IRC log of #maemo-ssu for Wednesday, 2012-08-01

*** MohammadAG has quit IRC00:07
*** xnt14 has quit IRC00:07
*** _rd has joined #maemo-ssu00:07
DocScrutinizer05there's "an app for that" (optification)00:35
DocScrutinizer05comes per default with each maemo500:35
DocScrutinizer05each semi-*recent* maemo500:35
DocScrutinizer05zeq: I just sumbled across osso-backup-restore-locale, nfc what it does00:38
zeq?00:39
zeqDocScrutinizer05: I'm trying to convince buildpackage to build the full locale cache00:40
zeqit does build a locale*.deb by default, but it contains very little.00:40
keriowould that fix the stupid thing where apt talks to me in english but it expects italian words at prompts?00:41
zeqkerio: that's possible, I've no idea what's contained in the binary blob locale pkg00:42
zeqas Doc mentioned earlier we might have some things missing.  It's entirely possible there will be other things currently which currently are missing appear.00:43
zeqFrom what I can gather recent debian libc builds have a locales-all target in addition to locale which by default just builds a package for the (build-)system current locale, or so it seems.00:45
keriozeq: would this end up in cssu-devel?00:46
zeqThere *must* be a way of cleaning all this up00:46
kerioor is it cssu-extras material?00:46
zeqkerio: don't know where it will end up.00:46
zeqI'd quite like to have locale packages to install support for different lanaguages rather all or crypt symbols!00:47
zeqcryptic*00:47
DocScrutinizer05kerio: maemo-optify-auto-opt.sh maemo-optify-firstboot.sh maemo-optify-make-mountbind.sh00:48
*** _rd has quit IRC00:50
DocScrutinizer05zeq: I juggled with all that shite some 2 years ago. My conclusion been it's not possible to do anything about it. IIRC the 'normal' locale dirs don't contain anything, it's all in that 23MB closed blob locale-cache00:51
DocScrutinizer05I tried to get rid of zc, ru, whatnot else00:51
DocScrutinizer05no way00:51
DocScrutinizer05there's a 23MB blob in your / which you need like 2% of its content but you can't do anything and cant do without it00:52
DocScrutinizer05ask timeless, he published his own improved translations - or was it crashanddie?00:53
keriois mmcblk0 faster than emmc?00:54
kerioer00:54
keriois mtd faster than mmcblk000:54
zeqDocScrutinizer05: I don't give up too easily ;)00:55
DocScrutinizer05kerio: allegedly yes - slightly00:58
keriowhy can't we install everything there?00:58
DocScrutinizer05because it's only 200MB?00:59
kerioer00:59
kerioi meant on the emmc00:59
kerioi can't believe there's no way to make a kernel that can use a rootfs on the emmc00:59
zeqOne good thing about mtd, in theory it should be possible to use XIP, no idea if it is used, or whether it's even possible in this case.  I've not looked into it for years.01:01
zeqXIP == eXecute In Place01:01
kerioooooh, that's neat01:01
zeqsame applies for data, it can be directly mapped on the storage01:02
zeqthis doesn't mean to say it's an overall win, or possible in this case though :)01:03
keriohmm, i wonder why the swap on mmc is used, given that there's plenty of space on the sd swap and it has a higher priority01:04
DocScrutinizer05hah?01:05
kerio*emmc01:05
kerioDocScrutinizer05: swap on mmcblk0p3, priority 0, used 16mb, swap on mmcblk1p2, priority 10, used 16mb01:06
DocScrutinizer05zeq: I'm not sure about XIP. I think even ROMBL loads mtd xloader/NOLO to ram01:07
DocScrutinizer05zeq: don't you mix that with NOR flash?01:08
zeqThere's no meaningful shortage of RAM on the device, and it's going to be a lot quicker than XIP, so that could be the reason.  XIP is a cool technology though.01:08
kerioi suppose the kernel is doing this because it's faster01:08
zeqDocScrutinizer05: I think you're right01:08
keriobut fucking hell, i would expect priorities to be respected01:08
zeqAs I said, it's been a while since I looked into it01:08
zeq5 years01:09
zeq+01:09
DocScrutinizer05NOR can XIP, NAND not though, afaik01:09
zeqSounds right01:09
kerioi wonder if compcache is actually worth it01:11
zeqkerio: probably depends on workload.01:12
DocScrutinizer05nothing can replace more ram, except even MORE ram01:13
kerioDocScrutinizer05: yeah but since you still haven't found a way to get more ram, we do what we can with what we have01:14
DocScrutinizer05though... since my laptop with 8GB RAM I actually think sometimes you can have *too much* RAM01:14
keriononsense, you can't have too much RAM01:15
DocScrutinizer05I *had* too much RAM... for that 32bit OS01:15
kerioDocScrutinizer05: what's your opinion on swap on uSD?01:16
keriotoo slow compared to emmc?01:16
DocScrutinizer05nasty01:16
DocScrutinizer05only for hardcore freaks01:16
kerioignore the backcover removal peril potential01:16
keriohm01:16
kerioi tried it a bit01:16
keriocompcache+swap on SD01:17
keriobut... i'm noticing slowdowns on fucking SOLITAIRE01:17
keriowait, why "only for hardcore freaks"?01:17
DocScrutinizer05you already mentioned one of the main reasons01:18
keriomaybe it's just compcache!01:18
* zeq is off to bed01:18
* DocScrutinizer05 meant backcover01:18
kerioenabling both would be like striping, and it would increase speed! and stuff!01:18
keriooh, i see01:18
keriowell, there's the Estel_ route of gluing a magnet on the sensor01:18
kerio:D01:18
zeqkerio: if you want to stripe you need to have both swaps at the same priority01:19
kerioi could try that01:19
kerioi wonder if it'll be faster than just emmc01:19
DocScrutinizer05well, I'd rather start messing with sys/class/gpio/* rather than doing that01:19
kerioconsidering that emmc is still a lot faster than uSD01:19
zeqanyway, goodnight guys01:20
kerionight01:20
DocScrutinizer05n801:20
Estel_kerio, mtd isn't faster than emmc, sadly01:20
