**** BEGIN LOGGING AT Fri Sep 09 02:59:57 2011 Sep 09 07:26:09 nbd * r28202 /trunk/package/mac80211/patches/586-ath9k_fix_rx_flush.patch: ath9k: fix an rx path race condition that might be the cause of the "ath: DMA failed to stop in 10 ms" logspam as well as various connection stability issues Sep 09 07:26:59 nbd * r28203 /branches/backfire/package/mac80211/patches/586-ath9k_fix_rx_flush.patch: ath9k: fix an rx path race condition that might be the cause of the "ath: DMA failed to stop in 10 ms" logspam as well as various connection stability issues (backport of r28202) Sep 09 09:33:48 if I have a linux app that can compile with $(CC) -Wall -g `pkg-config --cflags --libs libmodbus` $< -o $@ Sep 09 09:34:07 what's the magic variable to put to run pkg-config, when it's called from the openwrt buildchain? Sep 09 09:34:40 I've got my app to build by manually putting in the -L and -I paths to the staging_dir/target/usr/[lib,include] Sep 09 09:35:01 but I see that there's pkg-config in the host tools as well, so it would be nice to be able to use it Sep 09 11:47:16 karlp: $(STAGING_DIR)/host/bin/pkg-config Sep 09 11:48:43 ahh, so nothing magic like the way $(CC) magically does the right thing then? Sep 09 11:52:38 well PATH is set Sep 09 11:52:46 at least it should be Sep 09 11:52:55 configure usually picks it up Sep 09 15:56:01 linux notes for gpio refer to gpio in sysfs, but I only seem to have it as gpiodev, is that expected? is there something I can enable to use the sysfs interface? Sep 09 17:09:48 hmm, more importantly, gpio_dev.h doesn't get copied into the staging_dir, Sep 09 17:10:15 jow * r28204 /branches/backfire/target/imagebuilder/files/Makefile: [backfire] merge r27872 Sep 09 17:20:05 karlp: you mean like CONFIG_GPIO_SYSFS=y Sep 09 17:23:43 karlp: gpio_dev is way obseleted Sep 09 17:23:45 dont use it Sep 09 17:24:01 use the /sys/class/gpio/... interface instead please Sep 09 17:24:23 gpio_dev was a ugly hack from a time when CONFIG_GPIO_SYSFS did not exist Sep 09 17:24:32 ... please Sep 09 17:30:45 jow * r28205 /trunk/target/imagebuilder/files/Makefile: Sep 09 17:30:45 [imagebuilder] Various enhancements from Manuel Munz Sep 09 17:30:45 - don't regenerate Package index unless the packagedir changes Sep 09 17:30:45 - remove opkg lists from rootfs, saves > 1MB uncompressed Sep 09 17:30:45 - don't direct errors on index creation into packagelist Sep 09 17:33:23 jow: ugh Sep 09 17:33:42 there's a dicussion on this Sep 09 17:33:47 s Sep 09 17:34:36 certainly stripping opkg lists with impunity is wrong Sep 09 17:35:12 jow * r28206 /branches/backfire/target/imagebuilder/files/Makefile: [backfire] merge r28205 Sep 09 17:35:19 sigh Sep 09 17:36:49 jow: are you still ignoring me? Sep 09 17:37:05 nbd: pong Sep 09 17:37:31 Chocky: ping Sep 09 17:37:37 (reverse ping/pong) Sep 09 17:37:39 yes Sep 09 17:37:57 jow's last commit is at odds with the last patch you made from me to imagebuilder Sep 09 17:38:33 jow_laptop: shouldn't the opkg file stripping be optional? Sep 09 17:38:52 so, now we have the optional remove bit, *and* the forced removed Sep 09 17:39:38 nbd: the lists not imo Sep 09 17:39:45 nbd: theyre more or less temporary data Sep 09 17:39:49 hm Sep 09 17:40:05 <__trine> jow_laptop, is going to finish up with the channels biggest ignore list :) Sep 09 17:40:05 a normal system does not even populate /usr/lib/opkg/lists/ Sep 09 17:40:07 it uses /var Sep 09 17:40:11 oh, right Sep 09 17:40:15 was thinking of something else Sep 09 17:40:17 alright then Sep 09 17:40:20 hm, ok Sep 09 17:40:46 jow_laptop: Chocky is wondering if you're ignoring him Sep 09 17:41:22 probably Sep 09 17:41:32 he's still upset I challenged the purity of uclibc :p Sep 09 17:41:39 anyway Sep 09 17:41:43 <__trine> nbd ask for me as well Sep 09 17:41:46 <__trine> heh Sep 09 17:42:08 Chocky: purity? Sep 09 17:42:12 __trine: I'm ignoring you Sep 09 17:42:26 nbd: I'm being facetious Sep 09 17:42:56 well, uclibc is a bit crappy in a few places, but it's the only suitable libc for small devices Sep 09 17:43:00 nbd: anyway, thanks for clearing that up Sep 09 17:43:24 so it's useful to make as much stuff as possible work with it and fix bugs when they show up Sep 09 17:43:25 that's true enough, but it's not suitable for everything. Sep 09 17:43:45 it's suitable for most openwrt users though Sep 09 17:43:45 I've never debated any of that. but it's not suitable for my usage Sep 09 17:43:55 why not? Sep 09 17:44:07 it's a whole lot slower with PHP. Sep 09 17:44:17 that's probably not hard to fix Sep 09 17:44:22 or at least, some of the PHP stack/library interactioni Sep 09 17:44:37 * __trine is in a good mood today as he won at court again Sep 09 17:44:44 __trine: congrats Sep 09 17:44:55 yes, likely. I'm