#qi-hardware IRC log for Sunday, 2011-06-05

qi-bot[commit] Maarten ter Huurne: Surface: Work around bug in SDL that loses per-surface alpha. http://qi-hw.com/p/gmenu2x/2394a0704:05
qi-bot[commit] Maarten ter Huurne: PNG: Use libpng directly instead of via SDL_image. http://qi-hw.com/p/gmenu2x/7aeb7a404:05
mthI finally got rid of SDL_image :)04:05
DocScrutinizerhttp://www.gizmoforyou.net/site/en/gizmocommunity/photos/photo.html?groupid=12&albumid=15#photoid=329  LOL MUHAHA04:18
DocScrutinizerwhere's my Aspirin?04:19
DocScrutinizeris it just me, or is this definitely illegible as a schematics?04:20
DocScrutinizerNB the latter is from the 1970s, still unchanged and available, except for the original set obviously been from a time where PNP was standard, GND=+, and NPN really exotic stuff (no kidding)04:27
wpwrakDocScrutinizer: seems that some EEs were begrudging us the obfuscated C contest ...04:29
DocScrutinizerGod, I *loved* this Braun Lectron stuff back when I was 1204:30
wpwrak"Lectron Gen-Regulation" ;-)04:35
larsc"EHEC-Zuechten leicht gemacht"?04:40
DocScrutinizerhttp://www.lectron.de/images/stories/pdfs/lectron_1010_bausteinkatalog.pdf p9_2803 :-)  also p6_2478(.1)   ///  p8_2612/13/14 p7_2487  ;-P04:54
DocScrutinizerwpwrak: how about the gizzle stuff? X-P04:56
DocScrutinizerwpwrak: 2933 Leerbaustein: 10stueck 6.5005:02
Fusingood morning Qiots05:19
wolfspraulmorning :-)05:20
Action: Fusin installed yesterday Haiku on an ol eeePC70105:21
DocScrutinizermorning wolfspraul 05:21
Fusinhi Doc05:22
DocScrutinizerhi Fusin 05:22
Fusinof couse too: hi wolf ;)05:22
Action: DocScrutinizer kicks his eeePC70105:22
Fusinfunny little box05:23
DocScrutinizerbattery lasted like 6 months05:23
Fusinthe right size for takin' with you, when you make longer sessions @ toilet :->05:23
DocScrutinizeronce suspended to ram and forgotten, it was dead05:23
DocScrutinizerspecial inferior Taiwan 701 version05:24
wpwrakup to 6 months battery life doesn't sound too bad. many laptops don't even manage 6 hours ;-)05:24
DocScrutinizerhaha, I knew somebody would say this05:25
larsctook me some self-control to not say it ;)05:25
kyakwejp: there is no option to select 0.9.32 uClibc within the backfire, it was not supported. You need to build trunk, where 0.9.32 is enabled by default. btw, the current image is based on backfire and uses uClibc.. 05:40
DocScrutinizerof course the battery is sealed (ultrasonic welded) - no way to open this thing without cutting off 2 fingers of your hand05:42
FusinDocScrutinizer: that left 3 for use, isn't that enough?05:44
DocScrutinizerwhat if my estimation been a bit off05:44
Fusinno risk, no fun :P05:45
DocScrutinizerhmmm cell dia ~18mm, length 65mm05:45
DocScrutinizer4 cells05:46
DocScrutinizerrating 7.4V 4400mAh05:46
DocScrutinizerLiIon Pack05:47
DocScrutinizer>> BY AIRMAIL << :-o09:36
whitequarkhow something like that can happen?09:42
wejpas the current firmware image seems to be incompatible with earlier ones, i have been trying to rebuild an up-to-date toolchain from scratch. but this fails unfortunately. when building the minimal gcc, configure stops with "checking for C compiler default output file name... configure: error: C compiler cannot create executables" any ideas what's going on there? the wiki is of no help :/11:26
whitequarkpost the relevant error from config.log11:27
wejphow do i find the relevant error from that file?11:27
whitequarkfind the line with your error message (C compiler cannot create executables)11:28
wejpat first sight the config.log looks okay to me11:28
wejpthere is no such line in the config.log11:28
whitequarkthere is a lot of configure scripts in gcc source11:28
wejplet me check if there are other config.log files in the directory11:29
whitequarktry "find . -name config.log -exec grep 'C compiler cannot' -ln {} \;"11:29
