#qi-hardware IRC log for Sunday, 2011-01-09

kristianpaulmorning wolfspraul :-)04:54
kyakit was a very interesting discussion yesterday about openwrt06:08
qi-bot[commit] David Kühling: gnuplot-gfx: reworked truecolor fixes to upstream-submittable http://qi-hw.com/p/openwrt-packages/52ae1d706:09
qi-bot[commit] David Kühling: gnuplot-gfx: update to latest upstream version 4.4.2 http://qi-hw.com/p/openwrt-packages/c20e4cd06:09
kyakactually, wolfspraul expressed many of my own concerns06:10
wolfspraulhah, let's wait until the netsplit is over :-)06:10
kyakyeah :)06:10
kyakxMff: you were asking about the list of apps that require iconv/intl - should I just grep apps that DEPEND on iconv/intl from qi's openwrt-packages or you need something else?06:15
kyakif you wanted to have a look at those apps, you could just add qi's openwrt feed and see it yourself06:16
bartbesmore interestingly, *your* failing packages06:16
bartbesthe efl feed has 'enlightenment'06:16
bartbesit's upstream, and doesn't compile06:16
bartbesI'm guessing there are more of those06:17
kyakoh, ok06:17
bartbeskyak: that was at xMff, I have no idea what he asked for06:18
kyaki think you got it right06:18
kyakanyway, i got stuck at fbterm compilation06:18
kyakwith "tiny" libiconv06:18
kyakmaybe something else was affected, i don't know, because then i wanted "full" libiconv back06:19
kyaknow i'll try to build against tiny libiconv and see what else fails06:19
kyakand i have to believe xMff that the tiny libiconv is a real replacement for libiconv, and if an app builds fine, there will be no problems during runtime06:22
kyakbut to tell the tryuth, i'm not sure about that06:22
kyakspeaking about communicatino issues, there is stil no reply to this ticker here https://dev.openwrt.org/ticket/860306:23
lekernelwolfspraul: (mentor graphics) what free tools? :)06:23
kyaki guess i should write to openwrt mailing list?06:24
lekernelor you mean, free of charge?06:24
wolfspraullekernel: the fedora/free electronic lab06:24
lekernelthere isn't any free equivalent for leonardo spectrum, which is the most useful (and pirated) mentor product06:25
wolfspraulkyak: what other openwrt tickets are open right now?06:26
kyak8413 from xiangfu06:27
kyak8594 about glib2 not able to build against libiconv-full06:28
kyak8282 - i also reported that (not very important, but still there was no feedback)06:29
kyakand this ticket 8289 - is a perfect example of how things can go well06:30
kyakreport - check - fix :)06:30
kyakquick fix of that helped a lot while porting netsurf06:31
qi-bot[commit] David Kühling: use ccache also for C++ compilation http://qi-hw.com/p/openwrt-xburst/fba4b7106:31
kyaki think that's all from me, xiangfu might have other opened tickets06:31
wolfspraulok we have 4 open ones right now - 8603, 8413, 8594, 828206:34
kyakyes; btw, 8413 is what has been bugging bartbes for a while now06:35
kyakbartbes: am i right?06:37
bartbesI get a link error06:37
wolfspraulkyak: can you build bootable/working images now? I was a little surprised that you and xiangfu said you could not build a working image since mid-December06:37
wolfspraulmy one and only focus is how we get to the next image, with improvements and all the stuff contributed in recent weeks, and without regressions06:37
kyakyes, we can now that we've overriden libiconv, gettext and glib206:37
wolfspraulok I see06:37
bartbesbut then again, I patched it to link against libiconv-full06:37
kyakwhat's the error?06:37
bartbesmissing iconv_open, iconv_close and iconv symbols06:38
kyakoh, pretty strange.. are your libiconv-full and gettext-full different from the ones in openwrt-packfges?06:38
