**** BEGIN LOGGING AT Thu Jun 18 02:59:57 2020 Jun 18 04:48:40 luci / 09a3b4730a769c43067ff1fb3e4aad339271fbc3 ("Translated using Weblate (Polish)") - my eyes are bleeding Jun 18 05:52:49 jow_: is this code used by LuCI at all? https://git.openwrt.org/?p=project/rpcd.git;a=blob;f=sys.c Jun 18 06:03:49 build #148 of sunxi/cortexa7 is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/sunxi%2Fcortexa7/builds/148 Jun 18 06:36:57 rmilecki: no Jun 18 06:37:56 rmilecki: iirc it was added sometime ago for some alternative gui. The "luci" references in there are copy-paste Jun 18 06:38:12 luci itself uses none of these procedures Jun 18 06:50:43 jow_: ok, thanks for update Jun 18 06:50:47 *info Jun 18 06:51:17 jow_: i assume fs.exec allows calling any command? Jun 18 06:52:22 jow_: should we switch to some task-specific ubus calls where possible & eventually deprecate/drop fs.exec? e.g. replace fs.exec('/sbin/reboot') with ubus call to "procd" "reboot" Jun 18 06:55:05 The packagelist function is used by attended sysupgrade Jun 18 06:56:11 rmilecki: it allows any command whitelisted in the ACL (check e.g. https://github.com/openwrt/luci/blob/master/modules/luci-mod-system/root/usr/share/rpcd/acl.d/luci-mod-system.json#L136 the "exec" permissions) Jun 18 06:56:26 jow_: ah, nice! Jun 18 06:58:26 aparcar[m]: what is attended sysupgrade? Jun 18 07:02:03 rmilecki: ideally something that keeps peoples devices updated https://github.com/aparcar/asu Jun 18 07:05:27 rmilecki: what you don't like on 'aktualizacji OR lub URL' ? :p Jun 18 07:23:14 jow_: is there a way to force linkage of a specific version of a library? I'm having issues with ppp and libcrypt Jun 18 07:28:00 mangix: none that I am aware of. If the libraries have the same basename it gets complicated Jun 18 07:45:02 h4m unfortunate Jun 18 07:45:28 you're trying to link xcrypt ? Jun 18 07:56:39 no. ppp does. I'm trying to force ppp to link to the libc libcrypt. Jun 18 07:57:28 but yes, I'm trying to introduce libxcrypt for shadow-utils Jun 18 07:58:49 the difference is in version. libxcrypt is .2 libc is .1 Jun 18 08:00:18 the only reliable way is staging libxcrypt in a nonstandard path Jun 18 08:00:41 e.g. $staging_dir/usr/lib/libxcrypt/libxcrypt.so.2 Jun 18 08:01:01 then change the ldflags of all packages that should use it Jun 18 10:44:08 I got flash to work! Was going on the assumptions of what the previous person had written in the hardware table, which to me seemed like would work with jedec Jun 18 10:44:14 it did not. Jun 18 10:45:08 only when i looked at the boot log more closely did I realise that it (or at least my device) has an entirely different flash, and needed brcm nand Jun 18 11:00:35 xx_ns: good progress! Jun 18 11:19:58 jow: do you prefer assigning "rpc.declare" returned objects to: variables OR object properties? Jun 18 11:20:10 e.g. startup.js does: callInitList: rpc.declare({ ... }) Jun 18 11:20:21 but system.js does: callInitList = rpc.declare({ ... }); Jun 18 11:20:40 rmilecki: I started settling on the former because it introduces less scope headaches Jun 18 11:21:02 ack Jun 18 11:21:14 often `this` is not available / annoying to reach (requiring binding all parent function scopes) Jun 18 11:21:49 so having the rpc helper assigned to variables makes them easier to call in all contexts Jun 18 11:22:29 wait, does "former" in english mean first or last? Jun 18 11:22:56 first :) Jun 18 11:23:14 like the alphabetical order of both words Jun 18 11:23:25 ah, so you *started* with formar, but then decided to go with variables? Jun 18 11:23:35 *former Jun 18 11:23:43 I started with assigning to object properties Jun 18 