wejpyeah, doing just that right now11:30
wejpok, found it11:31
wejpthis should be it http://pastebin.com/cjsVD8d611:32
wejplooks like the assembler is doing something wrong there11:33
whitequarksomething is seriously screwed up in your gcc11:33
wejpthat is strange, it works fine in all other cases11:33
wejpah, found the problem. it is not my gcc but the configure script that is screwed up. it fails when the working directory is in the PATH11:38
wejplooks like it is compiling okay now. thanks for pointing me in the right direction11:40
kyakwejp: the current image IS compatible with previous ones, see my comment here soem hours ago13:06
kyakhowever, this doesn't have anything to do with the error you got..13:08
wejpi know, the error is unrelated to that, but it looks like it is not compatible. i don't know when something changed though. it does not necessarily have to be the last image, but could very well be an earlier one13:10
wejpi skipped a few updates13:10
kyakthere was only one release based on trunk (using 0.9.32), long time ago.. You must be very outdated anyway13:12
kyakthat broke binary compatibility13:12
kyakbut then images were based on backfire again13:12
wejpmy old toolchain was based on, so that wasn't it obviously. i thought it could have been an uclibc update, because similar behavior (application just terminating unexpectedly) sometimes happend with such upgrades13:13
qi-bot[commit] kyak: lingot: not only guitar tuner using GTK http://qi-hw.com/p/openwrt-packages/a1b9d1113:49
mthkyak: would you be interested in reviewing Ayla's work on gmenu2x in the install_locations branch?15:52
mthwe want to merge it to master, but it would be good if someone checks whether it breaks stuff on the NanoNote first15:52
kyakmth: sure, i'll have a look (it could take some time though)16:56
mthmaybe Ayla can explain what the changes do, to speed up reviewing?17:03
kyakthis would be helpful, of course17:08
kyaki remember he mentioned once the changes were related to read-only rootfs of some other device, so user setting for gmenu2x need to live in user's home dir17:09
mthyes, the rootfs for OpenDingux is read-only17:09
mthah, there he is17:10
mthin general unix apps shouldn't write in their install dir, but with a read-only rootfs it's not a should but a must17:11
kyakof course17:11
kyakmoreover, these apps shouldn't depend on their configuration files being in CWD17:12
kyakcd /usr/share/gmenu2x && ./gmenu2x is plain awful17:12
mthwell, Ayla fixed those things and tested it on OpenDingux, where it works17:13
mthbut we'd like to check that it won't break on the NanoNote before merging it to master17:13
kyakhow should i do? merge remotes/origin/install_locations to master and see how it builds/works on NanoNote?17:14
kyakor just work on remotes/origin/install_locations?17:14
kyak(i.e. checkout this branch only)17:14
mthI think testing the branch should be sufficient17:15
mthbut Ayla recently merged the master changes into install_locations, so merging back is unlikely to cause serious conflicts17:15
kyakah ok17:15
Aylashouldn't be a problem, yes17:16
Aylayou guys are using a packaging system for the nanonote right?17:16
AylaI believe that you could create a package for gmenu2x17:16
kyakthis is want i will do, a gmenu2x packages based on install_locations17:17
mththere was also a new configure flag added, I don't remember the details but probably Ayla does17:17
kyakso that would be --enable-platform=nanonote17:18
Aylathat is required as the data to install in /usr/share for the nanonote is not the same for the dingoo17:18
Aylafor instance, the key configs are different17:19
kyaki see17:19
kyakit has built fine17:28
kyakthis is the list of installed files: http://dpaste.com/550791/17:29
Aylait looks ok17:30
kyaknow let's try to get rid of that awful scripts..17:31
