**** BEGIN LOGGING AT Thu Mar 28 02:59:57 2019 Mar 28 06:27:06 mangix: thanks, so we should figure out what went wrong and fix it, instead of using it as excuse to switch to proprietary codeload Mar 28 06:39:59 jow: one theory was that his gzip binary was using a different default compression level Mar 28 06:42:01 huh interesting. download.mk doesn't specify a compression level for gzip of bzip2 Mar 28 06:42:15 *or Mar 28 09:32:31 jow: could I please get your ack for https://github.com/ynezz/openwrt-fstools/commit/51454459c2d535cdb3b73a3de98e5732d1714d96 and https://github.com/ynezz/openwrt-fstools/commit/72a783e45b74fcc52848a0b16d61afde9c3e96d1 ? Mar 28 11:40:54 ynezz: feel free to add my acked-by Mar 28 11:58:52 pushed, thanks Mar 28 12:13:59 ynezz: before you bump the fstool packaging, I'd like to push some other fixes too Mar 28 12:14:15 ah you already did. nvm then Mar 28 12:34:35 sorry, fast fingers Mar 28 12:50:01 mangix: about 8513, the uncompressed tar files were different, so it may be a tar issue. Mar 28 12:52:07 yeah i checked. looks like it. the gzip sizes were not that different. so that's not the issue Mar 28 12:57:52 Hi people. Mar 28 12:59:24 the router maker Wally have sent me a mesage on the forum Mar 28 12:59:57 asking me if I would like any of there kit to get it to work with openwrt Mar 28 13:00:47 Is there any one who woulg like to have a look at what they can do as I cant code and I don't want to take any things if I cant help with coding Mar 28 13:01:19 March 28 Mar 28 13:01:19 Thanks very much for your kindly reply. Could you please help to suggest how to submit code ?As we don't know how to operate it. Mar 28 13:01:19 http://www.wallystech.com/ this is some of our main products, please have a review. We can also ship the samples to you if you are interested in it. Mar 28 13:01:19 By the way, Mar 28 13:06:52 i noticed on HEAD/openwrt-18.06 for xrx200: Symbol: PACKAGE_hostapd-mini [=n] (Wireless will not work without it because of missing /usr/sbin/hostapd) is that the default on purpose? shouldn't it be =y? Mar 28 13:07:32 no, we use wpad normally, which combines hostapd and wpa_supplicant Mar 28 13:08:14 for me on master CONFIG_PACKAGE_wpad-basic gets seletced on the xrx200 subtarget Mar 28 13:08:34 wpad-mini shows for me Mar 28 13:08:45 JyZyXEL:but the docs could be expanded for that, sure. Mar 28 13:08:56 and on the built image i get "* opkg_files_cmd: Package wpad-mini not installed." Mar 28 13:10:50 i suppose the wpad doesn not get included in the sysupgrade image if its set as a module? Mar 28 13:14:19 so probably the right thing to do is to set "Symbol: PACKAGE_wpad-mini [=y]" and leave hostapd-mini out? Mar 28 13:16:05 karlp: ah, the head of 18.06 doesn't yet have the wpad-basic, so it's probably some kind of migration happening from hostapd -> wpad? Mar 28 13:17:50 it's been wpad for years now, might have been just wpad, with mini being a cut down version. Mar 28 13:22:03 hello guys Mar 28 13:22:19 I got an mt7612 mini pcie card and I wanted to use it on my laptop Mar 28 13:22:27 but it looks like its getting some problems Mar 28 13:22:34 Im running kernel 5.0.5 Mar 28 13:23:13 “ping: sendmsg: No buffer space available” Mar 28 13:23:38 can I use the mt7612 driver from openwrt repo? Mar 28 13:23:46 s/repo/git Mar 28 13:47:25 jow, mangix: about 8513, the tar file differences are that the directories have SGID set. Mar 28 14:00:59 cotequeiroz: the new one has it set and the old one with the bad checksum not, or vice versa? Mar 28 14:04:37 The old one has it set. I think the solution is simple: add --mode=a-s to the tar command. Mar 28 14:13:38 I can confirm that adding the flag produces the same file as the new one, even when doing it from a sgid directory. Mar 28 14:13:43 Tar is used in download.mk, from which it is run, and also from image.mk. Should this be added to both invocations? Mar 28 14:16:11 jow:should ABI_VERSION get