11:23:49 later I mixed Jun 18 11:23:56 recently I only assign to variables anymore Jun 18 11:24:08 now I get it, thanks! Jun 18 11:24:11 (ot: former is often used in conjunction with 'latter', which means the last thing) Jun 18 11:25:25 I guess I should start switching to ES6 Jun 18 11:25:37 that would solve most of the scope headaches at least Jun 18 11:27:26 i can't get used to arrow functions Jun 18 11:27:39 I hate the syntax as well Jun 18 11:27:49 but I like the fact that they don't have own scopes Jun 18 11:34:12 I see from the dts that nearly all of the devices similar to mine have a "wfi" partition, with 'compatible = "brcm,wfi"' - I tried googling it but didn't find anything informative, what is this and is it something I should look out for? Jun 18 11:35:22 currently I've mapped the partitions as was originally specied by the OEM, and that didn't specifically have a "wfi" partition Jun 18 11:42:20 xx_ns: probably git grep is your friend here too Jun 18 11:43:44 PaulFertser: yep, thanks Jun 18 11:44:13 turns out i probably do need it Jun 18 11:46:20 in acl.d should I use "read" vs. "write" basing on method (not) accepting arguments OR (not) providing result? Jun 18 11:46:56 read for things without side effects, write for actions with side effects Jun 18 11:47:05 thx Jun 18 11:47:15 e.g. "query wireless interfaces" -> read, "scan wireless" -> write (because it interrupts service) Jun 18 11:47:49 got it Jun 18 11:47:49 "list packages" -> read, "install package" -> write Jun 18 12:32:58 i have ubus method that returns UBUS_STATUS_OK Jun 18 12:33:13 when called using jsonrpc reply looks like this: Jun 18 12:33:19 [{"id":4,"jsonrpc":"2.0","result":[0]}] Jun 18 12:33:48 my LuCI code: rpc.declare({ object: 'foo', method: 'bar', expect: { result: 0 }}) Jun 18 12:34:23 i get: "The bar command failed with code 0" Jun 18 12:34:59 any hints? Jun 18 12:35:30 oh, envermind... Jun 18 12:35:37 error comes from .then() handler Jun 18 12:35:43 ignore all above Jun 18 12:58:40 can someone merge the olsr PR? https://github.com/openwrt-routing/packages/pull/580/files then I can finally compile olsrd again :D Jun 18 13:13:07 done Jun 18 13:36:14 build #181 of octeontx/generic is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/octeontx%2Fgeneric/builds/181 Jun 18 13:38:59 PaulFertser: apologies for bothering you once again, but how exactly is one meant to figure out the correct GPIO pins for the driver? I'm feeling there's got to be a better method than flat out guessing. I tried looking through the OEM fw bootlog for any hints, but I'm not exaclty sure what I'm looking for Jun 18 13:39:06 @ https://openwrt.org/toh/zte/zte_zxhn_h368n#oem_bootlog Jun 18 13:42:44 build #182 of malta/be is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/malta%2Fbe/builds/182 Jun 18 13:44:35 xx_ns: probably by connecting a voltmeter to MDIO pins of the switch and then trying to find it same as you find LED gpios? Jun 18 13:46:18 PaulFertser: ah okay, i'll do that, thanks! Jun 18 13:46:22 xx_ns: alternatively, probably there's some reference board schematics for this SoC. Or you can try finding OEM firmware source code. Jun 18 13:46:51 xx_ns: probably someone else here has a better idea, someone who had more experience with Broadcom than me. Jun 18 13:47:13 re: OEM source Jun 18 13:47:28 well, their sources site doesn't exactly load... https://opensource.ztedevices.com/ Jun 18 13:49:37 overloaded :p Jun 18 13:49:51 xx_ns: is the switch IC in a not-too-horrible package so that you can probe the pins? Jun 18 13:56:40 build #180 of mpc85xx/generic is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/mpc85xx%2Fgeneric/builds/180 