Estel_due to compression01:20
Estel_it is technically, but compression kills it01:20
kerioyeah but ubifs is the shit01:20
Estel_freemangordon did extensive tests on that01:20
Estel_ubifs is ok, but nokia decided to enable transparent compression01:20
keriodid someone try a lighter fs?01:20
Estel_ubifs without comrpession would be OK01:21
keriohm, can it be disabled?01:21
Estel_well, nokia did it, because before Pr 1.2 or P1.1 whole system ended up i nrootfs01:21
Estel_they hoped for no need for optification01:21
kerioyeah right01:21
Estel_Pr 1.1 or 1.2 (can't remember) killed this ideas01:21
Estel_but compression remained01:21
Estel_well, considering that thumb makes 70-100 MB fre eon rotfs for average user...01:21
Estel_and that moving things to opt for increased performance actually makes sense for some things...01:22
Estel_disabling compression is an option to be considered01:22
kerioi don't actually know how ubifs works01:22
Estel_it wil lfor sure be investigated more after thumb01:22
Estel_it just works ;l001:22
DocScrutinizer05mkfs.ubi /dev/mtd401:22
Estel_BTw, as for patches01:22
keriono, i mean how to make a ubifs volume01:22
kerioDocScrutinizer05: i thought it had weird stuff wrt volumes01:22
kerioit makes a device in /dev/01:22
Estel_no idea if it will land on cssu-testing or cssu-devel or cssu-whatever, but for sure will land in cssu-thumb fork :)01:22
Estel_(zeq patches)01:22
keriocan the transparent compression be transparently disabled?01:23
Estel_never tried, no idea, but disabling it and restoring from backupmenu should work01:23
Estel_honestly, you could do everything from backupmenu01:23
Estel_i.e. creating filesystem ubifs without compression and extracting rootfs01:23
keriooh right, backupmenu does a rm -rf and then a copy01:23
Estel_but BEFORE doing so, you should optify what is worth to be optified01:23
kerioit doesn't remake the fs, for root01:23
kerioisn't stuff optified by default anyway?01:23
Estel_to be 100% sure that without compression files will fit into rootfs01:23
Estel_but01:24
kerioalso i only have 55mb of free space01:24
kerio:(01:24
DocScrutinizer05dafaq, mkfs.whateveryoulike /dev7mtd401:24
kerioalthough ubifs disk space is weird as shit01:24
keriobecause of the compression01:24
DocScrutinizer05dev/mtd4 even01:24
Estel_kerio, ubifs is for now only sensible filesystem to use wioth rootfs01:24
kerioDocScrutinizer05: explain /dev/ubi0, /dev/ubi0_0 and /dev/ubi_ctrl01:24
Estel_and few closed bits expect it there during boot01:24
Estel_so don't try with other filesystems01:24
kerioEstel_: do they expect it to be the transparent flavour?01:24
Estel_probably not.01:25
Estel_they shouldn't care01:25
kerio>you could do everything from backupmenu01:25
keriois backupmenu like zombo.com?01:25
Estel_well, 256 nAND mtd would be superfast for swapp, but would also live quite short01:25
Estel_kerio, you can enter root console in backuypmenu01:25
Estel_reformat mtd with ubifs without compression01:26
Estel_and then restore rootfs backup, one done before, AFTER optification01:26
Estel_BTw, check your rootfs, I have 70 MB free without any effort01:26
Estel_i.e. doing nothing special to have it this way01:26
DocScrutinizer05mtd4 mtd4 mtd4 mtd4 mtd4 mtd4 mtd4 mtd4 mtd4 mtd4 mtd4 mtd4 mtd4 mtd4 mtd4 mtd4 mtd4 mtd4 mtd4 mtd4 mtd4 mtd4 mtd401:26
Estel_DocScrutinizer, why (trendy word) trolling?01:27
kerioEstel_: used space?01:27
Estel_kerio, free space01:27
Estel_70 mb free01:27
keriono, i mean01:27
keriowhat's your used?01:27
DocScrutinizer05Estel_: seventeen *exponent* shouldn't%01:27
Estel_256-70?01:27
keriohaha01:27
keriosize 228M, used 169M, avail 55M01:28
keriosomehow i think it'll fit01:28
kerio:D01:28
Estel_should be 59 free01:28
Estel_but estimations01:28
Estel_some things like free are estimations in compressed01:28
kerioyeah01:28
DocScrutinizer05meh, lamers01:28
Estel_and I would be carefull with "should fit", as some files have awesome compression ratio in ubifs01:29
Estel_for example 10 MB file becomes 100 kb (and it's not /dev/zero derived ;) )01:29
Estel_kerio, easy experiment01:29
kerioEstel_: i think 169 are true 169MB01:29
Estel_take Your rootfs backup tar01:29
keriohold on, i'll DU root01:29
keriooh right01:29
Estel_and unpack it01:29
kerioi also have that01:29
Estel_on normal fs01:29
Estel_check space used01:29
Estel_job done01:29
* DocScrutinizer05 checks actual size of mtd4, checks what fs to format it with, and which speedtests to run on top of that fs01:30
Estel_speedtest kills nand quite quickly01:30
Estel_wouldn't stress it, if others did it already01:30
kerioDocScrutinizer05: apparently to properly use ubifs you need ubiattach and all sorts of shit01:30
Estel_kerio, google for ubifs, maybe there is a flag to just disable compression ;)01:31
Estel_but I doubt it01:31
Estel_but with thumb and speed advantage due to upgraded compiler, priority now is for compiling whatever possible and applicable with thumb01:31
DocScrutinizer05kerio: rtfm01:31
Estel_i.e. 3rd party programs, too, from extras and friends01:32
keriojust -x none in mkfs-ubifs01:32
keriomkfs.ubifs01:32
DocScrutinizer05Estel_: I kill and torture my mtd4 as long as I like01:32
Estel_kerio, might want to check actual flags used for current ubifs on N90001:32
kerioDocScrutinizer05: you monster01:32
DocScrutinizer05it's made for that01:32
Estel_and then, use same flags, EXCEPT compression01:33
DocScrutinizer05or do YOU use an initrd on your N900?01:33
Estel_to re-create it01:33
kerioi don't even know where they are01:33
Estel_mount01:33
Estel_;)01:33
Estel_mount as root01:33
keriooic01:33
keriorw,bulk_read,no_chk_data_crc01:34