full versions lik "2.0.1" ? if that's what the upstream makefiles are creating for .sonames? Mar 28 14:21:53 is anyone familiar with setting up a nand partition spec in a dts file? i'm struggling to figure out what's required, and how to find values Mar 28 14:22:53 it's bcrmnand, and the soc is 63238 Mar 28 14:23:03 63268 sorry Mar 28 14:24:56 donn_: bcm63xx/nand uses values from nvram/cfe; you will need to write a parser for that Mar 28 14:26:04 KanjiMonster: ok, does that mean i can have a dts without a nand entry then? Mar 28 14:26:25 can you point me to an example parser? Mar 28 14:27:21 donn_: https://elixir.bootlin.com/linux/latest/source/drivers/mtd/bcm63xxpart.c does the parsing for NOR based bcm63xx devices Mar 28 14:27:31 you will need to do something similar for nand Mar 28 14:30:55 thanks Mar 28 14:31:27 out of curiousity, why can't this info come from dts? Mar 28 14:33:00 because dts is static, while the partition information in the nvram is what the bootloader actually uses Mar 28 14:38:00 cotequeiroz: yes please! Mar 28 14:39:03 karlp: it should follow SONAME Mar 28 14:39:28 karlp: e.g. on my desktop (random library chosen): Mar 28 14:39:30 $ readelf -d /usr/lib/libkidletime.so.4.14.26 | grep SONAME 0x000000000000000e (SONAME) Library soname: [libkidletime.so.4] Mar 28 14:39:35 => ABI_VERSION:=4 Mar 28 14:40:19 readelf -d doesn't have SONAME for this... Mar 28 14:40:27 the makefile writes .so.2.0.1. Mar 28 14:40:40 what library is it? Mar 28 14:40:59 https://github.com/romanbsd/statsd-c-client Mar 28 14:41:19 I imagine most people using C just re-implemented it in their own, never actually bothered linking it. Mar 28 14:41:25 but I thought I'd start easy. Mar 28 14:42:39 the above makefile looks a bit like cargo-culting, yes Mar 28 14:43:17 yeah, it's nothign nice. Mar 28 14:43:19 KanjiMonster: i've built an initramfs image that boots, is there a way to probe/read the nand when that's running? Mar 28 14:43:21 usually, when linking, ld will use the value of the SONAME elf section of the resolved .so file to decide what to put in dt_needed of the application being linked Mar 28 14:43:34 when no SONAME is present, it'll simply link foo.so Mar 28 14:43:40 the version symlink serves no purpose Mar 28 14:43:54 right, that makes sense. Mar 28 14:44:01 donn_: if it's in the dts, it will be probed Mar 28 14:44:12 I'm going to drop the symlink install and the ABI version for now in my own repo at least. Mar 28 14:44:22 they're just making symlinks because they thought that's how it worked. Mar 28 14:44:28 karlp: but back to your question, since that library apparently is not able to provide abi guarantees, it makes sense to set ABI_VERSION to its full relese version number, including minor and micro version Mar 28 14:45:00 should I still try and install the .so.ABIVERSION symlink? Mar 28 14:45:06 no Mar 28 14:45:14 KanjiMonster: i guess that means i'll just have a flat mtd device? Mar 28 14:45:27 donn_: FYI the main reason for brcm63xx not yet supporting nand is the braindead design decision to include most of the bootloader in the on-flash image (strict requirement - you can't flash an image without it) Mar 28 14:45:42 setting the abi version to the full release number will not prevent opkg file clashes but it will prevent accidential updates of the library to potentially incompatible versions Mar 28 14:46:31 an alternative approach is simply picking an ABI_VERSION (e.g. "0") and increment it whenever incomptabile symbol changes happen usptream - that requires close supervision from the package maintainer on every version bump though Mar 28 14:46:33 KanjiMonster: i see...fwiw it's this router https://openwrt.org/toh/zyxel/zyxel_p8702n Mar 28 14:47:03 but either way, without patching the lib to actually include a SOVERSION, it all is mood more or less Mar 28 14:51:46 jow: how about https://github.com/cotequeiroz/openwrt-1/commit/3d09bcfca3426fff43a034f3c7cff985c679e0c6 