#qi-hardware IRC log for Wednesday, 2013-03-20

wpwrakwhee ! back online again !01:53
wolfspraulwpwrak: welcome back01:53
wolfspraulwhat happened?01:53
wpwrakPC disk started to fail. then i got a blackout (power). after than, the PC didn't want to come up anymore. when trying to nurse it back to life, another blackout struck. after that one, the disk didn't even get to the grub menu.01:55
wpwrakthen i tried to reinstall with one of my out-of-service disks. four disks later, still no luck.01:56
whitequarkwpwrak: use UPS?01:57
wpwraktoday, i finally went and bought a new one. this one seems to fine.01:57
wpwrakwhitequark: it beeped for about 30 seconds, then it shut down01:57
whitequarkwpwrak: man 1 battery-maintenance ;)01:59
wpwrakman pointless :) power failures here are either very short (mere brown-outs) or > 1 h. there's hardly ever one in between. so buying new batteries all the time would be an exercise in futility02:01
whitequarkmakes sense02:02
wpwrakthe quality of harddisks really seems to be going down the drain these days. this disk was only something like two years old. and its predecessor didn't live much longer either.02:03
whitequarkwpwrak: manufacturer?02:10
wpwrakWD and Samsung02:12
wpwraki miss Maxtor. those were immortal.02:13
wpwrakin my file server, i have two maxtor. must be about 10 years old by now. still working perfectly. then two WD. some of the first 1 TB disks. one is still okay. the other goes offline after transferring a few hundred MB.02:18
wpwrakah yes, there was some added fun: last week, i got an eye infection (conjunctivitis), which of course just about peaked around saturday/sunday. made crawling around the pc and swapping disks particularly enjoyable. murphy obviously believes in synergy :)02:23
kristianpaulwpwrak: wb :)02:37
wpwrakthankx :)02:40
hellekinwpwrak: you're cracking me up. Murphy's synergy :} Glad that your eyes got better03:04
wpwrakstill very sensitive. so reading text at daytime is painful. but yes, they're much better than a few days ago. may still be a week or so until they're back to normal, though.03:08
wolfspraulwpwrak: did you take antibiotic eyedrops?03:12
wolfspraulif it's as serious as you describe you most likely should03:13
wolfspraulthere are good eye cleaning solutions too, boric acid I think03:13
wolfspraulbut don't mix yourself :-)03:13
wolfspraulthat's just for cleaning though03:13
wpwrakthe ones they prescribed me are antibacterial but not antibiotic. seem to do the job, though. just takes a while.03:15
wpwrakfor washing the eyes, the doctors recommend ... tea :)03:15
wolfspraulok :-)03:16
whitequarkisn't boric acid poisonous?03:18
wpwrakfor the cleaning, it seems the worst of the muck gets stuck in places where it doesn't leave easily. and you can't pry it out yourself. yesterday, my doctor did a round of cleaning and removed something like 1 cm^3 of goo from one eye. thursday will be another round.03:18
hellekinwpwrak: you could sell it on eBay :)03:20
wpwrakdo they have a category for body fluids/slimes ? :)03:22
whitequark1 cm^3?!03:23
whitequarkwhere was it03:23
wpwrakinside the eye, above and below the eyeball. some mucus membranes, i think.03:28
whitequarkso... did he something like a blunt syringe tip into your eye? and then drain it?03:29
wpwrakthe eyeball is fine (besides being a bit irritated). a conjunctivitis is an infection of the tissue surrounding the eyeball.03:31
whitequarkI see03:40
mwcampbellCan the NAND flash memory on the Ben be addressed as if it were RAM? I'm asking about the capabilities of the hardware, not of Linux.15:18
whitequarkmwcampbell: nope15:22
Action: whitequark has only recently heard of XIP NAND15:23
mwcampbellXIP was what I was thinking of.15:25
mwcampbellI guess that has mostly been done with NOR flash.15:26
pcercueimwcampbell: no15:30
mwcampbellWhat is the shortest boot time that has been achieved on the Ben? From power-on to a shell15:31
mwcampbellor at least, from power-on to running some user-space program15:31
whitequarkmwcampbell: actually nevermind, I never heard of XIP NAND flash ever.15:32
whitequarkit was a NOR with an SPI interface, which made it somewhat more convenient.15:32
