing around with getting external toolchain working Nov 27 13:07:44 why do we need include/site/? Can't we just run autotools and figure that information out for each package? Nov 27 13:09:54 because you're cross compiling? Nov 27 13:11:10 iircn ot all autotools tests can be performed at compile time Nov 27 13:11:41 some need runtime testing (which cannot be done during cross compilation), hence the site script to pre-seed certain common values Nov 27 13:11:56 apart from that you do not really want to test sizeof(int) for every pacakge Nov 27 13:12:20 however to bring up your toolchain you can try to simply "touch" an empty site file Nov 27 13:12:31 then see how far you get Nov 27 13:14:24 amazin' thanks again jow Nov 27 13:14:31 karlp you can cross compile autotools Nov 27 13:14:57 autotools often compiles tiny test programs and runs them to figure out values Nov 27 13:15:02 ^^ Nov 27 13:15:03 this cannot be done during cross compilation Nov 27 13:15:17 you cannot compile e.g. a mips test executable and execute it on the x86 build system Nov 27 13:15:55 so you need to preconfigure certain values to skip the test program phase Nov 27 13:16:25 (as a nice side-effect this also speeds up ./configure since not every value has to be compile-exec-compare'd over and over) Nov 27 13:16:54 still too mcuh checking for sunOs headers though :) autohell is still precious. Nov 27 13:17:42 yeah or supporting shells straight out of the eighties Nov 27 13:18:00 better make sure we're not in csh. better check again, just in case they were tricking us. Nov 27 13:18:23 yeah, and then fall over anyway because someone added a clever custom bash-only macro Nov 27 13:18:34 no, your autowin package isn't new enough :) Nov 27 13:19:25 or too new Nov 27 13:19:47 sorry abotu that luci ordering ticket btw, thought that was going to be an easy fix, but there's ordering bits in a few to many places for me to follow what was the intent :| Nov 27 13:19:56 s/fix/improvement/ Nov 27 13:20:07 karlp: can't you just give an explicit order? Nov 27 13:20:34 this is where I have a bunch of separate controllers all adding to the same menu. Nov 27 13:20:35 jow/karlp its good to see you guys also find autotools to be a tad precious haha Nov 27 13:20:45 don't want to try and synchronize them Nov 27 13:20:52 ok Nov 27 13:20:56 but yeah, you can just close it if you like, order is what's definied to work, Nov 27 13:21:16 and realistically it will be just me with all the pages, even if in theory others can write them. Nov 27 13:22:25 I'll see if there's an easy fix Nov 27 13:22:52 its probably just a matter of adding a private "get_ordered_children()" to the dispatcher, then use it in the various places Nov 27 13:24:11 lnslbrty: I'm curious, what kind of remote do you need to use the ptunnel-ng ? Nov 27 13:24:26 lnslbrty: simply another host also running ptunnel-ng ? Nov 27 13:24:48 jow: yep, another remote running in forward mode Nov 27 13:25:15 jow: even just my patch makes it "work" for all the _functional_ cases I can see. Nov 27 13:25:20 jow: the server runs usually on a modem-router in mode and the client as forwarder Nov 27 13:25:26 I was just a bit lost what the other orderings were even for, given that mine was enough Nov 27 13:25:28 *proxy_mode3 Nov 27 13:25:57 karlp: big ball of mud Nov 27 13:26:03 jow: if you are using an external modem, it needs to support icmp forwarding Nov 27 13:26:27 jow: eh, just skip it then. Nov 27 13:26:42 karlp: nah, will take a look at it, it seems sensible Nov 27 13:26:50 jow: like the draytek vdsl/vdsl2 modems Nov 27 13:27:11 :+1: Nov 27 13:29:35 jow, I'm trying to build Lua. For some reason float-abi=hard is set on the gcc options. Lua doesn't use autotools so I assume it's not a site issue. Do you know where