*** Estel_ is now known as sfdsdgfsd01:34
*** sfdsdgfsd is now known as Estel_01:34
*** Estel_ is now known as asdasdas01:34
DocScrutinizer05mtd4 is *unused* empty space! (well h-e-n copies all your p0rn from your attached memsticks to it, but otherwise it's never touched)01:35
*** asdasdas is now known as Estel_01:35
kerioDocScrutinizer05: can't we use it?01:35
keriofor... idk, rootfs01:35
kerioit's like 200000 somethings01:36
DocScrutinizer05mkfs.ubi mtd401:36
kerioi meant for rootfs01:36
keriodoes mtd have a partition table or something01:36
DocScrutinizer05incredible how LOUD I have to shout until you start to notice I have something to say01:37
* kerio hugs DocScrutinizer05 and apologizes01:38
kerioi meant merging the partitions or something01:38
DocScrutinizer05flasher and NOLO have partitioning hardcoded01:38
keriosizes too? .s01:38
kerio:s01:38
kerioshame01:38
keriobut ok01:38
keriowhy the fuck is there a "rootfs.jffs2" in my COMBINED01:39
DocScrutinizer05ubi been jffs201:40
DocScrutinizer05for sentimental reasons the names stayed01:40
kerioi see01:40
DocScrutinizer05or rather, for "flasher got it hardcoded, as does NOLO" reasons01:41
kerioam i a bad enough dude to try this uncompressed rootfs thing?01:44
DocScrutinizer05I guess they used maemo4 and installed it to first n900 prototypes. then each crew started from that and wrote improved versions of their own little gem, so it would still work in m4 environment. One group made a better rootfs but called it rootfs.jffs2 so none of the other bits breaks. another crew made a NOLO that could also handle USB charging, but otherwise did everything like N8x0 NOLO01:45
DocScrutinizer05and that's where initrd/mtd4 comes from01:46
DocScrutinizer05though it *never* been used for maemo501:46
keriomeh, i'm not a bad enough dude to try it01:50
kerioalthough apparently there's a way to disable compression for single files01:50
kerioi wonder how flasher makes the FS01:51
Estel_it's still not sure if nolo cares for sizes, though01:52
Estel_flasher remakes them for sure01:52
kerioDocScrutinizer05: anyway, mkfs.ubifs /dev/mtd4 won't work at all01:53
Estel_but having working device and no need for flasdhing it via flasher - at least closed one - maybe you could alter size, just keep layout01:53
Estel_jsut a wild guess01:53
kerioubifs can be mkfsed on an ubi volume01:53
keriowhich is a layer between mtd and ubifs01:53
Estel_so create ubi volume first ;) but I would try with changing size of whole aprtition layout on NAND, just keep layout itself (adjusted sizes)01:53
kerioi don't think there even *is* a partition01:54
Estel_werll, don't expect that DocScrutinizer will hug and apologize You in return ;)01:54
kerio*partition layout01:54
Estel_it must be, as You have this mtd401:54
Estel_unused01:54
Estel_and our nice compressed ubifs used01:54
kerioyeah but it's something passed to the kernel at boot time01:54
Estel_doesn't matter, it doesn't reformat it on every boot. Just use linux trools for handling ubifs and mtd01:55
Estel_changte sizes01:55
Estel_recreate ubifs01:55
Estel_extract rootfs01:55
keriono, i mean01:55
kerioit's something that gets passed to the kernel on every boot01:55
Estel_s/trools/tools/01:55
Estel_lol01:55
kerioor it's compiled into the kernel01:55
kerioit's not something that you can change easily01:55
Estel_and there is a chance that it doesn't care about sizes :) I hope01:55
kerioooh01:58
kerioeven easier01:58
keriolike, a lot easier01:59
kerioIn UBIFS it is possible to enable or disable compression individually for each inode by setting or cleaning their compression flag. Note, the compression flag of directories is inherited, which means that when files and sub-directories are created, they inherit the compression flag of the parent directory.01:59
kerioi don't know if i'm a bad enough dude to disable compression on /02:04
kerioEstel_: wanna try?02:04
Estel_i'm dist-upgrading Ed now on device02:05
Estel_jsut unpack your rootfs basckup on normal filesystem02:05
Estel_and check it's sizes02:05
kerioyeah, it's probably faster02:05
Estel_if it will fit, disable compression02:06
kerio313M02:06
kerio:s02:06
Estel_if not, optifity something, until it will02:06
keriohow big is the nand?02:06
kerioer02:06
kerioyeah, the nand02:06
Estel_see biggest files, and optify them02:06
kerioaww, 256mb02:06
kerioshame02:06
Estel_228 in fact02:07
Estel_probably due to unused mtd402:07
keriohehe02:07
Estel_nand is 256 indeed02:07
kerionah, initfs is 0x200000 bytes02:07
keriojust 2mb02:07
Estel_200kb?02:07
keriooh, maybe02:07
keriono, 2mb02:07
keriosays /proc/mtd02:08
Estel_so something else sits there :) not unexpected, like bits for nolo?02:08
Estel_well02:08
kerioit's just wasted i think02:08
Estel_optify, until You will fit there :)02:08
Estel_in 228, not 2 mb02:08
kerionah, i'll pass02:08
Estel_kerio, Yuou use kp51r1, yep?02:08
kerioyeah02:08
Estel_when you use Pali's charging module, does the diode keep green one like during emergency charging, despite, in fact, charging normally?02:09
Estel_it annoys me, that it stays even with screen on02:09
Estel_and I have own pattern for charging ;)02:09
Estel_I wonder if this is not implemented yet, or a bug02:09
kerio\_o_/ i still use bme02:09
Estel_test is easy and safe02:09
Estel_as root02:09
kerioi like my battery indicator02:09
Estel_stop bme02:09
Estel_sure, but during hostmode You doj't use bme02:09
kerioof course not02:09
Estel_also, Pali's drivber, just like charge.sh by shadowJk, allow to charge my uber battery with higher current ;)02:10
keriohehe02:10
Estel_so, if You could02:10
Estel_stop bme02:10
keriooh shit it's 1am02:10
Estel_modprobe bq2415x_charger02:10
Estel_and connect charger02:10
Estel_then tell me about diode02:10