paul_boddieActually, when it comes to boot time, is it possible to augment the kernel so that the CPU just starts running it (the augmented bit of the larger payload, I suppose) at power on? Or does there need to be a smaller payload like U-Boot that then loads and jumps to the larger kernel payload?16:14
whitequarkpaul_boddie: well, you need a bootload16:22
whitequarkbecause nand can have bad pages16:22
whitequarkplus, the ROM primary bootloader only knows how to load a few leading NAND pages.16:23
paul_boddieWell, I was just thinking of the situation you have where things like U-Boot starting to accumulate Linux kernel code, and then you start to wonder whether the essentials of the bootloader couldn't just be incorporated into the kernel somehow. Or rather, the kernel payload.16:24
whitequarku-boot needs to set the environment up for kernel16:24
whitequarkwhen it starts to be portable across platforms, it quite naturally has an intersection with the kernel field of responsibility16:25
whitequarkgiven they're both in C16:25
paul_boddieSo instead of the CPU loading U-Boot, initialising the hardware, figuring out where the kernel is, loading the kernel, jumping into the kernel, and all the stuff that then tends to go wrong (the kernel doesn't have the support for hardware that U-Boot was quite happy about), why not just have a single payload?16:25
whitequarkpaul_boddie: for one, that would require compiling a distinct kernel for each device16:26
paul_boddieYes, but there is only one NanoNote. ;-)16:26
whitequarknot sure about MIPS, but with ARM, you can more or less build in the support for every single board out there.16:26
whitequark(now with device tree support)16:27
whitequarkMIPS stuff iirc has a similar mechanism16:27
whitequarkpaul_boddie: well, you're not the only user of the kernel16:27
paul_boddieIt's not like some random distro loading modules for all sorts of arcane, ancient datacentre crap just in case that modern CPU is connected to some weird networking interface made by some dinosaur organisation in 1993 for ten whole minutes.16:28
whitequarkpaul_boddie: u-boot is also useful in its arguable bloat.16:28
jekhorhello. In latest snapshots of openwrt-xburst for nanonote mpd seems broken (exits with message "Failed to initialize input plugin 'ffmpeg': No protocol"). Does someone know something about this?16:28
whitequarkpaul_boddie: I've reflashed dead kernel through tftp more than once. Without u-boot, I'd be fucked.16:29
paul_boddiejekhor: Maybe the MPEG exclusion in the configuration is broken?16:29
paul_boddiewhitequark: I agree that U-Boot is very useful, but I just wondered that since U-Boot is probably always expanding, why not just combine it with the kernel? Let them share their hardware expertise.16:30
whitequarkpaul_boddie: I guess because U-Boot's source is also of abysmal quality16:31
jekhorpaul_boddie, this is tail of 'strace mpd' output: http://paste.debian.net/243137/16:31
paul_boddieI guess there is some movement towards eliminating the initial root disk versus running system discrepancies, but I can't say that I read all the details.16:32
paul_boddiejekhor: It seems to look for /dev/crypto and not find it. I wonder if this device is supposed to be created somewhere by some script.16:33
paul_boddieThe whole initrd thing is probably well illustrated by things like Knoppix or distro live CDs which manage to run on systems, but when you try and install the actual system, you can end up with something that won't boot. I seem to recall seeing that.16:34
paul_boddiejekhor: I'm no expert on things like device creation. That's another area where things have changed from static device creation for a load of devices you'll never use to more clever stuff, and maybe even getting the kernel to share its knowledge of the devices it supports, but I can't say I've looked at it much, especially with OpenWrt.16:36
paul_boddieA simple mknod with the right arguments might work, but it depends on whether the kernel is able to support the functionality. Alternatively, you might be able to disable the plugin causing the error, but this might be something you actually want enabled, so that wouldn't really help you.16:38
whitequarkafaik the kernel now creates the device files for every device which exists automatically16:38