Jun 18 14:30:48 build #180 of ramips/rt3883 is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/ramips%2Frt3883/builds/180 Jun 18 14:32:51 ynezz: if there is something we can help with from FreeBSD side just let me know Jun 18 14:35:22 decke: yeah, using GNU tools would make life easier :) Jun 18 14:35:46 ynezz: we know that CI with FreeBSD is not easily available but we're working on it and in the meantime we can also provide buildbots on FreeBSD etc. Jun 18 14:36:27 ynezz: yeah they are available in FreeBSD ports - 32555 - pick one Jun 18 14:36:49 well, if there's gitlab-runner for macOS I assume there is one for FreeBSD as well Jun 18 14:37:05 ynezz: gitlab yes, github no (.net issue) Jun 18 14:37:32 we have package builders running on gitlab so yes that works fine Jun 18 14:37:37 aparcar[m] is probing sr.ht as there seems to be FreeBSD available as well Jun 18 14:37:49 yes that works, circle.ci as well Jun 18 14:38:11 so it's just a matter of time/effort Jun 18 14:38:21 sorry cirrus.ci - I alway keep mixing them up Jun 18 14:38:53 yeah the .net stuff is a bit tricky, but .net 5.0 seems to be building fine on freebsd now :) Jun 18 14:39:21 so it's only a matter of packaging that up into the portstree Jun 18 14:39:31 then we can also run github actions on self hosted runners Jun 18 14:40:52 build #179 of ar71xx/mikrotik is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/ar71xx%2Fmikrotik/builds/179 Jun 18 14:42:32 build #177 of layerscape/armv8_64b is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/layerscape%2Farmv8_64b/builds/177 Jun 18 14:56:41 build #178 of mvebu/cortexa72 is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/mvebu%2Fcortexa72/builds/178 Jun 18 15:16:52 build #176 of armvirt/32 is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/armvirt%2F32/builds/176 Jun 18 16:09:24 aparcar[m]: I think i found why it does not compile - I have never clang and it pass checks Jun 18 16:09:44 4.2.1 Jun 18 16:09:54 -- /usr/bin/cc -dumpversion Jun 18 16:10:09 on my system 10.0.0 Jun 18 16:10:27 $(CC) -dumpversion | grep -E '^(4\.[8-9]|[5-9]\.?|10\.?)' Jun 18 16:12:30 it is very strange because clang 8.0.1, not 4.2.1 Jun 18 18:13:06 Hi, everyone! One question: any particular reason we're not yet building the userspace by default with -thumb for ARMv7 targets? Jun 18 18:13:38 don't know of any. Space constraints maybe? Jun 18 18:14:12 Well, I wager space is cheap on these targets. RAM on the other hand… Jun 18 18:14:29 … and i$, especially, that's another story. :) Jun 18 18:15:10 at least some time back we discovered that compressibility of mips16 is worse than that of standard opcode Jun 18 18:15:32 Thumb-2 is also less compressible, of course, since the encoding is denser. Jun 18 18:15:52 But ath79 targets are usually extremely space-constrained. Jun 18 18:16:07 are thumb-2 binaries runnable with non-thumb-2 libraries? Jun 18 18:16:51 I don't see why not. It's just a separate instruction set which the CPU supports. Jun 18 18:17:56 then there's probably no reason. I'd guess current toolchain defaults were inspired by the respective vendor SDKs Jun 18 18:17:58 Actually, I'm running a full (kernel + userspace) Thumb-2 build on my Omnia, at the moment. Jun 18 18:18:38 Great! I'll cook a patch for it soon. Jun 18 18:19:42 I already sent a patch a couple of hours ago enabling Thumb-2 for the kernel itself on ARMv7 MVEBU targets. Jun 18 18:22:33 My vmlinux is over *two megabyes smaller* with Thumb-2 when compared to pure ARM. That's a lot of code. Jun 18 18:24:41 The compressed image, howerver, is about 300 kiB larger. I believe it's a reasonable