Estel_after finishing, rmod or modprobe -r bq2415x_charger02:11
Estel_and start bme02:11
kerioi'm gonna go to sleep actually :302:11
Estel_hm, can't do short test? Pity, but I won't stop You02:12
kerioEstel_: diode stays on constantly02:13
kerioprobably because it's the best the kernel can do, it can't really know your mce patterns02:14
Estel_it can02:14
Estel_it know during hostmode and booston :)02:14
keriowell of course it can, it's the fucking kernel02:14
Estel_I think it's jsut not implemented yet02:14
Estel_thanks a lot for confirming02:14
kerioEstel_: no, it's usbmode.sh that does the cool light shows02:14
kerioduring hostmode02:14
Estel_right02:14
Estel_no problem to have something that does light shows during charging, in userspace02:15
Estel_after all, booston is kernel situation, too02:15
Estel_i.e. kernel does it, and something take lightshow02:15
Estel_in stock bme does it02:15
Estel_in oour case <whatever> may do it too ;)02:15
Estel_and this dioe is eatin 0.5 mA of my charging current! :P:02:16
keriohaha02:16
kerioanyway, wait for pali's userland bme replacement02:16
keriohm, i just realized that there would've been a problem with decompressing rootfs by recreating the fs02:19
keriothe ubitools are on the rootfs02:19
*** jonwil has joined #maemo-ssu03:22
*** arcean has quit IRC03:55
*** nox- has quit IRC04:09
*** LaoLang_cool has joined #maemo-ssu04:33
*** jonwil has quit IRC04:58
*** amiconn has quit IRC05:31
*** amiconn_ has joined #maemo-ssu05:31
*** amiconn_ is now known as amiconn05:31
*** zeq1 has joined #maemo-ssu07:54
*** _rd has joined #maemo-ssu08:41
*** _rd has quit IRC08:50
*** LaoLang_cool has quit IRC09:14
*** jonwil has joined #maemo-ssu09:50
*** Pali has joined #maemo-ssu09:52
zeq1as long as the SDK stays as is, is there a good reason not to have a more recent glibc on device?10:22
*** Pali has quit IRC10:27
zeq1freemangordon: I've held back putting the cssu0 build up until I know what's happening with the linux-header dependencies.10:30
*** Pali has joined #maemo-ssu10:30
*** infobot has quit IRC10:32
*** luf has joined #maemo-ssu11:14
lufcssu version fixed in curl, obexd and eds in git.11:14
kerioyo freemangordon11:51
keriowtf is tinymail11:51
kerioand do i have to reboot after upgrading it?11:51
Palilibrary for modest11:54
kerioohai pali11:57
kerioPali: Estel_ was a bit confused by the values in the sysfs entries of your battery module12:00
keriowhat units are those? why 1800 for 950?12:03
kerioor something12:03
Palikerio, all values are in standard units (mAh, mV, mA, ...)12:05
kerioyeah but the charging current was 1800 for Estel_12:06
Palicat registers and check them in datasheet12:06
Palialso rmmod driver and use i2c get/set12:07
Paliand ask Doc about this problem12:07
PaliI ported charger shell script with info from datasheet12:07
Palicheck if same problem is i charger script12:08
kerioi see12:08
*** luf has left #maemo-ssu12:15
*** Pali has quit IRC12:45
*** macmaN has quit IRC13:08
*** macmaN has joined #maemo-ssu13:12
*** andre__ has quit IRC13:43
*** lizardo has joined #maemo-ssu13:54
freemangordonzeq1: ok13:58
zeq1freemangordon: I've confirmed again it is working fine with 4.2, trying with 4.6...13:59
freemangordon4.6?13:59
*** Pali has joined #maemo-ssu14:01
zeq1linaro-4.614:01
zeq1rebooting now14:02
*** zeq1 has quit IRC14:06
zeqfreemangordon: linaro-4.6 does exactly the same thing.  I14:15
freemangordonand it is?14:15
zeqI can't believe the generated locale format is compiler dependent, so it must be something not building correctly on later compilers.14:16
freemangordonwell, lets stick to 4.2.1 for libc then14:17
zeqThat why I asked ^ whether we *really* need to stick to this old version for anything other than the SDK14:17
freemangordonold version of? libc?14:17
zeqyeah14:18
zeqwhat stops us moving closer to Debian?14:18
freemangordonwell, i have no clue what will break if we do that14:18
freemangordonreally, having pselect working should be enough, even without newer compiler14:19
zeqit's a separate thing for sure14:19
freemangordon:nod:14:19
zeqThere must be security vulnerabilities in this version at least?14:20
freemangordonsure, but we don't know about them14:20
zeq(libc I mean)14:20
freemangordonbut it has nothing to do with the compiler14:20
freemangordons/but/and/14:21
freemangordon~ping14:21
freemangordonhmm14:21
zeqno, but being able to get Maemo closer to a debian distro will pick these things up for free.14:21
freemangordonbut we don;t know what will break. though libc should be ABI compatible.14:21
zeqthere *shouldn't* be breakage (in theory)14:22
freemangordonyeah14:22
zeqbut won't know without testing14:22
zeqIf I get my dev device I think I will make an experimental target and get as close as i can14:22
*** LaoLang_cool has joined #maemo-ssu14:22
freemangordonzeq: well, lets split that.14:22
freemangordon1. pselect with 4.2.114:23
freemangordon2. libc upgrade14:23
freemangordonagree?14:23
zeqabsolutely!14:23
freemangordondeal :)14:23
zeqdo we keep --without-fp in the glibc rebuild?14:24
zeqI think we should probably if only there must have been some reason to have it...?14:24
freemangordonhmm, did you find any rationale for that?14:24
zeqno14:24
zeqbut there must be one14:24
zeqsurely14:24
freemangordonglibc? or libc?14:24
zeqit's in the armel.mk of the Maemo glibc build -> libc614:25
freemangordonzeq: ^^^14:25
zeqit's actually eglibc afaik14:25
freemangordonzeq: though it might be there by historical reasons14:25
zeqcould be14:25
zeqsupport ARM CPUs without FP hw14:25
freemangordonbecause AFAIK on N8X0 vfp is veeery slow14:26
zeqor very slow FP indeed14:26
freemangordonnot sure how it is on A8, but should not be that bad14:26
zeqmight be worth benching with and without?14:26
