**** BEGIN LOGGING AT Thu Apr 25 03:00:02 2019 Apr 25 03:16:18 Grrrrrrrr the build bot hits some sort of race condition that prevents building. Apr 25 03:20:31 on 20 cores Apr 25 03:20:37 8 cores works just fine Apr 25 03:34:19 thagabe: make -j 20 V=s Apr 25 03:35:06 mangix Built the latest base + packages using $make -j9 V=s and built fine Apr 25 03:35:25 went ahead and $make clean with my actual .config file Apr 25 03:35:35 $make -j21 V=s Apr 25 03:35:39 error Apr 25 03:35:47 $make -j9 V=s and built fine Apr 25 03:36:32 issues with parallel building happen from time to time (imho a bit more 'recently'), but they're always hard to debug Apr 25 03:38:37 Yep, because running same exact command again and stuff just works.. Apr 25 03:38:43 yep Apr 25 03:39:00 and logging by itself also slightly changes the order of things Apr 25 03:41:22 rip parallelizing Apr 25 03:41:24 oh well Apr 25 03:41:36 thanks for the info, I thought i was going nuts Apr 25 03:41:50 well Apr 25 03:42:02 if it errors with make -j 21 V=s, you cacn check the error Apr 25 03:42:09 will take a while though Apr 25 03:42:30 could pipe it to a file Apr 25 08:28:30 What on earth are "Intel(R) 82576 Virtual Function" drivers/support? Why do/don't I want it? Apr 25 08:57:33 ahh, enabling direct access to the host NIC from virtual machines. I won't be requiring that then :-) Apr 25 09:25:23 hmm, my apu2 isn't remembering config changes across reboots - why would that be? Apr 25 09:25:35 ldir: sounds like tmpfs overlay Apr 25 09:26:03 CONFIG_PACKAGE_mkf2fs=y ? Apr 25 09:26:12 * ldir checks Apr 25 09:27:07 it is now! Apr 25 09:27:23 ;) Apr 25 09:27:45 have fun with your APU2! Apr 25 09:28:31 I've come from devices that have 'flash rom' ie archer c7,mir3g,hh5a etc etc so x86 is erm, strange Apr 25 09:29:10 will f2fs package just make it work(tm) ? Apr 25 09:30:03 it depends on how big is your image Apr 25 09:30:34 rootfs_data > 100M then it's f2fs, rootfs_data < 100M then it's ext4 Apr 25 09:32:36 https://lxr.openwrt.org/source/fstools/libfstools/rootdisk.c#L272 Apr 25 09:32:54 s/ext4/squashfs/ no? Apr 25 09:33:06 o_O Apr 25 09:33:17 ehr. lol Apr 25 09:33:32 that didn't make sense at all. I meant jffs2. or did we completely ditch that Apr 25 09:35:47 nope, jffs2 is still being used Apr 25 09:38:18 but jffs2 needs mtd layer Apr 25 09:38:53 ah, right. we ditched block2mtd Apr 25 09:38:57 yep Apr 25 09:39:48 that means we might be able to drop 440-block2mtd_init.patch and 441-block2mtd_probe.patch ? Apr 25 09:40:17 ah, seems mediatek still uses it Apr 25 09:40:46 ok nvm Apr 25 09:40:49 * stintel goes back to work :P Apr 25 10:01:01 right - sorry - phone went! Apr 25 10:01:21 am back now - so I have apu2 with 16GB SSD module thing in it. Apr 25 10:02:12 one of these - https://linitx.com/product/linitx-apu2-d4-4gb-(3nic+usb+rtc)-pfsense-msata-firewall-kit---silver/16019 Apr 25 10:02:26 I have 2 of those in use since early 2016 ;) Apr 25 10:02:42 not the linitx kit but ordered from PCEngines directly Apr 25 10:03:28 are you planning to do some benchmarks with SQM? Apr 25 10:03:44 no - I just want it to work Apr 25 10:04:55 ok :) Apr 25 10:06:05 hmm this doesn't look right https://pastebin.com/P65eCBGz Apr 25 10:06:21 two /boot Apr 25 10:06:41 I have it as well Apr 25 10:06:54 so do i Apr 25 10:06:56 not sure I ever reported it Apr 25 10:07:37 and both are mounted ro Apr 25 10:08:25 both rw here Apr 25 10:10:30 yay - it's remembered basic LED config across a reboot. That's progress. Apr 25 10:10:43 * ldir turned the power led into heartbeat Apr 25 10:12:07 what did you do with the led on the mir3g? Apr 25 10:15:18 I think I flashed one of the colours for heartbeat and left the other as a 'power'. mir3g coming out once I get my head around the APU Apr 25 10:22:39 *yawn* Apr 25 10:22:57 i just wasted 10 minutes wondering why udhcpc wlan0 aint working Apr 25 10:23:03 better add taht -i Apr 25 10:37:23 * ldir breaths sigh of relief, extra partitions do survive a sysupgrade Apr 25 10:38:16 \o/ Apr 25 10:42:03 Sooo, the remaining 14GB of my SSD....what should I format it as? ext4 ? Apr 25 10:49:39 f2fs Apr 25 10:51:13 * ldir reformats partition Apr 25 10:51:46 mine are ext4 Apr 25 10:57:55 