Mar 28 14:52:18 ccache_cc -fPIC -o libstatsdclient.so.2.0.1 -shared -D LINUX -D AMD64 statsd-client.c Mar 28 14:52:36 hardly going to be creating the right thing anyway there. crap makefiles getting too much from the host Mar 28 14:53:01 I reproduced the old archive with master by setting tmp sgid, then applied the patch and ran it again, ending up with the right file. Mar 28 14:53:17 isn't PKG_INSTALL:=0 meant to stop it from running "make install" ? Mar 28 15:01:46 cotequeiroz: that looks quite well. I think we only need some confirmation from an OS X user that it works there as well (or do we invoke gtar there, anyway?) Mar 28 15:10:59 jow: we ship our own tar in tools Mar 28 15:36:09 Hauke: ah right, I forgot already. cotequeiroz: your patch should be fine then Mar 28 15:45:39 I'm trying to sync kernel configs for armvirt/64, doing `make kernel_oldconfig CONFIG_TARGET=subtarget; make kernel_oldconfig CONFIG_TARGET=target' is that correct? Mar 28 16:17:24 oh, heh, PKG_INSTALL=anything_at_all_even_0 does a "make install". Mar 28 16:17:34 you need to make sure it's not defined at all if you want to turn it off. Mar 28 16:24:29 jow:ended up just overriding compile and everything :) https://github.com/remakeelectric/owrt_pub_feeds/commit/f54f982860317920c7f3fca7a02c6dc21ac8fba7#diff-cfc7fbdc4ae7b3ef5c29502f2f284cc3R39 Mar 28 16:28:25 karlp: ah, then you can add -Wl,-soname=libstatsdclient.so.2 as well while you're at it :) Mar 28 16:29:17 * karlp laughs Mar 28 16:29:40 if this works out I might just write a "new" implmentation of it Mar 28 19:49:42 Hauke: valgrind folks have fixed the mips32 revision detection, allowing it to be built with LTO enabled. I am about to refresh https://github.com/openwrt/openwrt/pull/1634 . Shall I enable LTO again? Mar 28 19:51:41 cotequeiroz: yes thanks Mar 28 19:51:52 but I think mips16 is still unsupported Mar 28 19:52:35 We can live with that, I suppose. Mar 28 19:52:57 I think they want to do a new release soon Mar 28 20:58:51 Hey all, I have a question. I got the zbt-we1326-BK which is only different from the regular one by having 256MB ram instead of 512. I know someone on openwrt had a fix to compile builds for the lesser ram, but I have spent like an hour looking for that post and I can not seem to find it. What should I change to make it work on the 256MB one? Mar 28 21:03:32 dedomraz: I don't know how those devices differ in ID's and such... but generaelly 256mb of ram is plenty well for openwrt, so in that sense it should just work... but what is the real problme in workings? does it not flash? does it not boot? what is the actual error Mar 28 21:04:24 The problem is that the original supported zbt we1326 comes with 512MB ram, and I just flashed that from my fresh compile, and it does not work. As in - no DHCP, no luci, no responding at 192.168.1.1 Mar 28 21:04:54 But my device only has 256MB ram, so I want to change from 512MB to 256MB basically Mar 28 21:07:08 Change dts memory node to Mar 28 21:07:09 memory@0 { Mar 28 21:07:09 device_type = "memory"; Mar 28 21:07:09 reg = <0x0 0x10000000>; Mar 28 21:07:11 }; Mar 28 21:07:14 ? Mar 28 21:07:19 While I can't be certain from there with only your text, I'd still say it has nothing to do with amout of ram, but something else being at fault... Mar 28 21:08:03 Thanks @Rene__ Mar 28 21:08:09 try that and then report how wrong I was ^ ;D Mar 28 21:08:10 I will try that right now Mar 28 21:08:48 Yeah from what I remember - thats what needs to be changed, just couldn't find the exact thread on OpenWRT :-) Mar 28 21:17:05 I just changed it, compiling now. About 99% sure that this will make it work :-) Will report in about 15 mins probably Mar 28 21:18:48 ok Mar 28 21:47:22 @Rene__: thank you so much!! It worked :-) Mar 28 21:47:43 @olmari: changing ram did work Mar 28 21:48:02 dedomraz: Great! Mar 28 21:54:05 nice :) **** ENDING LOGGING AT Fri Mar 29 02:59:57 2019