kyakbecause as i mentioned yesterday, libiconv-full and gettext-full are basically renamed to libiconv and gettext and now present in openwrt-packages (overriding the ones from openwrt feeds)06:39
bartbesI heard it was some header issue probably06:40
bartbesoh, I know06:41
bartbesthat might have been an error I did06:41
bartbessince gettext creates no actual lib06:41
kyaki just wonder, why you have this error06:41
bartbesnothing can link against this06:41
kyakhm, it created libintl06:41
bartbesoh, I probably messed up, I'll try again later06:41
bartbesit.. did?06:41
kyakhave a look into Makefile06:41
kyakyou should link against libintl06:42
bartbesbecause I'm pretty sure the makefile doesn't install anything but headers and autoconf data06:42
kyakas libintl is provided bby gettext package06:42
kyak$(CP) $(PKG_INSTALL_DIR)/usr/lib/libintl.{a,so*} $(1)/usr/lib/gettext-full/lib/06:43
kyaklooks like a real lib to me :)06:43
bartbesalso, as far as I can tell nothing is compiled eithr06:43
bartbesbut that's gettext-full06:43
kyakpaths are messed up, however06:43
bartbesnot gettext06:43
kyakyes, so what?06:43
bartbesand yes, gettext-full fails all paths06:43
bartbeshence nothing links to it without heavy, heavy patching06:44
bartbesalso, did it link against (stripped) libiconv?!06:44
kyakoh, so you are spending your time trying to link against stub lib?06:44
bartbeslibiconv-full also uses insane paths06:44
bartbesand I'm not sure gettext-full was patched06:45
kyaknot patched06:45
bartbesin any case, the paths are one huge pile of suck06:45
kyakyes :)06:45
kyaklike this one:06:45
bartbesso, did gettext-full compile?06:45
bartbesagainst stub libiconv?06:45
bartbesor did you have to patch manually?06:45
kyakagain, both *full versions are fixed (paths) and renamed to libiconv and gettext06:47
kyakthen all build fine06:47
kyak(except for glib2, which was changed upstream for the tiny libiconv)06:47
kyakbut it's overriden, too06:48
kyaki still have a feeling you don't completely understand me06:48
bartbesbut you manually fixed all that?06:53
bartbesor did somebody just revert all stubs?06:53
kyakthese packages have been added to openwrt-packages06:57
kyakthus effectively overriding the ones from openwrt feeds06:58
kyakyou have to change the order of feeds in feeds.conf for that06:58
kyakthis is temporary, until a switchable (between "tiny" and "full") solution is implemented07:01
kyaklike xMff mentioned, something like global $(LIBICONV)07:02
bartbesso if I want to switch I simply change the order in feeds.conf07:02
bartbesthen update and install07:03
bartbes(using scripts/feeds)?07:03
kyakyes, should be easy like that07:03
kyakchange the order; rm -rf feeds tmp; make package/symlinks07:03
kyakthen you will be using libiconv, gettext and glib2 from openwrt-packages07:04
bartbestime to reboot then07:04
kyakand these are full versions07:04
kyakhappy reboot then ;)07:05
bartbesput qipackages highest in the feeds.conf07:12
bartbesrm completed07:16
bartbesI just rememberd07:18
bartbesI killed my patched efl libs as well :)07:19
bartbesalright, time to reapply some patches then..07:27
bartbesI always wonder why the toolchain takes so long beforeit actually does something07:36
Action: bartbes tries compiling enlightenment07:37
bartbesoh I would be so happy if I got e17 running..07:37
kyakindeed would be cool07:37
kyakwhat is your build host?07:37
kyaki mean, is it powerful machine?07:37
bartbesnot really07:37
bartbesstill singlecore :(07:38
bartbesat least I have hyperthreading07:38
kyakit takes here a while, too.. but i don't feel like i have to wait too long07:39
bartbeshmm I guess I have to manually recompile libiconv and gettext (aka libintl)07:41