kyakand this is the build log here: http://downloads.qi-hardware.com/people/kyak/tmp/gmenu2x.log17:33
Ayladoes it work? :p17:37
kyakwell, it works if i start it as cd /usr/share/gmenu2x && ./gmenu2x :) I copied /usr/share/gmenu2x/gmenu2x as /usr/bin/gmenu2x.bin and try to start it like /usr/bin/gmenu2x.bin, and i get black screen17:40
Ayla/usr/share/gmenu2x/gmenu2x shouldn't exist17:41
kyaki figured that we still need that shell script for sourcing /etc/profile, because other apps (when started from gmenu2x) might require some env variables17:41
Aylathe binary should be /usr/bin/gmenu2x17:41
kyakAyla: it exists because it's isntalled currently like that17:41
kyak$(INSTALL_BIN) $(PKG_INSTALL_DIR)/usr/bin/gmenu2x $(1)/usr/share/gmenu2x/17:42
Aylaah ok, it's the script17:42
kyakso it's ok so far17:42
kyakno-no, it's the binary17:42
kyakthe script is installed as /usr/bin/gmenu2x17:42
kyakthis is the Install section of Makefile, i haven't modified it..17:43
Aylasorry, my computer use to reboot from time to time17:49
kyakAyla: hm... your last commit to origin/install_locations is from Thu Apr 14 19:35:50 2011.. Is it right?17:56
kyakah, now i see, heh17:58
kyaki didn't do checkout correctly :)17:58
Aylathe last commit should be from the 1st june18:00
kyakthat's right18:00
kyakAyla: ok, it started now18:06
kyakbut there are no icons18:06
kyakexcept for "Explorer"18:06
kyakand a new "emulators" section18:07
Aylayou have the settings, wallpapers icons?18:07
kyaki can change the wallpapers, too18:08
Aylawell, on dingoo we don't have any global application that should be on gmenu2x from the start, so my guess is that it does search only on ~/.gmenu2x18:09
kyaki have /root/.gmenu2x directory18:10
Aylabut then that's a bug18:10
kyakthat has these empty sections18:10
kyakok, we have a number of applications that should be in gmenu2x by default18:11
kyakso it should search in /usr/share/gmenu2x/sections/, too18:11
kyaki just did "cp -r /usr/share/gmenu2x/* .gmenu2x/" and it works18:12
Aylaaccording to your install log, some sections are installed to /usr/share18:12
Aylagmenu2x should be able to list them as well18:12
kyakseems that gmenu2x doesn't see those sections under /usr/share/gmenu2x/sections/18:13
kyaki see a bunch of 18:17
kyakopen("/usr/share/gmenu2x/skins/Default/icons/.png", O_RDONLY|O_LARGEFILE) = -1 ENOENT (No such file or directory)18:17
kyakopen("/usr/share/gmenu2x/skins/Default/icons//usr/bin/hnb.png", O_RDONLY|O_LARGEFILE) = -1 ENOENT (No such file or directory)18:17
kyaki tries to open some files in /usr/share, but doesn't use correct naming for them18:17
kyak*it tries18:17
Aylait searches for skin files in /usr/share, yes18:21
Aylaas well as on ~/.gmenu2x18:21
kyakAyla: i don't see that gmenu2x searches for sections somewhere else other than GMenu2X::getHome()18:40
Aylait does not18:44
Aylabut it should18:44
kyakAyla: did you do it intentionally that data/platform/nanonote/ contains only scripts and sections?19:00
Aylawhat should it contain?19:00
kyaki noticed it because the data/skins/320x240/Default is different from the currently used19:00
kyaki wonder how it got lost...19:02
Aylathe skin directory is the same that is on the repo19:02
kyakyeah.. i'm looking at history, looks like someone accidentally commited another default.png (older one)19:04
kyakanyway, skins are not meant to be platform-specific?19:04
mthskins should be resolution specific rather than platform specific, I think19:07
mtha platform implies a resolution, but multiple platforms can share the same resolution19:07
mthfor example NanoNote and Dingoo are both 320x24019:07
kyakok, i see19:09
Aylait prevents duplication19:10
mthif we'd want to avoid the Qi-branded wallpapers on OpenDingux and vice versa, maybe we should have both shared and non-shared wallpapers19:11
mthor we just ship everything and let the user sort it out19:11
