**** BEGIN LOGGING AT Mon Sep 24 02:59:59 2012 Sep 24 03:03:53 is there way to compile a single host tool? like packages/mksh/compile? I can't seem to feature out the correct target Sep 24 03:04:41 nm, freaking package not packages Sep 24 03:22:28 so miniupnpc breaks build on Darwin because it does a $(uname -s) in its Makefile, which returns Darwin, instead of Linux. I patched it by just hard coding "Linux" in place of the uname call. It works. Is that the best way to patch it? Sep 24 03:22:39 I ask before I send email to mailing list Sep 24 03:50:19 build #78 of brcm63xx is complete: Failure [failed compile_7] Build details are at http://buildbot.openwrt.org:8010/builders/brcm63xx/builds/78 Sep 24 06:36:21 build #69 of pxcab is complete: Failure [failed shell_11] Build details are at http://buildbot.openwrt.org:8010/builders/pxcab/builds/69 Sep 24 06:42:23 build #68 of rb532 is complete: Failure [failed shell_11] Build details are at http://buildbot.openwrt.org:8010/builders/rb532/builds/68 Sep 24 06:47:52 build #68 of ppc44x is complete: Failure [failed shell_11] Build details are at http://buildbot.openwrt.org:8010/builders/ppc44x/builds/68 Sep 24 07:29:07 build #50 of iop32x is complete: Failure [failed shell_11] Build details are at http://buildbot.openwrt.org:8010/builders/iop32x/builds/50 Sep 24 07:32:47 build #50 of etrax is complete: Failure [failed shell_11] Build details are at http://buildbot.openwrt.org:8010/builders/etrax/builds/50 Sep 24 09:07:52 build #48 of gemini is complete: Failure [failed shell_11] Build details are at http://buildbot.openwrt.org:8010/builders/gemini/builds/48 Sep 24 15:24:00 nbd * r33531 /trunk/package/openssl/ (3 files in 2 dirs): Sep 24 15:24:00 openssl: use asm SHA1, SHA256, AES on MIPS. Sep 24 15:24:00 On a lightly loaded 24Kc, as measured by 'openssl speed sha1', shows between Sep 24 15:24:00 27% and 120% speedup depending on block size. Sep 24 15:24:00 SHA1 is notably used in Transmission for piece verification. Sep 24 15:24:00 Patch by Catalin Patulea Sep 24 15:24:03 nbd * r33532 /trunk/package/base-files/files/etc/sysctl.conf: base-files: remove obsolete entries from sysctl.conf (#12236) Sep 24 15:26:20 cool, fellow romanian's patch :-) Sep 24 15:26:48 hi all Sep 24 15:26:57 hey there Sep 24 15:27:01 can anyone comment on my asio patch (submitted by email)? Sep 24 15:27:26 <[florian]> pocock: it's ok, but we are in the middle of a release, so we don't commit new packages, yet Sep 24 15:28:07 is there a branch where they can be committed until the release is done? Sep 24 15:28:46 <[florian]> not at the moment Sep 24 15:30:11 what would be the timescale for following up on them? As I also do upstream releases of reSIProcate, I'm keen to follow through on any issues that arise making it all work on OpenWRT Sep 24 15:32:27 pocock: as long as it takes Sep 24 15:32:28 ;) Sep 24 15:32:37 i am starting a test for the beta2 build today Sep 24 15:32:46 i would guess 2-4 weeks from now until we have a final Sep 24 15:33:21 I don't want to rush anybody or ask for guaranteed dates - but I'll be happy to come back to it in another 6 weeks then and just see how things are progressing Sep 24 15:34:07 there is no guarantee Sep 24 15:34:07 incidentally, I'm testing the 12.09 beta at the moment on WR1043ND, particularly the 3G modem support Sep 24 15:34:12 read the GPL license ;) Sep 24 15:34:19 pocock: good Sep 24 15:34:22 does it work ? Sep 24 15:34:27 it varies Sep 24 15:34:34 it works well with my HTC Desire Z Sep 24 15:34:45 it worked instantly, using the cdc_ether mode Sep 24 15:35:06 but with the TP-Link MA180, I'm getting a connection and then it drops, lcp echo timeout Sep 24 15:35:20 usually I can browse the web for 20-30 seconds and that's it Sep 24 15:36:06 tried a usb hub ? Sep 24 15:36:17 maybe your dongle eats too much power on the spikes Sep 24 15:36:23 no, you suggest a powered hub? Sep 24 15:36:53 yes Sep 24 15:37:15 is there any convenient way to measure power issues without a multimeter? Sep 24 15:37:55 no Sep 24 15:38:01 even a multimeter is bad for that Sep 24 15:38:05 you wont see spikes on that Sep 24 15:38:44 ok, I'll go out and get a hub tomorrow and see if that solves the problem Sep 24 15:39:10 make sure to get a powered hub Sep 24 15:39:47 otherwise, it seems to detect the dongle