bartbesor not..07:41
bartbesnvm, my fault07:42
bartbesI tried enabling ccache, well, it helps if it's installed07:42
bartbesoh wow07:51
bartbesthe first bit of gettext is configure script after configure script07:51
bartbesyou see it configure for 10 mins, then 1 compile, then another configure (10 mins is a huge overstatement, don't worry)07:51
bartbeswell well well, that is one hell of a compile07:55
kyaksomething has changed in tiny libinconv07:58
kyakoh.. i think i should build from scratch.. perhaps fbterm is taking the older lib..07:59
bartbesfinally gettext finished08:01
bartbesyeah I needed to recompile libiconv and gettext manually as well08:02
kyakit wasn't really necessary to make it manually, i think08:03
bartbesmaybe not, but I did it anyway08:05
bartbesso now the efl is building again08:06
bartbeskyak: hmm08:32
bartbesI'll pastebin08:32
bartbesI'm thinking it means I need to patch08:33
bartbesoops, wrong window08:33
bartbesit can't find libiconv.so08:35
bartbesso maybe it is in the makefile?08:36
kyakthat's maybe because InstallDev didn't take place08:37
bartbesit did, it's there08:37
kyakcan you show more complete log?08:37
kyakis it from efl?08:37
bartbesfrom e1708:38
kyaki think it's the problem08:39
bartbesthat'd be in the package makefile08:40
bartbesI'll see if disabling that works08:40
kyakshould be something like TARGET_LDFLAGS+= -Wl,-rpath-link=$(STAGING_DIR)/usr/lib08:40
bartbesI just commented out the line and started a recompile08:42
bartbesif it fails I'll try that08:42
bartbesalright, added your LDFLAGS line08:47
qi-bot[commit] David Kühling: gnuplot-gfx: cleanup patches for 4.4.2 version http://qi-hw.com/p/openwrt-packages/234363908:47
qi-bot[commit] David Kühling: plplot: adapt font size scaling to smaller nanonote display http://qi-hw.com/p/openwrt-packages/a05abc908:47
bartbesalright, that's a lot better, now it fails on some configure options08:49
bartbesor not..08:49
bartbesanyway, something else now08:49
kyakthe round is won :)08:49
bartbesI need to disable some X11 stuff08:50
wpwrakzrafa: funny .. all the libs (ecore, edje, efreet, ...), but no e17 in jlime ?08:56
bartbeswpwrak: in owrt it's called 'enlightenment' maybe same goes for jlime?08:59
wpwrakbartbes: i searched for e17, enlightenment, Enlightenment ... nada08:59
wpwrakbartbes: let see if i have more luck with qpkg :)09:00
bartbese16? :P09:00
wpwrakthat, too09:01
kyakwpwrak: e-wm, perhaps09:03
bartbesof course a single "e" isn't much of a search term :P09:04
wpwrakthere's some e-wm-* stuff, no e-wm per se, though09:06
bartbesI'll have to go investigate on how to disable the X modules09:06
wpwrakhmm, i think the qpkg parser needs a bit more work. packages "${IPKG_VARIANT}" and "(<1.0." :)09:07
kyakwpwrak: maybe you can poke in here http://cgit.openembedded.org/cgit.cgi/openembedded/tree/recipes/e17 and find what could be it09:07
bartbesapparently the wm needs X09:09
wpwrakkyak: ah, thanks. seems that e-wm could be the right name. but there's no e-wm package. well, i'll just wait for zrafa's comment :)09:10
wpwrakbartbes: if you didn't have X, what windows would the window manager manage ?09:10
bartbesbut it could at least give you a desktop09:11
wpwrakwithout windows ? hmm ...09:11
wpwrakbartbes: zrafa's been doing a lot of work around e* lately, maybe he knows if what you have in mind is a perversion raster has already indulged in09:16
bartbesedje can be used to create a launcher09:19
wpwrakhmm. now why did i leave that 0R in the clock line atusd ? doesn't get any more useless ...09:19
bartbeshi xMff10:59
kyakxMff: http://pastebin.mandriva.com/21613 the build log of fbterm with tiny libiconv11:01
