**** BEGIN LOGGING AT Thu Aug 19 02:59:57 2010 Aug 19 07:55:21 * swalker wonders who packaged dudders Aug 19 08:04:17 heh Aug 19 08:46:40 /j #ubuntu Aug 19 08:46:49 well damn :/ Aug 19 10:59:43 [florian], ping Aug 19 11:23:07 mb * r22712 /trunk/ (2 files in 2 dirs): Add Tahvo USB support Aug 19 11:23:49 mb * r22713 /packages/net/gpsd/Makefile: gpsd: Install all libraries Aug 19 11:26:17 <[florian]> Wipster: pong Aug 19 11:27:14 [florian], bit of a leap forward, in trunk I have both interfaces being detected properly and the switch driver registering :) Aug 19 11:27:45 <[florian]> Wipster: awesome !! Aug 19 11:27:46 [florian], no disabling of internal phy either Aug 19 11:27:53 <[florian]> Wipster: ok, that's even better Aug 19 11:28:42 <[florian]> Wipster: so you got swconfig working and so on? Aug 19 11:29:34 mb * r22714 /packages/Xorg/app/gpxview/ (6 files in 2 dirs): Add GPXView Geocaching application Aug 19 11:30:08 [florian], yeh swconfig detects it when I do help. I needed to add CONFIG_IP17XX_PHY=y and CONFIG_SWCONFIG=y manually despite setting them in kmod tho Aug 19 11:31:37 <[florian]> Wipster: ok, I see why Aug 19 11:31:55 <[florian]> Wipster: whe I added these as modules, my ethernet driver is also a module, this is the reason why they have the right loading order Aug 19 11:33:11 [florian], the standard mii registers in the switch are todo with phy4 between the switch and the ar7, so the link doesn't come up (MAC mode?) as it forces a duplex and speed. Aug 19 11:33:46 <[florian]> Wipster: since cpmac is built in the kernel, the swconfig and kmod-ip17xx should also be built-in Aug 19 11:34:24 <[florian]> Wipster: ok, that was to be exected I guess Aug 19 11:35:32 [florian], unfortunately I cant see anything in extended register description to say if there is an external link established Aug 19 11:35:59 mb * r22715 /packages/lang/python-sip/Makefile: python-sip: Support parallel build for both native and host package. Aug 19 11:38:08 [florian], but the stack is sending so it thinks a link is up somewhere, I think the other interface is linked.... to what I dunno Aug 19 11:39:33 <[florian]> Wipster: can you show me the dmesg, I'd love to see it :) Aug 19 11:43:26 [florian], http://pastebin.com/GLNzdNC3 Aug 19 11:44:27 <[florian]> Wipster: that looks great Aug 19 11:45:18 [florian], hmm getting link ok now, nice Aug 19 11:45:23 <[florian]> Wipster: do you have a patch that you could send to openwrt-devel ? Aug 19 11:47:36 [florian], not yet, bit embarrassing but I need to work out how to make patches properly hah Aug 19 11:55:19 <[florian]> Wipster: ok, do not be embarrassed Aug 19 11:55:37 <[florian]> Wipster: from my perspective, a good patch is "atomic" in a sense that you can revert it and not break the existing stuff Aug 19 11:56:07 <[florian]> Wipster: from the workflow perspective, it is usually convenient to create new patches in quilt directly in build_dir/linux-ar7/linux-2.6.32.16 and work from there Aug 19 11:56:26 <[florian]> Wipster: not forgetting to use make target/linux/refresh not to loose your newly added patches Aug 19 11:57:01 Wipster: http://wiki.openwrt.org/doc/devel/patches Aug 19 11:57:29 mb * r22716 /packages/Xorg/lib/qt4/ (5 files in 2 dirs): Aug 19 11:57:29 qt4: Add qws- prefix to QWS specific files. Aug 19 11:57:29 No functional change. This just moves the files. Aug 19 11:59:08 <[florian]> jow_laptop: thanks, I did not know that page either :) Aug 19 11:59:51 thanks :) Aug 19 12:09:56 me neither, I only knew the oldwiki one *bookmark* Aug 19 12:13:20 still trying to figure out a strategy to fade out oldwiki Aug 19 12:50:11 nbd * r22717 /trunk/package/kernel/modules/fs.mk: kernel: add a missing config symbol for ext4 Aug 19 12:50:20 nbd * r22718 /trunk/target/linux/generic/patches-2.6.31/942-ssb_add_dma_dev.patch: kernel: backport ssb dma_dev changes to fix b43 build Aug 19 12:50:24 nbd * r22719 /trunk/target/linux/generic/patches-2.6.31/965-arm_restore_sigmask_v2.patch: kernel: backport ppoll to 2.6.31 to fix udev build Aug 19 12:50:29 