not sure how to come up with a credible test case I could share though. Sep 09 17:45:05 Chocky: just do some profiling with your app Sep 09 17:45:10 I did. Sep 09 17:45:14 it's nearly 1/2 the speed Sep 09 17:45:47 https://forum.openwrt.org/viewtopic.php?pid=140543 Sep 09 17:46:07 but it's stuff I can't share. and it's pretty complex Sep 09 17:46:23 profiling, meaning looking at which functions steal the most cpu time Sep 09 17:46:26 might be sqlite or anything Sep 09 17:46:36 maybe its already over, but i think we should always delete packagelists like in the commit from jow Sep 09 17:46:50 Chocky: you are peter? Sep 09 17:46:53 especially in libc Sep 09 17:47:07 hrm, 3g/edge connection starting to fail again Sep 09 17:47:23 coverage in the northern part of germany sucks, at least if you're on a train Sep 09 17:47:35 nbd: understood. I did a lot of that for unrelated reasons at the PHP level; didn't get down into C functions. libc works for me now, so it's a bit hard to jsitufy the time Sep 09 17:47:46 soma: for my sins Sep 09 17:48:13 Chocky: well, uclibc will always receive more testing than (e)glibc Sep 09 17:48:22 fine. i'll try to explain better on a reply what my problem is, was a bit in a hurry before. maybe nbd can check this then, too Sep 09 17:49:07 nbd: I know. But it seems like we're trying to construct an argument to not support e/glibc. Sep 09 17:51:50 not really Sep 09 17:52:18 soma: the commit is fine, i only got mixed up with lists vs info Sep 09 17:52:30 my fault for confusing him Sep 09 17:55:36 nbd: well, the point I definitely do have, is that e/glibc support remains bust in SVN, and my outstanding patches address that. Even if some of them are messy, and need proper fixes. Sep 09 17:56:16 yeah, i still don't get around to going through patches very often Sep 09 17:56:34 i spend most of my time working on ath9k and netifd Sep 09 17:56:56 though netifd looks like it'll be mostly complete soon Sep 09 17:57:04 aye Sep 09 17:57:25 * nbd is so looking forward to getting rid of our crappy network setup scripts Sep 09 17:57:31 I think the .4 kernel bump broke USB serial support, at least for me, on ar71xx. Will report a bug on that later today Sep 09 17:57:57 and the PHP bump has broken something obscure for me too. Looking at that too Sep 09 18:04:01 moo Sep 09 18:05:10 http://geekandpoke.typepad.com/.a/6a00d8341d3df553ef01543533e604970c-pi Sep 09 18:06:19 are you saying that big endian is a pig? and small endian is bacon? Sep 09 18:06:41 no, it's not about the pig Sep 09 18:06:44 it's about the picture of the pig Sep 09 18:06:53 oh, about sniffing its own butt? Sep 09 18:07:27 if it was a donkey instead, I could say something about it being half-assed Sep 09 18:07:31 * nbd doesn't lije to explain jokes Sep 09 18:07:38 s/lije/like/ Sep 09 18:24:44 jow * r28207 /trunk/package/hostapd/files/hostapd.sh: Sep 09 18:24:44 [package] hostapd: introduce "wpa_pair_rekey" and "wpa_master_rekey" options, Sep 09 18:24:44 remove hardcoded wpa rekey intervals, remove hardcoded Sep 09 18:24:44 radius_acct_interim_interval as it might overrule the radius Sep 09 18:24:44 Acct-Interim-Interval attribute Sep 09 18:26:54 jow * r28208 /trunk/package/broadcom-wl/files/lib/wifi/broadcom.sh: [package] broadcom-wl: implement wpa_group_rekey option (mapped to nas -g) Sep 09 18:26:57 philipp64|laptop: where is CONFIG_GPIO_SYSFS found? I can't find it anywhere in menuconfig, nor in kernel_menuconfig yet. Sep 09 18:27:24 blogic: I'm more than happy to use sys/class/gpio, I just had working code from gpiodev from gpioctl :) Sep 09 18:28:24 is sysfs going to be much faster than gpiodev? it's currently far to slow to use as a software RTS pin for a 485 driver on the serial port. Sep 09 18:28:35 I'd rather not hack it with timing delays too much Sep 09 18:36:49 or will I really have to make my own new serial driver, that can use kernel calls to twiddle the gpio? Sep 09 18:38:10 karlp: it's in your target's config-default file. Sep 09 18:39:48 aren't there USB to RS485 dongles? Sep 09 18:40:07 sure, but this platform doesn't have USB Sep 09 18:40:38 from the help, it seems that sysfs for gpio requires either ARCH_WANT_OPTIONAL_GPIOLIB or REQUIRED Sep 09 18:40:46 which doesn't seem to be defined for atheros Sep 09 18:41:37 by target's configdefault, you mean the target/linux/atheros/config-2.6.30 file right? Sep 09 18:44:57 ah, my brain's full enough for this week anyway, back to this on monday Sep 09 19:40:36 i plan on running debian on these, but i have a bunch of kirkwood systems Sep 09 19:40:59 if people want to nominate another person to get one, i'd be happy to give some love to the https://dev.openwrt.org/wiki/WantedHardware **** ENDING LOGGING AT Sat Sep 10 02:59:58 2011