zrafawpwrak: I just checked.. Yes, it should be e-wm, and it is not there :P.. but.. I tried to build the packages with OE and it looks like it built the e-wm just fine :P11:01
kyakthere was this patch http://projects.qi-hardware.com/index.php/p/openwrt-packages/source/commit/2d7e398567856d8953b9f64576928573aae2c991/ for this issue, but it was dropped when we enabled uClibc with locale support11:01
xMffkyak: I'll take a look11:03
zrafawpwrak: because I think that jlime repo at qi is okey if we keep it stable (or frozen) I would prefer to upload these new packages on extra-packages. For opkg, jlime-pkg or qpkg would not have difference. So those packages would be there for installation11:03
zrafaSO jlime-pkg update; jlime-pkg install e_wm (or whatever) would work11:04
zrafawpwrak: what do you think?11:05
xMffkyak: they do really stupid stuff in their code :(11:07
kyakxMff: thanks! i'll try to stop what else packages might have problems.. most of all i'm afraid of runtime problems11:07
kyak*try to spot11:07
bartbesgreat, jlime boot failed11:08
bartbesit's never easy11:08
xMffkyak: basically (void *)-1 must be changed to (iconv_t)(-1)11:08
xMffkyak: in io.cpp11:08
kyakxMff: i agree, the code might be out of date and stupid.. but rewriting libiconb from scratch is not good, too :)11:08
zrafabartbes: why it is never easy?11:08
wpwrakzrafa: yeah, as long as it works, that's good :)11:08
kyakxMff: thanks for the hint11:08
xMffkyak: point is that it probably fail with other non-gnu iconv implementations too. iconv_t is not guaranteed to be void * (or a pointer at all)11:09
zrafawpwrak: yeah, that will work out of the box.. NO idea if e17 will work, we need to test :)11:09
wpwrakzrafa: i think we ultimately should move the jlime repo at qi-hw from "stable" (= frozen) status to "alive". but that's work for another day ;-)11:09
bartbeszrafa: I meant in general, not jlime11:10
kyakxMff: yeah, i think fbterm is build with gnu iconv in 99% of cases...11:10
bartbesin this case a reboot fixed it11:10
kyakxMff: thanks a lot for your explanations yesterday and patience, too11:10
zrafabartbes: ah :).. well, I remember that I failed to have the bootloader file.. and when I did the command for installation on NAND it did not install the bootloader.. so after that system did not boot (of course) but worse.. I was not able to have the usbboot again :P11:11
bartbesthere's always hardware usb boot11:12
zrafabartbes: yeah, but that was my first time trying that hardware usb boot.. so it took a lot of time I think11:13
wpwrakgah. i solder like a spider on dope. that death-by-caffeine breakfast probably wasn't such a good idea. hmm, perhaps a bit of beer will sooth my itchy hands ...11:17
kyakwpwrak: just for info, what's your time zone? you seem to be always active :)11:18
bartbeszrafa: so when can we expext an e-wm package?11:18
kyakwpwrak: or, you have that AI bot speaking instead of you, while you are sleeping :)11:19
wpwrakkyak: WTZ, werner's time zone :) the external stimulus (that day star) would be buenos aires time11:21
zrafabartbes: those built just fine, I should upload but no remember exactly how to do that :P11:21
zrafabartbes: checking my notes..11:22
zrafawpwrak: do you remember the wiki page link explaining details about how to upload stuff to downloads.qi.. etc?11:23
wpwrakwww-data@downloads.qi-hardware.com ?11:25
zrafabartbes: I have other nice things to upload to extra packages yet, like emacs as well which was cross compiled (with the final binary dump, no the huge thing which takes a lot to run), but lazy or no time to make the proper ipk these days :(11:26