ldir: i checked for some performance of ext4 vs. f2fs and there weren't too big differences Apr 25 10:58:05 ldir: so I went with ext4 as maybe better tested Apr 25 10:58:37 ldir: https://www.phoronix.com/scan.php?page=article&item=linux-50-filesystems&num=2 Apr 25 10:59:59 ldir: https://www.phoronix.com/scan.php?page=article&item=linux-416-fs&num=2 Apr 25 11:00:11 not sure I care about anything much other than perhaps wear levelling...but even then. I just didn't want 90% of my new disc sat unusable Apr 25 11:03:42 ldir: f2fs doesn't handle wearing directly, it's done by SSD's firmware Apr 25 11:04:02 ldir: so i'm not sure if you find any tests on wearing level/speed for different filesystems Apr 25 11:04:34 ldir: f2fs is probalby expected to be more friendly for SSD firmware handling NAND wearing Apr 25 11:04:48 ldir: but I didn't hear about any tests regarding that, ping me if you find any :) Apr 25 11:05:36 Mm, because with SSDs flash indeed isn't directly exposed to OS like with embedded flash chip is Apr 25 11:08:56 correct, it's not what mtd gusy call "raw" NAND access Apr 25 11:09:13 in case of raw NAND access you need ubi or similar to handle flash wearing Apr 25 11:16:57 actually f2fs was developed for emmc chips, it was a samsung fs for android Apr 25 11:22:40 oh, i didn't know that Apr 25 11:23:43 https://www.kernel.org/doc/Documentation/filesystems/f2fs.txt Apr 25 12:52:33 Is there a rule to make all packages and stop. Don't bother building kernel or creating image? Apr 25 12:57:08 jow: is it possible to get the ToH as a json file including all specs? Apr 25 13:05:47 aparcar[m]: no Apr 25 13:07:56 ok, I'll write a parser then :) Apr 25 13:09:37 I'm using an external toolchain hower CONFIG_SOFT_FLOAT is not being set automatically Apr 25 13:09:59 does anyone know what I need to change, I'd like to integrate this correctly instead of setting environment variable each time which leaves potential for human error Apr 25 13:10:01 thanks in advance Apr 25 14:31:44 still struggling with Makefiles for my package. is there a variable to address the source directory? in Package/install i can use relative paths but not so in Package/install-overlay Apr 25 14:52:00 jow: ping Apr 25 14:54:50 charolastra: PKG_BUILD_DIR should point at it. Apr 25 15:02:59 no, that doesn't construct a valid path. here it points to openwrt/build_dir/target-mipsel_24kc_musl/happyweb-1.0./ while it should be openwrt/build_dir/target-mipsel_24kc_musl/happyweb-1.0/ipkg-mipsel_24kc/happyweb/ Apr 25 15:12:50 ynezz: so that devmem2-airos of you seems to work, but I'm a bit lost how to figure out the address I need to read Apr 25 15:20:08 ar934x so I think I should start at 0x18050000 Apr 25 15:30:38 Is the plan for ath79 on master to move to 4.19 after v19.xx is branched? Apr 25 15:31:03 that's the plan for all targets Apr 25 15:31:31 Thanks! Apr 25 15:55:48 ynezz: nvm, starting to make something out of it :) Apr 25 15:58:51 Very off-topic question, but I think this channel knows if any... in linux when one has VLANed interface does the MTU reported in the "ip addr" and alike including the 8 bytes that vlan "takes" or not? :) I'd say not, but this is something I haven't yet figured out Apr 25 16:04:32 Example MTU on "eth0.X" shows 1500, is it really 1508 total or 1492 payload + 8 VLAN tagging = 1500? :) Apr 25 16:23:17 sudhanshug: pong Apr 25 16:23:19 charolastra: how about PKG_INSTALL_DIR, then Apr 25 16:25:09 Nevermind, it should point to 'openwrt/build_dir/target-mipsel_24kc_musl/happyweb-1.0/ipkg-/ipkg-install' in your example. Apr 25 16:27:32 You can use $(PKG_BUILD_DIR)/ipkg-$(PKGARCH)/ Apr 25 16:36:17 jow: I am an aspiring student for GSoC this year with Freifunk. My project includes creating a module(application) for automatic upgrades. Currently, I am learning developing LuCI modules. I have read the official documentation at https://github.com/openwrt/luci/wiki but still feel like a great amount of knowledge is missing. What would you recomme Apr 25 16:36:17 nd to me for learning the same. Can you specify something like a "ladder" to learn the LuCI developments? Any help is greatly appreciated Apr 25 16:51:08 depends on what you want to achieve, I guess its