trade-off for these machines. Jun 18 18:29:08 rsalvaterra: please check the libraries compatibility, there's a thing called "internetworking" and probably there're some additional measures needed to get it to work. Jun 18 18:29:31 Because the CPU needs to be actually switching between Thumb and ARM ISAs. Jun 18 18:30:31 E.g. when you use "bx" instruction it'll take care of switching to Thumb if the lowest bit of address is set, and switching to ARM otherwise. But regular "b" (or "bl") won't do that, and crash. Jun 18 19:06:31 Does ARM/Thumb interworking make sense in ARMv7? AFAICT it's an ARMv5 (pre-Thumb-2) thing, where you couldn't mix both ARM and Thumb states. Jun 18 19:07:14 https://gcc.gnu.org/onlinedocs/gcc/ARM-Options.html Jun 18 19:14:58 And according to the documentation… http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.ddi0344c/Beiiegaf.html Jun 18 19:15:17 Ivan_83: so what should I do? Install a never GCC version? Jun 18 19:15:17 "Thumb-2 is an enhancement to the 16-bit Thumb instruction set. It adds 32-bit instructions that can be freely intermixed with 16-bit instructions in a program." Jun 18 19:15:31 There is no ARM or Thumb-2 state. Jun 18 19:15:49 rsalvaterra: so only needed on pre-v5, ok. Still worth runtime checking compatibility I'd say. Jun 18 19:16:15 rsalvaterra: Thumb-2 is indeed part of Thumb. But Thumb is not part of ARM ISA. Jun 18 19:18:28 jow: what brought the sudden merge of https://github.com/openwrt/openwrt/pull/2271 GPIO labels? Jun 18 19:19:35 Oh, yeah, about GPIO… that's another can of worms on the Omnia with OpenWrt master. Jun 18 19:22:27 I had to compile the kernel without support for PCA9538 interrupts. Sometimes they go unhandled and the kernel screams bloody murder. Jun 18 19:22:36 rsalvaterra: there's still the T bit in the CPSR and it does switch with instructions like "bx", and same code would be interpreted differently based on the T state. Jun 18 19:23:08 aparcar[m]: please wait, not i stuck with Cannot use CONFIG_CC_STACKPROTECTOR_REGULAR: -fstack-protector not supported by compiler Jun 18 19:23:21 *now Jun 18 19:23:54 I do not want install gcc, it is very easy ) Jun 18 19:24:48 I can use clang on the VM as well, just tell me what to install Jun 18 19:25:47 if you want - you can try, I already update PR Jun 18 19:25:56 PaulFertser: again, from the documentation, «Thumb-2 instructions are accessible as were Thumb instructions when the processor is in Thumb state, that is, the T bit in the CPSR is 1 and the J bit in the CPSR is 0.» Jun 18 19:26:05 readme contains more packages to install Jun 18 19:26:17 Ivan_83: excellent Jun 18 19:28:42 There's no concept of "state" in Thumb-2 from ARMv6T2 onwards. Jun 18 19:30:55 rsalvaterra: yes, I do not argue that, Thumb-2 can be freely intermixed with Thumb as it's a Thumb extension. But I'm talking about potential compatibility issues between code using ARM instructions (with T unset) and code using Thumb (including Thumb-2) instructions. If everything is Thumb-2 the CPU never switches so it's covered, agreed. And on Thumb-2 capable core it doesn't make sense to use Jun 18 19:31:01 plain old ARM. But if you consider binary compatibility with old code the picture changes I tihnk. Jun 18 19:31:19 rsalvaterra: you just cited a doc with the words "processor is in Thumb state" Jun 18 19:31:34 Thumb. Not Thumb-2. ARM is very clear about the distinction. Jun 18 19:31:46 Ivan_83: https://builds.sr.ht/~aparcar/job/236228 here we go again Jun 18 19:31:53 rsalvaterra: Thumb-2 is a superset of Thumb. And Thumb-2 instructions are executed in Thumb state. Jun 18 19:32:20 My point is there're two