zrafawpwrak: is downloads.qi-hardware.com?.. I guessed that I used another server name11:27
zrafawpwrak: let me try..11:27
zrafawpwrak: yes.. works :)11:29
zrafabartbes: almost uploading..11:30
Action: bartbes powers on ben11:37
zrafaE : well.. uploaded... And I tried to install. It installs, but you have to modify jlime-pkg. replace into /usr/bin/jlime-pkg the line  "opkg install *.ipk"  with  "opkg  --force-overwrite install *.ipk"11:57
zrafathen: jlime-pkg update ; jlime-pkg install e-wm11:57
zrafaAlso, we should install the recommended packages that installation suggests. Those packages are on repo11:58
zrafaexport DISPLAY=:0 ; enlightenment_start12:02
zrafaI guess that we need the recommended packages installed before.12:02
zrafawpwrak: bartbes: I modified /etc/muffinman-extras/startx, commented out matchbox lines.. and added enlightenment_start after Xfbdev line..12:42
zrafait runs..12:42
zrafanow I do not know what to do :D12:42
bartbesget proper keybinds? :P12:42
zrafa(I installed the recommended packages suggested before)12:42
zrafabartbes: yeah12:42
zrafabartbes: let me check if I find some documentation useful12:43
bartbeszrafa: I can't run update via jlime-pkg?12:50
zrafabartbes: which jlime version you have? the one from qi?12:50
bartbesone I saw announced in the mailing list..12:51
bartbesI guess it's the one you introduced jlime-pkg in..12:51
bartbeshow do I get the version?12:51
zrafaah.. okey. You have the version under OE+jlime development. The current with the e-wm packages is the uploaded to downloads.qi12:51
zrafabartbes: let me give you the link12:51
bartbesso.. what's the diff?12:52
wpwrakbartbes: can your jlime'd ben see the intertubes, google, youporn, etc. ?12:53
zrafabartbes: several things. Mainly you have the version which is being merged with OE. The version at qi servers does not have packages with patented technologies.12:53
bartbeswpwrak: of course12:53
wpwrakbartbes: just checking ;-)12:54
bartbeswpwrak: why?12:54
zrafabartbes: http://en.qi-hardware.com/wiki/Jlime12:54
zrafabartbes: that is the main link of the version with the e-wm packages I uploaded.12:54
bartbesdoes it have an sd rootfs?12:55
bartbesoh it does12:55
zrafabartbes: mainly there are several jlime version because I did that version at qi for this project. But also B_Lizzard is working hard on merge his jlime version work for nn into OE. So there are some diffs. But more or less it is the same :)12:55
zrafabartbes: yes, there is a version for SD12:55
bartbeswhat's the diff between the normal and the bootstrap version?12:56
zrafabartbes: bootstrap is a minimal system.. it boots and lets you on a shell.. YOu can play to convert that version on your dreamed version for installing packages12:57
zrafayou want and GUI you want, etc12:57
bartbeshmm so that would be ideal for e17 ;)12:57
bartbesI'll stick with stock for now12:57
zrafabartbes: maybe.. yes, but full version would be faster to run perhaps.12:57
zrafabartbes: but yes. bootstrap is nice to play :)12:58
bartbesbart@bartbes:/media/jlime$ sudo rm -rf *12:59
bartbesand then jlime was gone..12:59
bartbesunpacking the new one now13:00
bartbesopkg_prep_intercepts: mkdtemp: No space left on device13:17
bartbeszrafa: that went well (see above)13:18
wpwrakseems that the intercept was successful13:22
kristianpaulzrafa: boostrap is nice for sie14:34
kristianpaule low-level software (ogusb-lite.exe) provided with the board takes the samples and dumps them to a file as 8 bit values...inefficient but easy to work with.15:50
kristianpauleasy  i guess then..15:51
zrafakristianpaul: yeah, I guess that it should be a good start point for sie16:05