freemangordonhmm, can we do some kind of benchmark?14:26
freemangordon:D14:26
freemangordonyeah, do you know any?14:27
zeqthere must be a libm benchmark....14:27
zeqI'll google14:27
freemangordonok14:27
*** infobot has joined #maemo-ssu14:30
*** ChanServ sets mode: +v infobot14:30
freemangordonthanks doc14:31
DocScrutinizer05yw14:31
zeqfreemangordon: nbench should be ok i think14:33
freemangordonok. will you do that?14:33
DocScrutinizer05~+uptime14:33
infobot- Uptime for purl -14:33
infobotNow: 3m 17s running infobot 1.5.4 (SVN) -- linux14:33
infobot1: 59d 8h 41m 19s running infobot 1.5.4 (SVN) -- linux, ended Sun Nov 14 18:39:57 201014:33
infobot2: 57d 3h 9m 23s running infobot 1.5.4 (SVN) -- linux, ended Fri Jun 26 20:39:27 200914:33
infobot3: 36d 20h 47m 14s running infobot 1.5.4 (SVN) -- linux, ended Tue Aug  4 17:38:59 200914:33
freemangordonzeq: hmm, stupid question14:33
DocScrutinizer05~+chaninfo14:33
infobot124/124 channels, 5033 users, 3618 unique: #debian/1182, #kde/353, #maemo/284, #gsoc/237, #meego/214, #asterisk/188, #harmattan/188, #oe/151, #wowuidev/138, #wowace/114, #webos-internals/111, #openmoko/95, #wowhead/88, #arm-netbook/86, #openmoko-cdevel/73, #uclibc/71, #n9/70, #utah/69, #bzflag/61, #edev/61, #htc-linux/59, #tomcat/56, #qi-hardware/54, #curseforge/51, #maemo-ssu/50, #asterisk-dev/45, #diablofans/43, #elinux/42, #flow3/41, ...14:33
zeqwhat is? :)14:33
freemangordonit is only you who can do it right now :D14:34
zeq:D14:34
freemangordonnbech with/out VFP14:34
zeqtrue14:34
freemangordonnbench14:34
*** mickname has quit IRC14:36
DocScrutinizer05whenever somebody notices infobot being unresponsive or even not joined, please ping me about it14:37
DocScrutinizer05*sometimes* I can do sth about it14:37
freemangordon:)14:37
DocScrutinizer05~+status14:39
infobotSince Wed Aug  1 11:30:17 2012, there have been 0 modifications, 2 questions, 0 dunnos, 0 morons and 3 commands.  I have been awake for 8m 43s this session, and currently reference 118668 factoids.  I'm using about 18484 kB of memory. With 0 active forks. Process time user/system 3.08/0.15 child 0/014:39
DocScrutinizer05thanks tim riker we got purl aka infobot aka ibot aka apt14:44
infobotDocScrutinizer05: de rien14:44
*** gri has quit IRC14:44
freemangordonWTF? I am running my device without OC for several hour without noticing it. hmmm, not sure if it is good or bad thing :)14:45
freemangordonhours even14:45
DocScrutinizer05definitely a good thing, it proves OC is placebo14:45
freemangordonno, it proves thumb2 is really good thing :P14:46
DocScrutinizer05hey, probably that too14:46
freemangordon;)14:46
DocScrutinizer05and then it demonstrates OC << thumb14:46
freemangordonbut of course14:46
freemangordonthat is for sure, no CPU can replace free RAM14:47
DocScrutinizer05I'm still amazed about that reported speed gain of thumb14:47
freemangordonDocScrutinizer05: you won;t believe it14:47
DocScrutinizer05I'm already pondering to experience it myself14:47
freemangordonbut I think it is a combination of less code and newr compiler14:47
DocScrutinizer05:nod:14:49
DocScrutinizer05since I just come to think of it: you usually compile bug proving code with -O014:49
freemangordonyeah14:50
freemangordonthough not necesserily14:50
DocScrutinizer05yeah, for bugs in compiler optimization you use whatever is needed ;-P14:51
* jonwil is still no closer to figuring out the mysteries of the universe (or rather, the mysteries of osso-wlan-security) :(14:51
DocScrutinizer05for demonstrating any other bugs elsewhere you either use -O0 or assembler ;-)14:52
DocScrutinizer05generally it's a good idea to look at assembler listing anyway14:53
DocScrutinizer05to make sure the binary does what you planned it shall do14:53
DocScrutinizer05for bug testing code14:53
freemangordon:nod:14:53
freemangordonPali: ping14:54
Palipong14:56
freemangordonPali: could you confirm tinymail bug fixed?14:56
freemangordonPali: in -thumb bersion14:57
zeqfreemangordon: just for fun (and because it was ready to go) here is linaro-4.6 compiled libc6 --with-fp: http://pastebin.com/mfATnev5  compare to sulu's nbench results for Easy Debian: http://talk.maemo.org/showthread.php?t=34550&page=26314:58
*** arcean has joined #maemo-ssu14:59
freemangordonPali: also, could you look at https://gitorious.org/community-ssu/tinymail/blobs/1955bad9fa612e7313ad753726e8cda68f14d0a8/libtinymail-camel/camel-lite/camel/providers/imap/camel-imap-store.c#line271714:59
freemangordonI still have a feeeling the code is buggy15:00
freemangordonzeq: ok15:00
zeqshould add, that's thumb too15:00
Palifreemangordon, fixed15:00
zeqmuch faster than the ED results15:00
DocScrutinizer05now who's going to prepare a generic "backup" containing all the "good stuff" *apps* we like to offer to our beloved customers?15:01
freemangordonzeq: yeah15:02
DocScrutinizer05idea: download the "backup" somewhere, and restore, picking which apps to restore/install and which not15:02
freemangordonPali: does not matter, please take a look at do-while cycle, i think if there is only one "item" to add, the list will contain 2 elements15:03
freemangordoni.e. item-> next will point to item15:03
freemangordonand that will cause a memory corruption. sorry, I am sick these days and cannot think :(15:04
freemangordonzeq: yeah, deffinitely better. I am curious how it will be with 4.2.115:06
zeqI'm rebuilding to make sure I have everything set correctly15:07
freemangordonok15:07
freemangordonzeq: could you look at the tinymail code above15:07
freemangordonseems Pali is too busy now15:08
zeqok15:08
zeqthe while loop?15:08
freemangordonyep15:09
freemangordonwhat will happen if we have only one element added to the list15:09