every time it boots, the only thing I'm not sure about is whether I should hardcode the serial port (this dongle has ttyUSB[0-4], and pppd must run on ttyUSB3) Sep 24 15:48:23 build #45 of octeon is complete: Failure [failed shell_11] Build details are at http://buildbot.openwrt.org:8010/builders/octeon/builds/45 Sep 24 17:32:15 nbd * r33533 /packages/libs/boost/Makefile: boost: move the + to the right place to fix build errors Sep 24 17:54:16 blogic: I've attached a powered USB hub now, but this doesn't completely solve the problem Sep 24 17:54:33 when I manage to get connected, the dongle works well now Sep 24 17:55:13 ok Sep 24 17:55:16 but it doesn't always work the first time the device boots up - sometimes I have to unplug the dongle from the hub, plug it in again, and click the `Connect' button in LuCI Sep 24 17:56:47 I'm guessing that because the hub provides a constant power supply to the dongle, it does not reset properly when I power cycle the router - is that possible? Sep 24 17:56:56 yes Sep 24 17:57:05 that is most likely correct Sep 24 17:57:14 ok, I'm testing that theory... Sep 24 17:57:31 power cycle the router and I'll see if it still has the post-modeswitch ID Sep 24 17:57:47 yes Sep 24 17:57:54 ATH will most likely help Sep 24 17:58:50 should I just wait for beta2 and try with that? Sep 24 17:59:15 that wont change anything Sep 24 17:59:21 ok, I can confirm that the post-modeswitch ID is still there after power cycling the router Sep 24 17:59:43 the problem is that the power provided by the MA180 is not enough Sep 24 17:59:50 if you're lucky your router can control usb power through a gpio Sep 24 17:59:53 the /dev/ttyUSB[0-4] are all there Sep 24 18:00:04 thats the only reliable way software way I know off Sep 24 18:00:20 but running gcom -d /dev/ttyUSB3 info Sep 24 18:00:21 software triggered usb port resets are usually not enough to actual bring back the dongle to its initial state Sep 24 18:00:21 fails Sep 24 18:01:12 now let me power cycle everything together and see if it comes up cleanly Sep 24 18:01:17 you can give the "usbreset" package a try, it sometimes helped in my tests Sep 24 18:01:28 it is probably also enough to replug the dongle Sep 24 18:01:45 yes, I can confirm that replugging the dongle makes it work Sep 24 18:02:00 but what I'm aiming for is to make this 100% stable for some non-technical users Sep 24 18:02:10 good luck with that Sep 24 18:02:16 ;) Sep 24 18:02:46 I already have a few OpenWRT boxes around various countries (in fact, on 3 continents now) running reliably, but none with 3G Sep 24 18:02:49 as I said, the only 100% stable way I am aware of is controlling the usb host power through a gpio Sep 24 18:03:05 then remove the power for 500-1000ms Sep 24 18:03:17 and bring it back, that will truly reset the dongle Sep 24 18:04:14 on the 1043nd gpio #1 controls the usb Sep 24 18:04:21 so try to toggle it through sysfs Sep 24 18:04:28 jow_laptop: he uses a powered hub Sep 24 18:04:43 the ma180 draw too much power and the 3G spikes kjill the dongle Sep 24 18:04:44 yes, it wasn't stable without the powered hub Sep 24 18:04:54 well then you're f*cked :) Sep 24 18:05:12 oioioi Sep 24 18:05:21 this is not #openwrt-swearing ;) Sep 24 18:05:27 I can confirm everything came up well together when I power cycled hub and router simultaneously Sep 24 18:06:22 I think I will just deploy it with one user for now, and when he gets fed up, he can subscribe to ADSL Sep 24 18:06:58 I've seen similar stability issues with Onda MT655 dongle, so I'm guessing it might work with a powered hub too Sep 24 18:08:06 can anyone recommend a dongle that is not so power hungry, or a router that can deliver more power over USB? Sep 24 18:08:16 during3G dialin the dongle starts to draw lots of power Sep 24 18:08:31 and the regulators on std tpl routers are cheap Sep 24 18:08:43 so they fail to provide enough power fast enough Sep 24 18:08:49 i did some basic testing a while ago Sep 24 18:09:11 the thing is that when increasing the power draw over the regulator slowly the regulator wont fail Sep 24 18:09:34 but if the slope is too steep it will simply breakin and not provided the needed power fast enough, thus killing the dingle Sep 24 18:09:42 if I disable wifi temporarily, will that allow the regulator to cope? Sep 24 18:09:56 