mostly displaying info and provide an button to trigger the upgrade, so maybe look at how the system backup/restore pages are implemented Apr 25 16:51:13 best is to look at existing code Apr 25 17:11:50 okay Apr 25 18:09:39 ping jow Apr 25 18:18:37 cotequeiroz: https://downloads.openwrt.org/snapshots/faillogs/x86_64/base/cryptodev-linux/compile.txt is causing long range build failures Apr 25 18:19:22 hmm nvm. probably a buildbot issue Apr 25 18:20:26 I'll take a look at it anyway. Apr 25 18:21:29 other kernel modules fail in the same way Apr 25 18:22:52 OK. What I can do to streamline building of openssl is to add a conditional to the kmod-cryptodev depends in libopenssl-devcrypto. That way, cryptodev will only compile if the devcrypto engine is selected. Apr 25 18:26:26 Same with afalg (crypto-user) and padlock. Apr 25 18:30:02 looking at your mips64 patch. there are more targets that fail than just mips64 (a bunch of the older ARM + arc). Apr 25 18:30:25 then again, nothing supports arc Apr 25 18:30:37 Do you mean my node patch? Apr 25 18:30:41 yes Apr 25 18:31:36 I will take a look at that and include them as well. The list of supported CPUs is rather simple. Apr 25 18:33:22 diizzy's method for x264 was to limit the ARM stuff to anything with cortex in the name. Seems mostly correct for node. Apr 25 18:33:34 My only doubt is if mips64 will work with mips64el's definition. I know mips is bi-endian, but I imagine you have to pick one or the other for the whole system, and not mix and match the libraries/binaries. Apr 25 18:33:53 i really doubt it Apr 25 18:34:13 having said that Apr 25 18:34:46 I submitted a patch for libgpg-error making the mips64 definitions match mips64el. Nothing seems to have blown up Apr 25 18:36:40 I don't have the hardware to test it, so I won't venture into that. All MIPS support has been dropped in node 12 because V8 dropped them for lack of maintainers. Apr 25 18:38:40 Another thing is why would you make a distinction in the ARCH (mips64el vs mips64) if you could just mix and match? Just a default? I would not have created a separate ARCH just for that--think MIPS16. Apr 25 18:39:16 I really doubt you can mix and match Apr 25 18:39:28 mipsel binaries will not runs on mips Apr 25 18:40:09 same with armeb vs arm Apr 25 19:14:18 hey guys. is there a way to get initramfs images to build by default? Apr 25 19:14:53 yes Apr 25 19:15:13 vi ~/.openwrt/defconfig Apr 25 19:15:24 blogic: hey :) Apr 25 19:15:38 in there you can add stuff you want to end up inside your .config when generating a new one Apr 25 19:16:06 I have these 3 lines in there Apr 25 19:16:07 CONFIG_CCACHE=y Apr 25 19:16:07 CONFIG_DEVEL=y Apr 25 19:16:07 CONFIG_TARGET_ROOTFS_INITRAMFS=y Apr 25 19:16:23 so my tree always uses ccache and enables initramfs Apr 25 19:16:24 thanks. can one set that per target though? e.g. my fritzbox needs an initramfs image to flash from stock to openwrt Apr 25 19:17:52 i think you can add it to FEATURES:= Apr 25 19:18:31 ok i'll play with that. would a patch for that be accepted? so the buildbots build the initramfs image as well? Apr 25 19:18:47 not sure Apr 25 19:18:56 jow: ^^ Apr 25 19:19:21 Might also be able to add it to the device's IMAGES entry in target/linux//image/Makefile Apr 25 19:20:23 jeffsf: no Apr 25 19:20:35 well we could patch it i guess but that feature does not exist atm Apr 25 19:21:46 blogic: no as in it won't work or it's a hack, on jeffsf's suggestion? Apr 25 19:23:49 blogic: i'm seeing FEATURES:= including ramdisk in quite a few targets already, so that shouldn't be an issue i reckon? :) Apr 25 19:24:33 Borromini: yep, which target is this ? Apr 25 19:24:44 blogic: lantiq/xrx200 Apr 25 19:24:53 ironically all the other lantiq subtargets have it Apr 25 19:24:54 ah ok, go for it, there are not many images Apr 25 19:25:13 then move it from the subtargets to the master Makefile please Apr 25 19:25:45 yeah was gonna ask that :) Apr 25 19:27:56 mangix: I've sent the openssl patch. If, for some reason, cryptodev does not compile, it will not keep you from building openssl without the cryptodev engine. Apr 25 19:41:55 blogic: apparently lantiq has five subtargets, adding initramfs to xrx200 would make