zeqI'm not sure what the data looks like...?15:11
freemangordondoes not matter15:11
freemangordonmy point is that if we add only one element (i.e. item), then its "next" member will point to itself15:12
freemangordoni.e. item->next==item15:12
zeqthat's true15:12
freemangordonthats a nasty bug, look at the code that frees the list15:13
zeqI take it, it's possible to have only one item be added15:14
zeq?15:14
freemangordonyes15:14
freemangordonhttps://gitorious.org/community-ssu/tinymail/blobs/1955bad9fa612e7313ad753726e8cda68f14d0a8/libtinymail-camel/camel-lite/camel/providers/imap/camel-imap-store.c#line262015:15
freemangordonzeq: ^^^15:15
freemangordonthe heap will be raped :D15:15
zeqouch15:15
zeqThis has hit people?15:16
freemangordonwell, heap corruption. who knows :D15:17
zeqlooks like it just needs an extra check somewhere15:17
freemangordontail->next = item; should be "else tail->next = item;"15:18
freemangordonaiui15:18
freemangordoncorrect?15:18
zeqyes, otherwise the tail=item is redundant anyway. must have been intended15:19
freemangordonthat comes from my last commit, it was even uglier before that15:20
zeqoh15:20
DocScrutinizer05(item-> next will point to item) classical single linked list, empty :-)15:20
freemangordonnot empty unfortunately, with one element pointing to iself15:21
freemangordon*itself15:21
DocScrutinizer05for a double linked list both item->next and item->prev point to item when empty15:21
freemangordonDocScrutinizer05: look at the code that frees the list, it is not done like that15:22
DocScrutinizer05each linked list consists of a root elem that's always 'virtual' and never holds real data15:22
freemangordon^^^15:22
DocScrutinizer05then the implementation is brainfsckd ;-D15:22
freemangordon:nod:15:22
freemangordonthis is how it was:15:23
freemangordonhttps://gitorious.org/community-ssu/tinymail/commit/1955bad9fa612e7313ad753726e8cda68f14d0a815:23
freemangordonsomehow that was working with gcc 4.2.115:23
freemangordonnewer compilers resulted in NULL being returned15:24
freemangordoninsted of pointer to first element15:24
zeqhmmm15:24
freemangordon?15:25
zeqitems = NULL;15:25
zeqtail = (struct imap_status_item *) &items;15:25
zeq?15:25
freemangordonyes :D15:25
zeqwtf15:25
freemangordonbrainfcked code, though I can understand what was the intention15:25
zeqmaybe it meant if items != NULL?15:26
freemangordonfirst time tail will hold the address of items15:26
freemangordonand it will be assigned the first element address15:26
freemangordonbut gcc 4.6.2 and 4.7.2 cannot deal with that15:27
zeqprobably because it's broken ;)15:27
freemangordontotally hackish and unreadable code15:28
freemangordonyes15:28
freemangordonanyway, goping to fix it15:28
freemangordon*going15:28
zeqok :)15:28
zeqI'll get back to rebuild libc615:28
zeqhandy thing I can just static link nbench to the SDK libc6 :)15:32
*** Pali has quit IRC15:33
kerioDocScrutinizer05: try thumb! it's great!15:34
* DocScrutinizer05 sucks his thumb15:34
kerioand then provide a service to upgrade the n900's ram!15:34
* kerio doesn't know if it's even theoretically possible15:34
DocScrutinizer05sure, I'll ship the .deb tomorrow15:34
DocScrutinizer05in theory we could add "ultrafast" RAM_as_uSD15:35
keriohow?15:35
keriodo they even exist?15:35
DocScrutinizer05adapter15:35
DocScrutinizer05anyway, telco waiting for me, cya l8r15:36
keriohttp://www.hwtools.net/ExtenderBoard/TFEX.html15:36
zeqbye doc15:36
keriolooks small15:37
zeqfreemangordon: can't we just say the non-sdk linux-kernel-*-headers "Provides: linux-kernel-headers" ?15:39
kerioEstel_: pls add a slot for DDR3 ram in your replacement aluminium cases15:40
zeqotherwise we need "... | ...power...| ...cssu..."15:40
zeqin the libc6 control15:41
freemangordonzeq: i think that is provided, lemme check15:42
freemangordonno, it is not :(15:43
zeqbug eh?15:43
freemangordonzeq: but what is the problem to build it with SDK headers?15:43
freemangordonwel,, not sdk, but maemo-extras15:44
zeqyou end up without any pselect() :P15:44
zeqat all!15:44
freemangordonwhy? is it something in headers?15:44
zeqyes, the headers hold the syscall tables15:44
freemangordoncan't recall what your patch is doing15:44
freemangordonaah, ok15:44
freemangordonhmm, it will take a while15:45
zeqfor some reason, if you tell glibc to use a later min kernel, it assumes pselect() et al is provided without checking and silently builds libc without it entirely if it isn't in the syscall list.15:46
freemangordonhehe15:46
zeqrather unfortunate :)15:46
freemangordonzeq: lemme finish with tinymail, we'll need Pali too15:46
zeqok15:46
amiconnfreemangordon, zeq: Afaik the "fp" in --without-fp means old arm hardfp, which no modern arm cpu supports. So this option makes sense. Using vfp otoh is a good idea (if not already in use)15:52
amiconnAfaik binaries compiled with hardfp do run on arm cores without hardfp support, but very slowly because hardfp will be emulated through exceptions (of course the emulation code needs to be present)15:54
freemangordonamiconn: ok, sounds sane15:55
amiconnOur arm cortex-a8 should support vfp, as it definitely does support neon, and iirc these two features are closely related15:57
freemangordonvfp is supported for sure15:57
freemangordonamiconn: why "old", there is some hand written assembly code?16:00
*** mase76 has joined #maemo-ssu16:01
amiconnI mean FPA, which is not binary compatible with VFP16:04
freemangordonok16:04
amiconnAnd what relation is there between "old" and "assembly"?16:04
zeqif it's FPA support mode, why isn't it called such ???16:04
zeqEABI dropped support for FPA afaik16:05
zeqhow can the switch do *anything*16:05
zeq?16:05
amiconnHmm, check the gcc --help output?16:06