nbd * r22720 /trunk/ (8 files in 2 dirs): speed up the build system by including include/shell.sh on shell commands only where necessary Aug 19 12:50:33 nbd * r22721 /trunk/package/pwm-gpio-custom/src/pwm-gpio-custom.c: pwm-gpio-custom: fix compile on linux 2.6.31 Aug 19 12:50:38 nbd * r22722 /trunk/include/package-ipkg.mk: speed up the ipkg build process some more by reordering commands Aug 19 12:50:42 nbd * r22723 /trunk/ (12 files in 12 dirs): toolchain: fix the sysroot mess by getting rid of $(TOOLCHAIN_DIR)/usr and moving it back to $(TOOLCHAIN_DIR), this change makes the toolchain relocatable again, which should fix the SDK Aug 19 12:50:47 nbd * r22724 /trunk/package/kernel/modules/ (other.mk pcmcia.mk usb.mk): kernel: speed up building kernel packages by getting rid of unnecessary CompareKernelPatchVer calls Aug 19 13:05:14 nbd: I'm currently wondering if it would be possible to run scripts/ipkg parallelized. Currently it seems to run sequencial and it takes a _lot_ of time. It's most of the build time, if lots of packages are selected and a change to one package was made. I looked through the makefiles, but I was unable to find out how that mechanism really works and why it's serialized. To me it seems a task like creating packages is a good candidate for parallel Aug 19 13:06:25 And thanks btw for your recent build system improvements, too :) Aug 19 13:06:33 mb__: nope regarding ipkg Aug 19 13:06:57 its a crappy shell script, if multiple instances access the status db for updating it will get corrupted most likely Aug 19 13:07:34 jow_laptop: Ok, one can fix that. But which part of the makefiles force it to run serialized? I would like to run some experiments on that. Aug 19 13:07:43 I bet reimplementing ipkg.sh in some real language will bring more speed improvements than parallel execution Aug 19 13:07:50 I also have some patches for host-opkg Aug 19 13:07:53 Yeah, that also Aug 19 13:08:09 mb__: we had it parallelized at some point, but it was racy wrt. stuff it was doing in /tmp Aug 19 13:08:28 however with the current structure it's not really easy to parallelize Aug 19 13:08:36 because the package makefiles are called with -j1 Aug 19 13:09:04 nbd: Ok, so I guess the easier way to make it faster is to rewrite it in a sane language? Aug 19 13:09:11 yeah Aug 19 13:09:13 or get host opkg going Aug 19 13:09:24 it works fine on linux, bsd and osx Aug 19 13:09:30 but there are some host libtool issues Aug 19 13:09:35 jow_laptop: What language is that written in? Aug 19 13:09:37 C Aug 19 13:09:44 ah, k Aug 19 13:09:51 its the same used on the target Aug 19 13:10:48 jow_laptop: So are your patches online somewhere? Aug 19 13:10:55 mb__: if you want you can try that: https://dev.openwrt.org/changeset/22406/trunk/include reversed Aug 19 13:11:09 it basically worked but I removed it due to the host libtool issues with opkg Aug 19 13:11:39 oops there was more Aug 19 13:11:41 moment Aug 19 13:11:52 jow_laptop: Those were just compile issues? Or also runtime issues? Aug 19 13:12:12 compile issues Aug 19 13:12:43 https://dev.openwrt.org/changeset/22406 (should apply with patch -R) Aug 19 13:13:00 IMO that ipkg script is really annoying. :) Aug 19 13:13:22 jow_laptop: Cool. Thanks a lot! Aug 19 13:14:41 mb * r22725 /packages/Xorg/lib/qt4/ (4 files in 2 dirs): Aug 19 13:14:41 qt4: Make the WS configurable and add X11 files. Aug 19 13:14:41 X11 config is still marked broken, because the configure script bits are missing. Aug 19 13:14:41 This should not alter current QWS build in any way. Aug 19 13:15:04 <[florian]> Wipster: so, how is it going? Aug 19 13:15:45 [florian], just come in from carpark cricket at work :) this quilt thing is really cool wont be long Aug 19 13:18:32 nbd: Hm, I just pulled in your changes. I guess I need to rebuild the toolchain? Because I get "arm-openwrt-linux-uclibcgnueabi-gcc: not found" Aug 19 13:21:11 What's the correct make command to do this? Aug 19 13:21:37 I did make toolchain/clean, but afterwards it still fails to build the toolchain. Aug 19 13:23:50 http://pastebin.org/613856 Aug 19 13:23:54 * mb__ scratches head Aug 19 13:24:09 you should do make dirclean Aug 19 13:25:56 cool, thanks Aug 19 13:26:35 nbd: no the treee is failing on mpfr even on a fresh checkout Aug 19 13:26:41 oh Aug 19 13:26:47 http://pastebin.com/YzAF7916 Aug 19 13:26:48 details please Aug 19 13:26:51 ah Aug 19 13:27:50 (Hm, clicking through pastebin.org posts.... 