and the 3g dial in is just short burst causing short spikes on the power consumption Sep 24 18:10:09 its not the max level you draw Sep 24 18:10:17 its the slope at which you require the power Sep 24 18:10:25 and this is also happening during operation mode Sep 24 18:10:38 as in, the dongle will veery few minutes have these spikes Sep 24 18:11:13 I chose the WR1043ND for the Gigabit switch and the 32MB RAM, can you suggest a better choice? Sep 24 18:11:36 it is hard to know about all these gotchas just from the information pages Sep 24 18:11:39 i am suprised the 1043 fail to be honest Sep 24 18:12:09 I have another one that was working more reliably with the dongle last week Sep 24 18:12:14 maybe they are not all the same Sep 24 18:12:34 it depends on where you are Sep 24 18:12:42 this one is v1.9, purchased in CH Sep 24 18:12:46 the better the 3g reception is, the less apparent the problem will be Sep 24 18:12:50 the other is a v1.8 purchased in Malaysia Sep 24 18:13:16 i used this recently with huawei and had no problems --> http://www.tp-link.com/en/products/details/?model=TL-MR3020 Sep 24 18:13:17 in this room, the reception is not so good, so maybe that is a factor too Sep 24 18:13:23 but its a tiny unit with no gbit Sep 24 18:13:36 the gigabit is just convenient, no need for an extra switch Sep 24 18:13:52 but the 32MB is essential for SSL (e.g. the reSIProcate package I'm working on) Sep 24 18:14:09 not just the 32MB, but the 8MB flash is necessary too Sep 24 18:20:56 ok Sep 24 18:21:11 no idea what a good unit would be Sep 24 18:21:23 maybe one of the wndr3700/3800 ones Sep 24 18:21:39 i think people have a good experienc with those Sep 24 18:22:19 I'll probably start one user with the 1043 while I get familiar with the other options Sep 24 18:22:25 a good unit is buffalo wzr-hp-ag300h Sep 24 18:22:33 and I may try some other dongles too Sep 24 18:22:38 nbd: does it have a gpio to toggle usb power ? Sep 24 18:22:38 because it doesn't have the hw version mess of the netgear routers Sep 24 18:22:45 yes, i think so Sep 24 18:22:46 let me check Sep 24 18:23:07 pocock: in that case the wzr-hp-ag300h would be a good choice Sep 24 18:23:15 yes, gpio2 is usb power Sep 24 18:23:19 but i cant guarantee, that the usb power prob does not show there Sep 24 18:23:41 pocock: however in general bufallo hw tend to be much better than tpl hw Sep 24 18:23:56 imho bufallo is japanese, while tpl is tw/cn Sep 24 18:24:08 it's all still built in china ;) Sep 24 18:24:23 qa is not done in china though Sep 24 18:24:44 but as i said, no guarantees, that the power wont break Sep 24 18:26:18 I can order that here, I don't mind giving it a try Sep 24 18:26:50 jow * r33534 /trunk/include/version.mk: [include] version.mk: ensure that %S is always populated with a subtarget identifier, fallback to "generic" if no subtarget exists Sep 24 18:27:13 actually, it has 128MB RAM and 32GB flash, so it is probably perfect for reSIProcate too Sep 24 18:27:43 I filled up the 1043 with OpenSSL, libstdc++, libdb47++ and reSIProcate itself Sep 24 18:28:02 and unlike the TP-Link, the buffalo device has tftp recovery in case something goes wrong Sep 24 18:28:04 had to run the sbin binaries from /tmp Sep 24 18:28:07 so you don't need serial to reflash Sep 24 18:29:04 I think it will be a good test to put it on the same workbench, with the same dongle, same crappy GSM signal that I've suffered with for 2 years... and see if it behaves or not Sep 24 18:31:06 I see some suppliers offering me the WZR-HP-G300NH (not wzr-hp-ag300h) - is there any difference? Sep 24 18:32:18 yes Sep 24 18:32:30 wzr-hp-g300nh only has one radio Sep 24 18:32:37 and is based on an older chipset Sep 24 18:32:49 and it also has less RAM Sep 24 18:34:10 i cant believe how cool runs my new-2-days wndr3800 with openwrt, all ports busy and 2.4 on Sep 24 18:34:37 same price though - but that is not unusual here in CH Sep 24 18:45:58 juhosg * r33535 /trunk/target/linux/ar71xx/ (7 files in 6 dirs): Sep 24 18:45:58 ar71xx: use different machtype and setup for the TL-MR3040 Sep 24 18:45:58 Should fix #12211. Compile tested only. Sep 24 18:47:23 anyway, the wzr-hp-ag300h appears to be available for delivery by Thursday or Friday, so I'll play with it at the end of the week **** ENDING LOGGING AT Tue Sep 25 02:59:58 2012