zeqamiconn: it's a glibc build switch16:06
freemangordonamiconn: it was a question, i.e "is there some hand written assembly code?"16:06
amiconnCode for what?16:07
zeqfreemangordon: I suspect there is since I get exactly the same nbench result with ARM gcc-4.2 as with THUMB linaro-4.6 compiled libcs16:07
freemangordonzeq: what ./configure --help gives about that switch?16:07
zeqfreemangordon: about to check that ;)16:07
freemangordonamiconn: scratch that :)16:07
amiconnok16:08
zeq  --with-fp               if using floating-point hardware [default=yes]16:08
freemangordonnice. now, WTF is that disabled?16:09
freemangordonzeq: who is the maintainer?16:09
amiconnHmm, not very descriptive. What type of floating point hardware?16:09
zeq*any* probalby16:09
amiconnI guess a quick check would tell16:09
zeqprobably16:09
amiconn(compile using --with-fp and check with objdump)16:10
zeqamiconn: I'm in the process of compiling with-fp and without-fp for benchmarking purposes :)16:10
zeqit is possible that it's a win to use softfloat in libm, who knows?16:11
freemangordonzeq: but your device will be busted if libc does a SIGILL16:11
zeqfreemangordon: that's okay because I'm just installing it in the SDK and building nbench with -static :)16:11
freemangordonaah, ok :)16:11
zeqJuha Kallioinen <juha.kallioinen@nokia.com> <- maintainer16:15
zeqI guess that means unmaintained?16:15
*** lizardo has quit IRC16:16
freemangordonzeq: well, unmaintained, but we can ask him why16:17
freemangordonif he can recall16:17
freemangordonzeq: but on the other hand if it works we can skip that16:17
*** lizardo has joined #maemo-ssu16:18
zeqit definitely works, it's on my device ;)16:18
freemangordonzeq: built with no --without-fp?16:18
zeqyeah, it was an accident16:19
freemangordonhehe, a brave one :D16:19
zeqI backed up /lib before playing with all this16:19
zeqso I could always use LD_LIBRARY_PATH16:19
freemangordonthats in case you can boot16:20
freemangordonand I am not sure you can with broken libc616:20
zeqother than the messed up locales, it's easy enough to test before rebooting :)16:20
zeqthe locales only break after reboot16:20
zeqwhen I tested the other day, and thought it was working, I hadn't fully realised16:21
freemangordonzeq: hmm, isn't dpkg-configure all needed?16:21
freemangordonor something like that, to rebuild the locales16:21
zeq?16:22
freemangordonyeah, lemme check16:22
zeqok16:22
freemangordondpkg-reconfigure locales16:22
zeqinteresting16:23
freemangordonnot sure that works on maemo though16:23
freemangordonhmm, no dpkg-reconfigure16:23
zeqnope16:24
zeqanother good reason to get closer to debian16:24
freemangordonzeq: BTW how are the locales broken?16:25
zeqwell, if using (apparently) any gcc other than 4.2 on reboot the UI translations are missing16:25
zeqjust symbolic labels appear for all the UI elements16:26
zeqmakes some things very hard to use!16:27
*** luf has joined #maemo-ssu16:27
freemangordonmissing as in? there is nothing in /usr/whatever_the_locale_are or gettext fails?16:27
zeqremember my what's that button do question the other day!16:27
freemangordonyes16:27
zeqthe locales are provided by posix-locales pkg16:27
zeqin /usr/lib/locales16:28
zeqor is it /usr/lib/locale16:28
zeqanyway it's just a huge bin only pkg containing the generated locales16:28
zeq(from nokia)16:28
freemangordonlocale-archive?16:29
*** Pali has joined #maemo-ssu16:29
zeqyes16:29
freemangordona huge 20MB file?16:29
zeqon rootfs16:29
zeq:)16:30
kerio:s16:30
keriocan it be optified?16:30
freemangordonso, the file is there, but cannot be read?16:30
zeqfor some *unknown* reason16:30
kerioalso Pali: can kp51 mount from the emmc before requiring anything in /usr/?16:30
zeqI don't see an error, is there a log I can check?16:31
zeqobviously it's not going to be in dmesg16:31
freemangordonsyslog?16:31
Paliand ubifs is rootfs16:31
zeqsilly question, but where is syslog kept?16:31
freemangordon"/var/log/syslog"16:32
zeqhonestly?16:32
zeqhehe16:32
freemangordonbut you should install sysklogd16:32
zeqah16:32
zeqmaybe that's why I didn't know16:32
freemangordonand reboot after that, so everyone is aware there is syslog daemon16:32
zeqok, that explains it16:32
zeqI don't currently have a /var/log16:33
zeqinstalling now16:33
keriocan /usr/ be moved away from the rootfs?16:33
zeqWell, I'll give it another go later, see if there are any errors16:34
kerio/etc/mtab is linked /proc/mounts, and unmount tries to change /etc/mtab which causes warnings.16:34
keriofukken messybox again?16:34
kerio...wait, why does my sbin/preinit have something like a menu to choose between flash and mmc boot? :o16:35
keriooh, r&d mode16:36
freemangordonPali: I pushed another fix to tinymail.16:38
keriowhat the hell, the boot process requires /usr/sbin/chroot16:39
kerioi really can't follow what happens in preinit16:39
Palifreemangordon, ok16:39
zeqI'm thinking nbench isn't a good way to benchmark libm16:42
freemangordonwhy?16:42
zeqI'm getting the same results, within ~1%16:43
zeqeven less than 1% actually16:44
freemangordonhmm, strange, with or without FP16:45
freemangordon?16:45
zeqbetween every test16:45
freemangordonaaah16:45
zeqwith-fp without-fp and linaro-4.616:45
zeqdiff says the bins differ though16:45
zeqwhat's a good fp instruction to look for in ARM disassembly?16:46
zeqI can't even think what the register symbols are16:47
freemangordonwell, check sin/cos, whatever trigonometric function, it should be done with one instruction AFAIK16:47
zeqgood idea :)16:47
zeqI'll look in the debug lib16:48
freemangordonor even better fsin/fcos16:48
zeqvcvt.f64.f32    d7, s1616:50
zeq?16:50
zeqfp?16:50
freemangordonyep16:50
zeqperhaps without-fp is ignored?16:51