could be setting this option? I tried to grep target in case it was being set there. But nothing showed up Nov 27 13:33:27 oh wait I think I see it in rules.mk. I'll continue checking Nov 27 13:33:59 it is set when CONFIG_SOFT_FLOAT is disabled and the target arch is ARM Nov 27 13:34:31 yeah just saw it, I'm adding it manually now thanks Nov 27 13:37:05 ok great all working now thanks. Nov 27 13:37:09 amazing stuff :) Nov 27 13:39:12 karlp: that modisort thing seems to be completely dormant code Nov 27 13:39:44 karlp: it relates to a feature called "tree modifiers" which stems back to the original code from steven 11 years ago, I think it can be ripped out completely Nov 27 13:40:15 yeah, it _looked_ weird and out of place, just wass't sure Nov 27 13:40:28 I /guess/ the tree modifiers were supposed to provide an api to register custom code which somehow mangles the internal node tree Nov 27 13:40:45 but I find no users of it, I do not understand it and it smeels Nov 27 13:40:48 so I'll throw it out Nov 27 13:40:54 * karlp cheers Nov 27 13:56:25 karlp: can you please test if you also get your desired behaviour if you compare "a" and "b" (which are the respective nodes names as seen in the url) instead of the titles? Nov 27 13:56:39 hrm, good idea, didn't even try Nov 27 13:56:48 karlp: I think the node titles will hold the localized names which would make the order undeterministic Nov 27 13:57:05 non-deterministic Nov 27 13:57:16 well, deterministic within the locale, which is probably.... better? Nov 27 13:57:39 hm, sure why not Nov 27 13:57:56 but then I'll need to rework _find_eligible_node() too to follow locale order Nov 27 13:58:00 will check Nov 27 13:58:59 eyah, sorting by "a" "b" sorts on the url path name, which is globally consistent, Nov 27 13:59:15 sorting by node.nodes[x].title gives "title" ordering. Nov 27 13:59:26 I could go either way on that :) Nov 27 14:06:02 re alternatives https://imgur.com/a/giC5oQm Nov 27 14:13:56 I think I prefer to sort on url name Nov 27 14:15:45 easy to explain, easy to understand. works for me. Nov 27 14:32:02 jow: one less use of setfenv now too ;) Nov 27 19:00:40 * Monkeh pokes blogic Nov 27 19:03:26 *shrug* Nov 27 19:03:34 ok honey tell me .... Nov 27 19:03:59 gimme that link again and i'll shot some mails into the right direction Nov 27 19:05:07 https://github.com/openwrt/openwrt/pull/1500 Nov 27 19:06:03 blogic: That was the one Nov 27 19:06:20 pm me your email Nov 27 19:07:26 and RALINK_SYSCTL_BASE + 0x1340 is what we are after > Nov 27 19:07:27 ? Nov 27 19:08:16 That and apparently 0x1350 and 0x1360 - and anything else which may be related to undocumented pin remapping Nov 27 19:09:04 I have no idea if that sample code actually does what it suggests or not, but it certainly sounds like there are some useful registers somewhere Nov 27 19:15:49 Monkeh: snet, you are on Cc Nov 27 19:16:27 Cheers Nov 27 19:16:34 np Nov 27 19:39:25 would someone please point out in this Makefile why it won't download from the SF git repo: https://paste.ubuntu.com/p/sVyyr6krzs/ Nov 27 19:41:54 it just sits there after executing download.pl - https://paste.ubuntu.com/p/H3MW5zhTs7/ Nov 27 19:49:08 Strykar: include $(TOPDIR)/rules.mk Nov 27 19:49:15 missing that line at the top ? Nov 27 19:49:25 ah no Nov 27 19:50:11 blogic: Nov 27 19:50:26 Strykar: drop PKG_SOURCE:=$(PKG_NAME)-$(PKG_VERSION).tar.gz Nov 27 19:50:46 and line 41 has example1/ Nov 27 19:50:58 also drop the Prepare block Nov 27 19:51:19 line 25/26 are b0rked aswell Nov 27 20:24:26 hi, how can I find the xl2tpd configuration that openwrt generates? Nov 27 20:44:08 (more or less) https://github.com/openwrt/packages/blob/master/net/xl2tpd/files/l2tp.sh Nov 27 20:52:21 ugh, why are packages from sf hard :( Nov 27 