mthI don't mind if someone wants to use a Qi-branded wallpaper on the Dingoo for some reason19:12
mthin any case, the majority of the wallpapers is not platform specific19:12
AylaI think you should be able to put platform-specific skins by putting them on data/platform/nanonote/skins/Default/19:12
mthah, that may already work indeed19:13
kyakyeah, that sounds good19:13
kyakwhat's been done is already a great improvement. Btw, i'd like to mention an old bug: change the skin to something other than default, save, go to skins again and change back to Default, save. Restart gmenu2x, you'll see blank screen19:14
Aylaeven on the new branch?19:15
kyakyep, even there19:17
kyaki gave the link to a log where i tried to dig a little bit..19:17
mthcompletely blank screen or black wallpaper?19:18
mthI changed the image loading code completely in the last week19:18
mthnot all of that has been merged yet to install_locations, I think19:18
mthbut it's present in master19:19
kyakmth: it's completely black19:19
kyaknot just the wallpapre19:19
mthI can't reproduce it on today's master19:20
mth"save" is just exiting the skins menu or is there more to it?19:21
kyakthat's sad for me :)19:21
kyaki press "s" and it returns back to main menu19:21
mththat depends... can you still reproduce it on today's master?19:22
kyakand the new skin is applied (i think so)19:22
mththe new skin is applied here19:22
kyakyeah, i've just tried it with install_locations19:22
mththe Dingoo doesn't have an "s" button though19:22
mthno keyboard19:22
kyakso, after you applied the new skin, can i go back to skins and change to the first skin?19:22
mthI used the start button to exit the skin dialog19:22
mthand then I started "read", exited that and gmenu2x restarted just fine with the default skin19:23
kyakare we using the same "Default" skin?19:23
kyaki.e. with "messageBoxBg=#00000080"19:24
mthinstall_locations is *almost* up to date with master, but not exactly and the changes made after the last merge are related to image loading and SDL surfaces19:24
kyaki'll try again with the laster master tomorrow19:25
mthah, my data is not from master, only my executable is19:25
qi-botThe build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-06042011-2219/19:25
kyakmth: this could be related to that "messageBoxBg=#00000080"19:25
mththat's RGBA?19:28
wpwraker ... read "what" instead of "that" ;-)19:29
kyakmth: hmm.. when i apply that "2010-12-14" skin, it always goes back to "Default" after restart19:29
kyaki'll try to test more tomorrow.. have to go now19:29
mthI'll test a bit as well19:30
mthsee you19:30
mthwpwrak: yeah, what I wanted to know is whether it was semi-transparent black or invisible blue ;)19:30
wpwrakmth: just set it to 0x00ff00ff and you'll see immediately ;-)19:35
wpwraknow, back to those most peculiar correlations i see in my "boundary scan". drive a pin to 1 -> reads as 1, drive it to 0 -> reads back as 1 too. grmbl...19:44
wpwrak(and the pin is nicely open. so it's all in the registers that somewhere get messed up.)19:45
mththe GPIOs can be pulled up or down depending on how they're configured19:46
mthand depending on how strong the outside world pulls, either the outside or the SoC wins19:46
wpwrakyes yes. i'm testing all four states. Z, R, H, and L. i also see the darn thing end up in all of them. just not the way the registers say it should.19:47
mthwe had a misconfigured pin that worked fine on one Dingoo model and caused problems on a clone19:47
wpwrakit basically seems that *somewhere*, data and direction get swapped. (it's on the AVR, which uses only those two for H/L/R/Z) but this doesn't make sense. i even read back the whole mess, and it looks okay. except that it isn't :-(19:48
mthah, it's not on the JZ19:49