freemangordoncould be16:51
zeqshouldn't be there anyway though16:51
zeqand it must be producing different code ...16:52
zeqotherwise the binaries should have matched16:52
zeqactually the libm bins do match16:54
zeqfor some reason the static nbench bin didn't16:55
kerioPali: what modules are needed to access the emmc?17:00
Paliomap hsmmc17:01
Pali+ deps17:01
kerioare they big?17:01
PaliI do not know17:02
keriohm, omap_hsmmc.ko is 21k17:03
*** DocScrutinizer has quit IRC17:06
*** DocScrutinizer has joined #maemo-ssu17:06
*** DocScrutinizer05 has quit IRC17:06
*** DocScrutinizer05 has joined #maemo-ssu17:06
*** LinuxCode has joined #maemo-ssu17:09
*** LaoLang_cool has quit IRC17:13
*** LaoLang_cool has joined #maemo-ssu17:32
*** LaoLang_cool has left #maemo-ssu17:42
*** Estel_ has quit IRC18:28
*** Estel_ has joined #maemo-ssu18:37
*** Pali has quit IRC18:47
*** ivgalvez has joined #maemo-ssu18:54
*** Pali has joined #maemo-ssu18:58
*** NIN101 has joined #maemo-ssu19:01
*** jonwil has quit IRC19:13
*** Pali has quit IRC19:20
*** Pali has joined #maemo-ssu19:23
*** ivgalvez has quit IRC19:28
*** Pali has quit IRC19:37
*** Pali has joined #maemo-ssu19:38
*** mase76 has quit IRC20:05
*** mase76 has joined #maemo-ssu20:27
*** mase76 has quit IRC20:27
zeqI've (re-) found a really good reason to upgrade glibc.20:38
zeqC99 inline semantics20:38
zeqas discussed on tmo:20:38
zeqhttp://talk.maemo.org/archive/index.php/t-81861.html20:38
kerio<freemangordon> afaik THUMB is broken on the N900's SoC, and the workaround's overhead > benefit of THUMB20:51
kerioLOL20:51
kerioi guess that programmers really *are* awful at guessing where's the slowness20:53
*** luf has left #maemo-ssu20:53
zeqYou don't know until you try ;)20:57
*** Estel_ has quit IRC21:14
*** Estel_ has joined #maemo-ssu21:14
*** taziff has quit IRC21:49
*** taziff has joined #maemo-ssu21:49
DocScrutinizer51well, producing decent working code is craftmanship21:55
DocScrutinizer51producing decent working *readable* code is an art21:56
DocScrutinizer51producing decent working readable *optimized* code that results in an optimauzed fast system, this is the dark science of magic spells21:57
*** arcean_ has joined #maemo-ssu22:00
*** arcean has quit IRC22:03
*** xnt14 has joined #maemo-ssu22:11
*** xnt14 has quit IRC22:12
*** xnt14 has joined #maemo-ssu22:13
*** xnt14 has quit IRC22:17
*** xnt14 has joined #maemo-ssu22:17
*** MohammadAG has joined #maemo-ssu22:25
*** xnt14 has quit IRC22:25
*** xnt14 has joined #maemo-ssu22:26
*** xnt14 is now known as xnt14_22:29
*** xnt14_ is now known as xnt14i22:29
*** xnt14i is now known as xnt1422:29
*** xnt14 is now known as _xnt1422:30
*** Pali has quit IRC22:32
*** _xnt14 has joined #maemo-ssu22:39
*** arcean_ is now known as arcean22:41
amiconnIf it's both working and fast, why does it need to be readable? :P22:46
DocScrutinizer05that's one of the magic words nobody outside the magic circle knows the meaning of: maintainability22:55
DocScrutinizer05one of the foundation concepts of our black magic22:55
DocScrutinizer05it's about heraitage and honour of your ancestors22:56
DocScrutinizer05transfer of sekrit knowledge to the next genration, and stuff like that22:57
DocScrutinizer05so basically nobody outside the magic circle needs or wants to know about22:57
DocScrutinizer05but occasionally you witness one of our black magic circle dispute with a charlatan at the IRC market, and then you can hear the old wise man shout something like "man 3p kill" and the charlatan falls to the ground hiurt deadly while the guru just turns away and walks outa the scene - then you witnessed the powers of the true dark science of magic spells23:08
*** _rd has joined #maemo-ssu23:09
keriohmm, what's 3p?23:09
DocScrutinizer05pshhhhh that's sekrit!23:09
keriooh, posix 200123:10
kerioor something weird like that23:10
*** infobot has quit IRC23:15
amiconnBlack magicians are able to deal with code that looks unreadable to normal programmers ;)23:19
kerioDocScrutinizer05: anyway...23:22
keriowho fixes the bugs in the decent working readable optimized code that results in an optimauzed (sic!) fast system?23:22
*** taziff has quit IRC23:30
*** taziff has joined #maemo-ssu23:30
DocScrutinizer05kerio: obviously such code is bugfree per definitionem23:31
kerioooh, tinymail updates again23:32
kerioi don't even use modest23:33
DocScrutinizer05( optimauzed (sic!) ) demonstrate that you tought your cat a better typing style ;-P23:33
amiconnlol23:34
keriooptimauzed is yours23:34
kerioand "tought", really23:34
DocScrutinizer05curse KDE4 bullshit, do you think it's possible to switch the spellchecker to en_EN in konversation's input field permanently?23:36
DocScrutinizer05mauze is what my cat types when he suddenly feels hungry23:37
*** _rd has quit IRC23:38
keriooh, btw23:38
keriowho recompiles the decent working readable optimized code that results in an optimized system as thumb2?23:39
DocScrutinizer05then he runs out and comes back with some mize23:39
DocScrutinizer05I told him to avoid thoise friggin warez sitez23:40
DocScrutinizer05recompile is gentoo23:41
DocScrutinizer05gentoo even nears to other OS23:42
Estel_DocScrutinizer, Pali to9ld me to contact You, about strange behavior of sysfs entries in bq2415x_charger module (sic!)23:46
Estel_as method of gathering them is, reportedly, derived from your work23:46
Estel_mind excuse wood faeries for a while, and look at it?;)23:46
Estel_and before You ask, I was also surprised that you're one to help with module written by Pali (despite methods used)23:47
*** NIN101 has quit IRC23:51
*** lizardo has quit IRC23:56

Generated by irclog2html.py 2.15.1 by Marius Gedminas - find it at mg.pov.lt!