99% of them are spam :D) Aug 19 13:28:03 OutBackDingo: is your scripts/config.guess and scripts/config.sub missing? Aug 19 13:28:56 nbd: For me the files are there. I guess it's a shell issue. I have dash installed, which doesn't expand {}, afaik Aug 19 13:29:03 ah Aug 19 13:29:26 and it also fails again for me after dirclean Aug 19 13:30:27 nbd: nope they are there Aug 19 13:30:46 cp: cannot stat `/usr/builder/devel/TEST/scripts/config.{guess,sub}': No such file or directory Aug 19 13:31:05 -rwxr-xr-x 1 dingo dingo 44973 2010-08-19 09:15 config.guess Aug 19 13:31:05 -rwxr-xr-x 1 dingo dingo 34898 2010-08-19 09:15 config.sub Aug 19 13:31:58 nbd: uhmmm `/ and sub}': dont match ?? Aug 19 13:33:40 nbd * r22726 /trunk/ (include/toplevel.mk rules.mk): fix compile errors by forcing bash to be used as a shell in all submake processes Aug 19 13:33:51 that commit should take care of this problem Aug 19 13:34:23 nbd: will test, thanks Aug 19 13:35:40 nbd: ok seems ok now Aug 19 13:36:00 yep Aug 19 13:36:47 [florian], ok worked that out, its ready Aug 19 13:49:03 wtf, I'm somehow hitting a shortcut for hibernation while typing Aug 19 13:50:53 [florian], well baring anything I might have done wrong it should be sent Aug 19 13:51:51 <[florian]> Wipster: allright, I will check my mail then Aug 19 13:52:41 <[florian]> Wipster: that's incredibly trivial Aug 19 13:53:02 [florian], yep just about the right level for me to find :D Aug 19 13:53:12 <[florian]> Wipster: I think that this phy_mask is completely wrong actually Aug 19 13:53:24 <[florian]> Wipster: looks like some remnants of the days where phylib needed it Aug 19 13:54:16 [florian], yeh I didn't really know what it was trying to do how how it devised there was a external switch but there ya go Aug 19 13:57:30 <[florian]> Wipster: I will give it a try later in the weekend on some devices Aug 19 13:59:53 [florian], cool Aug 19 14:00:23 <[florian]> Wipster: which switch model do you have? Aug 19 14:00:38 [florian], IP175A LF Aug 19 14:02:35 <[florian]> Wipster: exactly what I have, so I assume this should work Aug 19 14:09:44 [florian], strange thing I did notice when I unmasked the mdio before registering I saw get_phy_device in net/phy/mdio_bus.c return a phy for every id it tested.. only attached the first and the last though. Aug 19 14:10:30 <[florian]> Wipster: this is indeed a bit strange Aug 19 14:34:58 <[florian]> Wipster: actually I think I will commit this patch so that other people can also test it Aug 19 14:37:02 [florian], ooo exciting :) hopefully https://dev.openwrt.org/ticket/3124 wont get reopened heh Aug 19 14:37:28 <[florian]> Wipster: and #55 if I recall right can finally be closed :) Aug 19 14:57:11 florian * r22727 /trunk/target/linux/ar7/ (config-2.6.32 patches-2.6.32/970-remove_fixed_phy.patch): [ar7] remove fixed phy support, enable most ar7 switch drivers. Aug 19 14:57:13 florian * r22728 /trunk/target/linux/ar7/Makefile: [ar7] add swconfig to the default set of packages Aug 19 15:44:03 nice Aug 19 15:58:38 nbd * r22729 /packages/mail/xmail/Makefile: xmail: add $(LIBGCC_S) to fix ARM EABI compile errors Aug 19 15:58:41 nbd * r22730 /packages/lang/python/files/python-package.mk: python: fix compile error Aug 19 15:58:43 nbd * r22731 /packages/lang/python-sip/Makefile: python-sip: add missing dependency Aug 19 17:47:04 nbd * r22732 /trunk/include/autotools.mk: fix libtool fixup for libstdc++ Aug 19 18:21:39 re: ipkg.sh, why was it that we replaced the python version with shell anyway? Aug 19 18:25:32 Bartman007: maybe becuz pythins so fat Aug 19 18:27:08 OutBackDingo: one of these days I'll teach you by