wpwrakit's the atusb boundary scan. i send a pattern to the avr, it configures its gpios accordingly, then reports back all the registers involved. everything looks right, except the *somehow* direction and data end up being swapped. just how and where ? grmbl.19:50
wpwrakoh. i think i found it :) indeed, the only place where it's possible for a single bug to affect both setting and verification :)19:52
whitequarkwpwrak: can you take a look at a circuit? one fellow said he has shorted a certain pin to GND and now it does not charge. For me, it looks like the LED backlight should have died instead :)19:56
wpwrakwhat circuit ?19:57
whitequarkwpwrak: this one: http://files.whitequark.org/t/Vogue_01_sch.pdf19:59
whitequarkthe relevant parts are on page 3, the ones with VD1, VD2, VD319:59
whitequarkhe shorted C17 to GND, and there was smoke19:59
wpwrakif there's smoke, perhaps he how has an excuse to buy the last issue of playboy ? ;-)20:01
whitequarkthat's not equally easy for everyone in Russia. some live quite far from the places you can buy playboy in20:02
whitequarkanyway, I'm really curious why it could not charge now.20:03
whitequarkthat thing with VD2 (for everybody: 2R2 and a diode in series) is the ultimate chinese solution for charging Li-Ion batteries20:04
wpwrakyeah, i was looking for the charger. then it dawned on me that this contraption must be it ...20:05
wpwrakif the overcurrent blew 2R2 or VD2, that would explain the problem20:07
wpwrakyou would only see a bcklight failure if L2 or VD3 had died20:07
wpwrakah was the system on battery or USB power when it happened ?20:08
wpwrak(gpio scan) phew. just one misfit left. and that's the UART controller claiming control of the GPIOs20:10
whitequarkwpwrak: I'll ask him. I think it was on batteries20:10
wpwrakalso, how does he tell that the battery doesn't charge ? did he measure the battery voltage or did it just "not work" ?20:11
whitequarkthe firmware of player reads the voltage from ADC, and now it constantly says it's too low20:12
whitequarkand then eventually it shuts off20:12
wpwrakso it could also be VT1 that's broken20:13
wpwrakor is that BAT_CTRL ? searching ...20:14
wpwrakah yes. then it's directly the battery20:15
wpwrakyeah, my guess would be R9, VD2, or the battery itself20:15
wpwrakdid he notice where the smoke came from ?20:16
whitequarkwpwrak: when I've asked him whether the device was on USB or battery power, he answered twice, and on completely different question :/20:22
whitequarknow he says that on USB it works, and on battery just shows white screen20:23
whitequarkyou're probably right20:23
whitequarkwow, he posted that message (but in different words) third time20:25
wpwrakthe smoke must have been traumatizing ;-)20:27
Action: whitequark waits for the fourth20:28
wpwraki wonder if the unexpected attention from the "wrong" clientele kinda killed that idea of making "intelligent ads"20:44
wpwrakhmm, i have a table where pin values can be - among others - "L" (that would be input low with a pull-up), "l" (input low without pull-up), and "1" (output low). l and 1 are hard to distinguish. any suggestions for an letter/symbol for either ?21:46
wpwrak(other characters in the table are "H", "h", "Z", "z", "1", "x", and ".")21:47
wpwrakand =, :, !, / are reserved. as are all shell meta-characters21:47
rjeffrieswpwrak consider using zero 0 as symbol to represent input low w/o pull=up23:03
wpwrakah, zero is taken as well. forgot to mention that one. "0" is output low.23:04
wpwrakthis is what a typical table looks like: http://pastebin.ca/207527323:08
wpwraknow, depending on your browser font, you may see the ones in there or not ;-)23:09
wpwrakbah, no other suggestions ? how lame. maybe i'll pick "i" as an alias for "l" then ...23:50
DocScrutinizero=lower-0 = l23:58
DocScrutinizeror L *o* w, compared to *L* o w23:59
--- Mon Jun 6 201100:00

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