**** BEGIN LOGGING AT Wed Sep 28 02:59:56 2011 Sep 28 03:56:19 build #55 of rdc is complete: Failure [failed compile_7] Build details are at http://buildbot.openwrt.org:8010/builders/rdc/builds/55 Sep 28 06:04:27 build #55 of mpc52xx is complete: Failure [failed compile_10] Build details are at http://buildbot.openwrt.org:8010/builders/mpc52xx/builds/55 Sep 28 09:48:09 build #85 of s3c24xx is complete: Failure [failed compile_10] Build details are at http://buildbot.openwrt.org:8010/builders/s3c24xx/builds/85 Sep 28 11:40:31 build #77 of ps3 is complete: Failure [failed compile_6] Build details are at http://buildbot.openwrt.org:8010/builders/ps3/builds/77 Sep 28 13:44:14 build #73 of ppc44x is complete: Failure [failed compile_4] Build details are at http://buildbot.openwrt.org:8010/builders/ppc44x/builds/73 Sep 28 14:18:17 [florian]: are you sure I don't need to be using gpio_chip methods top use gpiolib and sysfs? As best I can tell, the board patch for ar231x adds it's own routines with the same names in arch/mips/include/asm/mach-ar231x/gpio.h Sep 28 14:18:52 <[florian]> karlp: for gpiodev, sure you do not, but to implement gpiolib yes Sep 28 14:19:00 ahh good. Sep 28 14:19:06 I misunderstood yesterday then :) Sep 28 14:19:10 yes, gpiodev works right now, Sep 28 14:19:14 <[florian]> karlp: sorry now that I rethink about your question yesterday I did not see the ambiguity Sep 28 14:19:38 <[florian]> yes, you basically need your own struct gpio_chip, most likely with a pointer to the base address register of the GPIO controller Sep 28 14:19:41 <[florian]> that should do it Sep 28 14:19:49 but I want to be able to follow the current documentation, and use the regualr gpio interfaces for some other work, so yeah, I copy the ar7 to make my own gpiochip Sep 28 14:19:53 yep, ok, Sep 28 14:20:06 I was looking at it, thinking I really would need that and getting rather confused :) Sep 28 14:20:33 <[florian]> yep Sep 28 14:20:38 I should be able to take the existing code though, and just wrap it up as the methods on a gpiochip, rather any real new code though Sep 28 14:20:46 <[florian]> yes Sep 28 14:21:00 <[florian]> beware that there are something which are done automatically by the gpiolib core code Sep 28 14:21:10 <[florian]> like for instance, if you have several banks of gpios, each with a different base number Sep 28 14:21:18 <[florian]> when your callback to set direction/input/output is called Sep 28 14:21:28 <[florian]> the gpio argument is already substracted with the proper gpiochip base number Sep 28 14:21:53 yeah, all the gpiochip calls get a gpio number that is 0...N relative to that chip right? Sep 28 14:22:00 <[florian]> indeed Sep 28 14:22:17 do you happen to know why the ar231x code never made it into the linux kernel? Sep 28 14:22:31 and it's just all in this one jumbo patch in openwrt? Sep 28 14:23:10 <[florian]> yep Sep 28 14:23:22 <[florian]> the patch is big but it was meant to be submitted for upstream inclusion Sep 28 14:23:38 <[florian]> having gpiolib support is a preriquisite do upstream submission anyway so your work is useful Sep 28 14:23:47 ok, goodo. Sep 28 14:24:07 <[florian]> in both cases, ar231x and ar5312 seem to both have less than 32 gpios Sep 28 14:24:13 <[florian]> so you will be fine with the same gpiochip register Sep 28 14:24:23 <[florian]> one for ar5315 and one for ar231x Sep 28 14:24:34 yeah, that's the way the current code does it Sep 28 14:24:43 they made their own gpiochip type struct Sep 28 14:24:47 with one each, Sep 28 14:25:55 <[florian]> the board code already seem to use gpio_* calls, so the conversion should be seamless Sep 28 14:26:27 yeah, the asm/mach-ar231x/gpio.h statically defined functions with the same names as gpiolib Sep 28 14:26:46 which is why it all failed to compile when I enabled