zrafakristianpaul: and you can add just you need16:05
zrafabartbes: no space left?.. poor jlime :)16:06
bartbespoor me16:06
zrafahaha :)16:06
zrafaI have a new 2gb uSD card.. old one broke16:07
bartbesmine is...16:07
kristianpaulliterally broke or just stop working?16:07
bartbesI just partitioned really well16:08
wpwrakroh, kristianpaul: for the "case", i was thinking of a U-shaped (in 3D) piece of silicone. slightly thicker but less wide than the board, so that one has to squeeze/tear it to make it fit, and it would stay in place mainly by friction16:31
rohwpwrak: sounds like some experiments are needed16:35
wpwrakroh, kristianpaul: this ought to be relatively simple to make, provided a two-component silicone preparation and a suitable release agent can be found. (the easily available acetone-based silicones used as bathroom/kitchen/etc. sealants may take a very long time to cure in this kind of mold, due to limited air flow.)16:35
wpwrakroh: aren't they always ? ;-)16:35
wpwrak"ten years of intensive research show that more research is needed" ;-)16:47
virichmm isn't there any bluetooth µsd device around?17:17
viricI imagine wpan is the only way to go :)17:18
wpwrak(BT) dunno. maybe, if there's wlan, there may be BT, too. but spectec themselves only have BT for mini-SDIO17:26
viricwhat is that 'io' after sd?17:33
viricanything relevant?17:33
wpwrakSDIO is an extension of the SD (memory card) protocol17:34
wpwrakadds new commands to access registers and also introduces an interrupt signaling mechanism17:34
qi-bot[commit] Werner Almesberger: atusd/Makefile: added targets "front", "back", "clean" http://qi-hw.com/p/ben-wpan/88b742b17:40
qi-bot[commit] Werner Almesberger: atusb/Makefile: removed claim that zones have to be filled in the .brd file http://qi-hw.com/p/ben-wpan/766b57517:40
qi-bot[commit] Werner Almesberger: atusd/Makefile: replaced board name by $(NAME) or pattern http://qi-hw.com/p/ben-wpan/cb2a30a17:40
qi-bot[commit] Werner Almesberger: atusd/Makefile: remove long-obsolete dependency on dtc123je.mod http://qi-hw.com/p/ben-wpan/ff6703f17:40
qi-bot[commit] Werner Almesberger: atrf-rssi.c: unmask IRQ_PLL_LOCK, for 231 compatibility http://qi-hw.com/p/ben-wpan/4d2395e17:40
qi-bot[commit] Werner Almesberger: tools/: renamed OpenWRT target from "ben" to "ben_openwrt" and added "ben_jlime" http://qi-hw.com/p/ben-wpan/45e55fe17:40
qi-bot[commit] Werner Almesberger: atusd/cam/mkmk: updated for making the 20110104 prototypes http://qi-hw.com/p/ben-wpan/e8d715017:40
qi-bot[commit] Werner Almesberger: atrf-txrx: updated for AT86RF231 interrupt handling http://qi-hw.com/p/ben-wpan/0c7146c17:40
qi-bot[commit] Werner Almesberger: tools/lib/misctxrx.c: decode AMI and CCA_ED_DONE interrupts (231 only) http://qi-hw.com/p/ben-wpan/1befc6917:40
viricThe SDIO-Linux project provides an open source SDIO-stack. The stack comes complete with drivers for the Host and Bus controllers, SD/MMC memory cards, SDIO-bluetooth and Atheros WLAN driver.17:44
viricso some SDIO-bluetooth card should work17:44
viricmaybe not µsd17:44
wpwrakdunno how compatible all the SDIO-BT cards are with respect to each other17:45
wpwrakthere's no difference in the protocol between SD, miniSD, and uSD17:45
qi-bot[commit] Werner Almesberger: atrf-txrx: small 231 continuous transmission mode cleanup (doesn't work yet) http://qi-hw.com/p/ben-wpan/9b498ae18:40
qi-bot[commit] Werner Almesberger: atusd: removed resistor of futility http://qi-hw.com/p/ben-wpan/8d4877318:40
wpwraktataa ! atusd bom cost reduced by 6 deci-cents apiece :-)18:40
zrafawpwrak: :D19:16
--- Mon Jan 10 201100:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!