committing a python md5sum implementation to fix some continued issues with non-gnu systems. Aug 19 18:27:22 it runs just as fast as the C-based gnu md5sum. Aug 19 18:27:37 :P Aug 19 18:27:45 Bartman007: i wasnt referring to speed wise i was more referring to size Aug 19 18:28:04 adding python requires like 500k ipkg Aug 19 18:28:40 OutBackDingo: that makes even less sense, ipkg.sh is used on the host, we use the compiled opkg binary on the targets Aug 19 18:29:32 oh, but maybe that is why... back in the WR days we did have the minimal ipkg-sh replacement for WR micro. thanks for reminding me. Aug 19 18:29:54 Bartman007: :P Aug 19 18:30:30 id be a happy puppy if a usable python would fit in say 200k :) Aug 19 19:09:08 claudio * r22733 /trunk/target/linux/at91/ (Makefile config-default): [at91] Include the usb-ohci module as default Aug 19 20:46:51 Hauke1: what's the status of the wgt634u fix? Aug 19 21:00:17 gah Aug 19 21:00:26 effing nickserv Aug 19 21:31:11 hi there Aug 19 21:31:26 I'm trying to flash a d-link router via JTAG, Imade awiggler adapter, it was working yesterday, I was able to detect the CPU and everything and now I have absolutely no response, I tested every sinlge pin of my adapter, I'm desperate :( Aug 19 21:33:37 wigglers are seldom reliable. atleast for me Aug 19 21:33:50 same computer on the parport? Aug 19 21:33:56 yes Aug 19 21:34:17 I have a digital scope, what does a normal JTAG detection look like ? I see activity on TCLK and TDI but nothing on TDO or TRST or SRST is it normal ? Aug 19 21:34:35 jtag is basically 'a long serial shift register' Aug 19 21:34:52 I've just read the wikipedia article ^^ Aug 19 21:35:14 so what you push into tdi, comes out ot tdo a fixed amount of clocks later. the number of clocks can deviate between different soc/cpus Aug 19 21:35:47 I guess on the d link router there will only be the CPU on the JTAG chain ? Aug 19 21:35:53 grapsus if you are interrested in that: there were some good talks on the llast 2 congresses of ccc.. there should be recordings/slides online Aug 19 21:36:11 grapsus if you are lucky, yes. Aug 19 21:36:12 roh: what is ccc, do youhave a link ? Aug 19 21:36:56 grapsus let me see.. ccc is the chaos computer club. a >20year old hacker community who does a public event once a year. Aug 19 21:37:44 http://events.ccc.de/congress/2009/Fahrplan/events/3670.en.html http://events.ccc.de/congress/2009/Fahrplan/events/3490.en.html Aug 19 21:38:22 thanks Aug 19 21:38:35 http://mirror.fem-net.de/CCC/26C3/mp4/ also has a video of the first talk. Aug 19 21:38:36 the fact that it was working yesterday drives me nuts Aug 19 21:38:48 felix does a very good introduction, compared to the dry literature. Aug 19 21:39:02 http://mirror.fem-net.de/CCC/26C3/mp4/26c3-3670-en-blackbox_jtag_reverse_engineering.mp4 Aug 19 21:39:33 ok I will download that one Aug 19 21:40:35 also check out his paper Aug 19 21:41:09 so even if you know nothing about the device after enough clock ticks you should find your data on TDO ? Aug 19 21:43:00 <_trine> roh thanks for that link there are some nice lectures there Aug 19 21:55:19 grapsus in principle... Aug 19 21:55:41 but there are differnt 'chains' in there, so it depends also on the data you push in there Aug 19 21:56:11 some are there all the time, some are device dependant and some may be there but not documented Aug 19 21:56:42 there should be a 'nop' Aug 19 21:56:47 and a boundary scan Aug 19 21:56:57 as different 'commands' or chains Aug 19 22:07:45 roh, those links 404 for me Aug 19 22:08:07 the fem-net.de ones Aug 19 22:11:14 <_trine> russell--, they work here Aug 19 22:11:39 <_trine> just watching one about the LHC at the moment Aug 19 22:28:32 _trine: working now Aug 19 22:28:59 <_trine> good Aug 20 01:28:23 cshore * r22734 /packages/net/freeswitch/files/freeswitch.init: [packages] net/freeswitch: Fixed remnants of UCI config that prevent FreeSWITCH startup with the UCI config not present (e.g. with -minimal or -upstream-defaults) **** ENDING LOGGING AT Fri Aug 20 02:59:56 2010