gpiolib yesterday Sep 28 14:26:54 <[florian]> makes sense Sep 28 14:27:12 I think I've got enough answers now to go and hack a little while, see what happens. Sep 28 14:31:24 <[florian]> don't hesitate if you have any question Sep 28 14:36:54 so then this single line: platform_device_register_simple("GPIODEV", Sep 28 14:37:05 that's registering for the old gpio_dev interface to exist right? Sep 28 14:37:16 and it can stay (for now) because they coexist happily Sep 28 14:43:21 <[florian]> indeed Sep 28 15:00:20 allmoo Sep 28 16:35:03 in ar7, there's int __init ar7_gpio_init. What calls that to make sure the registration happens? is the __init magical by itself? or do I have to add a call somewhere else forit? Sep 28 16:36:49 ok, found that for ar7, prom_init calls gpio_init, I guess it's the magical one? Sep 28 16:42:47 why is libpam marked as @BROKEN ? https://dev.openwrt.org/browser/packages/libs/libpam Sep 28 17:48:39 whee! it works Sep 28 17:49:05 I had to comment out the button detection for now, but yay, sysfs for gpios! Sep 28 17:51:36 hey groovy, I can even change the active_low flag from teh commandline Sep 28 17:52:58 My package's upstream SVN repo SSL certificate is expired. Is it possible to set an option in the OpenWrt Makefile to ingore SSL errors? Sep 28 17:58:50 can you just change the https url to a http one? Sep 28 18:13:21 crow: because it once didn't work -> https://dev.openwrt.org/changeset/10044 - sounds like it should now be usable (but probably should depend on shadow passwords or something) Sep 28 18:22:15 [florian]: ar231x has some code in arch/ for setting up a timer for a gpio button for reset. Neither ar7 nor bcm63xx seem to have anything similar. Sep 28 18:22:45 the existing reset stuff works by having this shared exported gpiodev struct, which I've replaced with the gpio_chip gpiolib interface Sep 28 18:23:05 should this reset button be setup somewhere else altogether? Sep 28 18:23:22 KanjiMonster i will enable it later and try to compile/use it. I already enabled shaddow password and enabled openssh (need to enable --with-pma to) Sep 28 18:23:45 or, should I just change the code to use the gpiolib functions themselves? Sep 28 19:18:59 <[florian]> karlp: yes, this should be replaced by gpio-buttons Sep 28 19:19:16 <[florian]> karlp: bcm63xx has such a feature Sep 28 19:48:35 nbd: ping Sep 28 20:02:30 build #80 of lantiq is complete: Success [build successful] Build details are at http://buildbot.openwrt.org:8010/builders/lantiq/builds/80 Sep 28 20:09:26 ^^ \o/ Sep 28 20:09:31 full build worked Sep 28 20:13:48 KanjiMonster are there any depends actualy for libpam... Sep 28 20:23:49 <[florian]> karlp: how is it going on? Sep 28 20:26:38 crow: dunno, never used/needed it Sep 28 20:42:23 KanjiMonster ok, well i tried to compile libpam with kernel but it fails :(, http://paste.pocoo.org/show/Tqu6UlOdLSxHu4rIFcTQ/ Sep 28 20:43:24 crow: sounds like it's actually still @BROKEN Sep 28 20:58:38 KanjiMonster seems so Sep 28 22:07:27 nbd * r28314 /trunk/package/kernel/modules/netdevices.mk: kernel: remove the x86 dependency from kmod-sky2 so that it can be used on other platforms as well Sep 28 22:33:33 [florian]: went home for the day, willhave a look at bcm63xx for gpio buttons tomorrow, poke a bit more. thanks for your pointers! Sep 28 23:00:51 build #78 of pxcab is complete: Failure [failed compile_10] Build details are at http://buildbot.openwrt.org:8010/builders/pxcab/builds/78 Sep 28 23:10:06 karlp: Nope. The repo is only available over SSL unfortunately. Sep 28 23:53:50 build #68 of avr32 is complete: Failure [failed compile_4] Build details are at http://buildbot.openwrt.org:8010/builders/avr32/builds/68 Sep 29 01:12:36 anyone heard of a 'signed' (im guessing encrypted) squashfs? **** ENDING LOGGING AT Thu Sep 29 02:59:56 2011