20:53:02 ty blogic, that fixed the download, but now when I add '$(CP) ./src/* $(PKG_BUILD_DIR)/' to my Build/prepare section it fails with - https://paste.ubuntu.com/p/PPTxdNnGVg/ Nov 27 20:53:24 and if I omit that line, it cant find CMakeLists.txt Nov 27 20:59:28 ugh, sry, I misread remove the prepare statement Nov 27 21:12:53 one last Q, what's this error mean: 'error: comparison between pointer and zero character constant [-Werror=pointer-compare]' Nov 27 21:13:48 Build fail log - https://paste.ubuntu.com/p/cjbCdwh3d3/ Nov 27 21:36:34 stintel: so i sent you this email Nov 27 21:36:53 stintel: and i'd expect you to stad up straight and do what i asked Nov 27 21:37:31 stintel: so i hope for your sake, that you simply did not have time to check mails Nov 27 21:38:06 stintel: otherwise you'll have an ear full next time we meet Nov 27 21:38:15 russell--: ^^ same for you Nov 27 21:38:26 greearb_: ^^ and you Nov 27 21:40:29 no hard feelings bur for real guys ... Nov 27 21:40:35 *but Nov 27 21:47:05 not sure how anyone here can help your public shaming, without any context. Nov 27 21:47:27 karlp: :-) pm Nov 27 22:03:20 eh? Nov 27 22:03:35 eh ! Nov 27 22:04:48 what is the issue? I missed some email? Nov 27 22:07:37 -> pm Nov 27 22:16:35 any sunxi experts know why uboot won't boot the ramdisk image? https://zerobin.net/?3721efc2696caf53#Ccbs8XoZ4uDAy/wBFGBQPdQ20x3sPtZfWYRzXmcPfuo= Nov 27 22:51:41 ooh nice... the wayback machine has lots of pages which are missing from the current old forum archive. Nov 27 22:58:28 ok, got the kernel to boot at least, now to make it find the ramdisk... https://zerobin.net/?4bb01756d81f8378#kA+dtPHPUvz5t5nS5MJ/PwieLy5h8EC+9FUIWFoQCfg= Nov 27 22:59:02 is there some attribute we can add to an interface to make it non-deletable from within LuCI? Nov 27 23:19:11 karlp: if the uncompressed kernel gets bigger than 32MB it will overwrite the compressed one wihle uncompressing Nov 27 23:19:17 try a higher loadaddr Nov 27 23:24:49 Hauke: so just using an explicit higher address with the tftp of the ramdisk _should_ have worked? Nov 27 23:25:22 I'm currently in loading kernel and dtb explicitly from the build dir, and it boots, was just looking at where to get the ramdisk from to load to $ramdisk_addr_r Nov 27 23:28:01 Hauke: doyou mean just "tftp 0x42000000[+128meg or so] openwrt-sunxi-cortexa7-sun8i-h3-nanopi-neo-initramfs-kernel.bin ? Nov 27 23:28:54 I just tried loading it at 0x4a000000 and it does the same thing. Nov 27 23:29:29 or do you mean changing the "Load Address: 40008000" bit? that's set somewhere in our buildroot right? Nov 27 23:34:30 whee I actually got it :) Nov 27 23:35:57 swwwweeeet Nov 27 23:36:02 now we can start the real work :) Nov 27 23:40:51 Hauke: great to see that 4.19 too, I'll probably want to use that for sunxi Nov 28 00:02:55 (the trick was to tell uboot that the kernel was the initramfs kernel, not play with addresses: http://false.ekta.is/2018/11/booting-openwrt-on-nanopi-duo2-or-a-nanopi-neo-vi-fel/ Nov 28 00:07:15 is there a particularly interesting sunxi based device at the moment (something like a jack of all trades, enough oomph for compiling, enough I/O for decent mass storage performance, without sacrificing the more devboard like features (spi for flashrom, i2c, etc.) and the upcoming cedrus acceleration) - on a budget? Nov 28 00:07:58 * karlp shrugs Nov 28 00:08:29 I'm exploring a nanoepi duo2/core module for an upgrade from an ar9331 board for industrial stuff. Nov 28 00:08:57 haven't really planned doig any compiling on it, just want faster, and to make things like node run on it and more apps faster for plugins. Nov 28 00:09:51 I'm mildly curious in playing with ARMv7/ ARMv8 (aside from running it on my router, but that doesn't lend itself to