three out of five need them Apr 25 19:42:20 so i won't add it to the master Makefile Apr 25 19:42:25 maginx: The arm failures appear to be different. I'll investigate them further. Apr 25 19:55:04 "Thank you" to whoever renamed to image/legacy.mk so I can leave that in the past! Apr 25 19:56:28 I definitely appreciate the clearer format of the current image-generation makefiles Apr 25 20:02:57 pkgadd: ping Apr 25 20:03:39 Borromini: pong Apr 25 20:05:45 pkgadd: hey. just wanted to run something by you - is 192.168.178.1 the default IP for fritzboxes? Apr 25 20:07:07 i know my 7490 uses that, and it looks like the 7362 SL does as well, but i haven't fired the latter up yet Apr 25 20:08:23 Borromini: yes Apr 25 20:08:35 pkgadd: danke schön Apr 25 20:12:40 Borromini: the ISP branded firmware only uses a light branding, e.g. you can just ignore the ISP specific activation codes and skip to manual configuration, but you can also unbrand the the firmware (german language, but https://praxistipps.chip.de/fritzbox-branding-entfernen-so-gehts_46681 it basically boils down to connect to the bootloader ftpd during early boot (adam2/adam2) and issueing "quote Apr 25 20:12:46 SETENV firmware_version avm") - then the firmware should be close to identical to your 7490 (there are v7.01 versions and beta versions, https://avm.de/fritz-labor/) Apr 25 20:14:30 (the international version would be "avme") Apr 25 20:18:20 pkgadd: thanks. i wouldn't be doing much else besides flashing openwrt :) Apr 25 20:28:59 putting an unmanaged switch inbetween the router and your client will help you retain the IP address/ avoid link-training and make it easier to hit the right spot for connecting via ftp Apr 25 20:29:27 that will probably require a few takes, other than that it shouldn't be too hard Apr 25 20:52:38 yeah i'll use a spare switch Apr 25 20:53:03 btw... it seems using the ramdisk FEATURES flag doesn't simply enable ramdisk images to be built Apr 25 20:53:38 i'm looking at the makefiles but it's not clear why some targets get their initramfs images built and others don't :( Apr 25 20:53:48 Trying to get ath79 to support spi-nand. While I can hand-edit config-4.19, is there a cleaner way to enable the kernel config such as CONFIG_MTD_NAND_CORE, CONFIG_MTD_UBI, and the like? Apr 25 20:54:36 jeffsf: i think you can use CONFIG_KERNEL_$opt Apr 25 20:54:37 I see USES_UBIFS and NAND_SUPPORT, but a clear way too add it to a board or, more likely, the target Apr 25 21:01:06 Yeah, I'll hack it in for now -- I've still got to get it running well enough that I can confidently post the patches to linux-mtd Apr 25 21:02:16 (Add support for GigaDevice GD5F1GQ4UFxxG) Apr 25 21:13:44 Open thought for anyone listening -- Given that ath79 is common for flash-constrained devices, should the NAND-based kernels be split from NOR-only kernels? Apr 25 21:31:53 TIL about ~/.openwrt/defconfig Apr 25 21:32:14 manginx: I don't get it. How can you select an arm-xscale arch and end up with FPU=y? I can't find a profile that fits that. node depends on FPU. arm-xscale is only used for TARGET_ARCH_PACKAGES if TARGET_kirkwood or orion (which is broken). Apr 25 21:52:33 jeffsf: that probably depends on how much larger the images actually get with NAND enabled (keep in mind that ath79-tiny is already required for the really ressource constrained devcies) and if there are other unwanted side effects of enabling NAND support Apr 25 21:53:27 Agreed Apr 25 21:54:03 cotequeiroz: no idea Apr 25 21:54:19 I don't think I've dealt with arm_xscale Apr 25 21:59:34 jeffsf: I wouldn't worry too much about that question until you have something ready to merge, refactoring NAND support into a dedicated sub-target should be easy if that's the results of the discussions for the merge request (after all the hard part is getting it working, not moving the finished parts into a dedicated sub-target). Apr 25 22:01:17 pkgadd: I dread the linux-mtd submission -- but at least I can recognize the spi-nand flash. Reworking the image build from legacy style now Apr 25 22:01:36 yep, understandable Apr 25 22:02:06 Just hoping I don't fry the thing before I get to that point! **** ENDING LOGGING AT Fri Apr 26 02:59:56 2019