distinct states: ARM and Thumb/Thumb-2. Jun 18 19:34:49 Ivan_83: so no GCC required at all> Jun 18 19:36:08 aparcar[m]: pkg version -v | wc Jun 18 19:37:11 Ivan_83: ? Jun 18 19:37:19 i see packeages but not see packages deps while installing on bootstrap Jun 18 19:38:49 python and perl must be installed as dep package Jun 18 19:39:00 and many other Jun 18 19:42:50 Ivan_83: perl and python are installed Jun 18 19:47:08 are you sure? show: pkg version -v Jun 18 19:47:15 PaulFertser: Whatever the case, this is something for the compiler to worry about. And gcc is doing the right thing, in both kernel and userspace, otherwise my router wouldn't be running the image I compiled. Jun 18 19:48:35 Ivan_83: http://sprunge.us/uy2fo3 Jun 18 19:51:08 rsalvaterra: I agree that's up to the compiler (and probably linker) to do the right thing. The question is whether you can freely mix your binaries and libraries with binaries and libraries downloaded from downloads.openwrt.org repository. Jun 18 19:52:00 aparcar[m]: can you also show: env Jun 18 19:52:00 before run CC=/usr/bin/cc ./scripts/feeds update Jun 18 19:52:23 and /usr/local/bin/gawk /usr/bin/awk, CC=/usr/bin/cc - not needed Jun 18 19:57:50 PaulFertser: Wait, we are worried about *closed source* binaries now? I'd say, if it breaks, you get to keep the pieces. Jun 18 19:59:12 Ivan_83: removing the gawk steps brings back the missing asort problem here https://builds.sr.ht/~aparcar/job/236239#bottom I'll stop for now, if you're interesting in making FreeBSD work in a CI you can use this manifest and tests build at sr.ht https://builds.sr.ht/api/jobs/236239/manifest . I have no bsd running on my own so I can't test it any other way Jun 18 19:59:27 rsalvaterra: no, I can't care less about closed source binaries. But if your change affects compatibility with binary code it should be explicitly noted as some care needs to be taken even if we consider OpenWrt upstream ecosystem only. Jun 18 20:01:22 aparcar[m]: ok, thanks Jun 18 20:02:37 PaulFertser: Sure, but the master branch is in constant flux, right? If these changes actually break existing libraries, we just have to wait for a complete rebuild cycle. Jun 18 20:03:19 rsalvaterra: not sure, jow will tell more. What I'm sure about is that it's much better to know in advance if it will break or not. Jun 18 20:03:20 And this is all hand-waving, I haven't actually seen any proof that this is something we should worry about. Jun 18 20:44:47 is there somewhere a good documentation about blob buffers? do I have to free them if I have a global variable? Should I only use local blob buffers, do a init, and after this a free? I have global ones (because of tutorial and other code) and I'm not sure if I maybe have a memory leakage Jun 18 21:44:34 I cant see that bump to Linux 5.4.47 in master! /s Jun 18 21:45:09 aparcar[m]: I began going through open PRs from the back Jun 18 21:45:25 aparcar[m]: and that one didn't look like it had any blockers Jun 18 21:49:45 Tapper: 5.4.47 has only been released yesterday... (if you want it nevertheless, ldır has mailed a patch to the mailing list) Jun 18 21:50:24 pkgadd did you see the ancer to his mail? Jun 18 21:52:17 That's why I put /s at the end of my statement /s =sarcasm Jun 18 21:53:08 yes, I did read them - but I'm apparently not good at reading tags like that Jun 18 22:27:01 how can i disable postint scripts when building an image? Jun 18 22:38:53 Hm… Has anyone tried to build master with CONFIG_USE_MKLIBS? Last time I tried (last week, I think) it completely broke the image, while it worked fine in 19.07.x. **** ENDING LOGGING AT Fri Jun 19 02:59:57 2020