playing around), but so far I haven't really found anything that would convince (as in compete with a ~100 EUR x86/ atom board) Nov 28 00:11:33 pkgadd, have you seen hardkernel boards? Nov 28 00:11:46 someone was comparing shitty intel with modern arm recently and said it was already better than shitty intel atoms and so on, perf wise Nov 28 00:11:58 I'm simply not even trying to do that sort of thing, so haven't even looked. Nov 28 00:12:41 Mister_X: yes and no, I haven't really found a lot of interesting sunxi devices beyond the A20. and I really want mainline support, so most of the hardkernel stuff isn't in the game Nov 28 00:13:07 here, this was it: https://www.cnx-software.com/2018/11/25/benchmark-x86-boards-vs-arm-sbc-rk3399-exynos/ Nov 28 00:13:17 I'm fine with linux-next or linux-next-next-next, but not with "vendor kernel for all eternity) Nov 28 00:13:39 pkgadd: what do you mean, I thought most of the hardkernel (amlogic at least) were well supported with upstream Nov 28 00:13:52 what's wrong iwth "past a20" for your use cases? Nov 28 00:13:59 I'm not familiar enough with arm to know which are mainline but from what I think, there aren't many Nov 28 00:14:10 (as in fully supported) Nov 28 00:14:37 karlp: mostly no SATA (fast I/O to main storage, I'm aware that the A20 isn't a performance monster for sata or ethernet either though) Nov 28 00:15:31 eyah, I've no use for any of those :) Nov 28 00:15:33 you probably know PCEngines? Nov 28 00:15:42 (for x86) Nov 28 00:15:51 100M ethernet is fine, and eMMC for core OS+apps, sdcard for teh rest is more than enough Nov 28 00:15:54 (for me) Nov 28 00:16:20 yes, but that's what I meant with needs to compete with a ~100 EUR x86 bottom of the barrel board, I'm interested in non-x86, but not at all costs ;) Nov 28 00:16:34 for your use cases,s ure. Nov 28 00:16:49 but x86 cna't even come _close_ to the size :) Nov 28 00:17:12 size doesn't matter, even less for my intended use cases (within reason, of course) Nov 28 00:18:02 42U 19" would be a bit much ;) Nov 28 00:18:11 not for you :) Nov 28 00:18:51 I'm ecstatic at the processing power I'm being offered in a comparable form factor to ar9331 modules. Nov 28 00:19:30 that said, there are three 3U 19" rackserver and a 2U 19" switch sitting on the desk next to me, a fourth wouldn't be a problem Nov 28 00:20:44 but that's 'boring' >11 year old x86_64 server gear Nov 28 00:45:51 pkgadd: the HardKernel Odroid-XU4/HC1/HC2 are supported upstream. Fedora 28 runs fine on them but the Fedora arm installer doesn't support them because of the requirement of a (signed) binary blobs needed for early stage boot support :-( Nov 28 00:46:42 I don't know if hardkernel's u-boot support has been submitted upstream, but kernel drivers have been. Nov 28 00:47:16 hmmm, signed bootloaders wouldn't quite be my preferred choice either Nov 28 00:52:09 nor mine, I haven't looked into that situation much, but I think it's a requirement of the Samsung Exynos 5422 that powers those boards. I'm hoping 2019 brings some more interesting mainline supported ARMv8 chips with A75 or A76 cores. These boards are A15/A7 (quad/quad) Nov 28 00:53:41 yep, unfortunately it's been "next year there'll be ARM servers" for the last 5 years already Nov 28 00:54:27 for only 1100USD you can obtain one of those Socionext 24-core A53 servers :-p Nov 28 00:56:19 well..... /pass sorry, far beyond my budget Nov 28 00:58:50 same here, I don't consider 24x A53 to be very interesting either. I keep hoping for 4-8x A72-A76 cores in the sub-$200 range eventually. Nov 28 01:00:10 yes, totally agreed on that. admitted for my use cases a large number of a53 cores would work as well (as at least compiling can be parallelized) **** ENDING LOGGING AT Wed Nov 28 03:00:01 2018