whitequarkthe job of udev was reduced to setting the correct permissions16:38
whitequarkthat is, with devtmpfs.16:38
paul_boddieSounds familiar.16:39
mwcampbellWhat is the maximum payload that the Ben's ROM bootloader can handle?16:43
paul_boddieI saw this kind of thing with Emdebian, I think.16:43
paul_boddiehttp://en.qi-hardware.com/wiki/Ben_NAND#NAND might be informative.16:44
mwcampbelland how many bytes are in a NAND page?16:46
paul_boddie4096 blocks * 128 pages/block * 4096 bytes/page = 2147483648 bytes (from the wiki)16:48
paul_boddieI don't know what kind of restrictions there are when it comes to the SoC pulling stuff out of NAND in order to boot.16:49
mwcampbellPerhaps the main reason for using a boot loader is to ease reflashing and avoid bricking.16:51
pcercueipaul_boddie: at startup the JZ loads the first 8kB from NAND16:52
pcercueiso the bootloader has to be smaller than 8kB16:52
pcercueiu-boot isn't, so it's divided into two parts, the first one just init the memory and load the second bootloader from the NAND16:53
mwcampbellSo the first part of the boot loader could just load the kernel instead, right? But then reflashing and alternative boot options would be more difficult, I guess.16:56
whitequarkmwcampbell: it is not trivial to fit the NAND FTL code into 8kb16:57
whitequarkeven less so if you have something fancy (and convenient!) like ubifs16:58
mwcampbellWhat does FTL mean in this context?16:58
mwcampbelljust need the expansion, then I can google16:58
whitequarkflash translation layer16:58
paul_boddiepcercuei: Thanks for the insight!16:59
paul_boddieFaster than light! ;-)17:00
mwcampbellSo the first part of u-boot loads the second part form an area of NAND flash that isn't subject to the FTL?17:00
pcercueimwcampbell: yes, that's what ubiboot does (the bootloader we use on the gcw zero) 17:00
pcercueiit weights only ~5kB, and loads a Linux kernel from UBI or SD (FAT)17:01
pcercueimwcampbell: FTL is Flash Translation Layer17:01
pcercueiit's a software layer that maps logical blocks to the real blocks of the NAND17:02
pcercueiso that when one block breaks, it doesn't break the whole memory17:03
whitequarkmwcampbell: first block of NAND is guaranteed to have much better durability than the rest17:05
whitequark1-2 orders of magnitude better at least17:05
mwcampbelland I suppose it will be even more durable if it's reflashed infrequently.17:05
whitequarkwell, it will surely last longer17:10
jekhorI am trying to install and load kmod-crypto-ocf which creates /dev/crypto, but it depends from kmod-crypto-manager and kmod-crypto-core, which doesn't exist.17:10
mwcampbellHere's why I'm asking about the boot process. http://www.loper-os.org/?p=300 argues that computers only come in two speeds, fast and slow, where "slow" means any user-perceptible delay. I'm still trying to understand what makes computers slow, even with processor speeds increasing. One obvious hypothesis is that multi-layered systems and multi-step processes, as opposed to just doing what the user wants in the most direct way possible, 17:12
whitequarkyour message was truncated after "way possible,"17:14
whitequarkbtw I can explain. there are roughly three reasons imo.17:14
mwcampbelloops, too long a message17:14
mwcampbellthe last few words were "are to blame"17:15
whitequarkfirst, there are genuine delays, perhaps network delays. they can be ameliorated by using proper caching, scaling, shaping etc algorithms but not removed17:15
whitequarkit's also sometimes an UI issue ("block the app" vs "add an operations queue")17:15
whitequarksecond, there are poorly designed systems which underutilize available resources17:15
whitequarke.g. a list box with 10k elements which does not cache/prerender items will appear to scroll slowly. it could be fast on the same hardware.17:16
whitequarkthird, there is some genuine increase in the complexity of systems. but more often than not it's intrinsic, not something due to poor design.17:17
whitequarksay you see that your IM client renders a megabyte of accidentally pasted text very slowly. you complain. but what does it do? it invokes the freetype engine which is ought to render every single character ever existed on this planet, with kerning, antialiasing, hinting and whatnot17:18
whitequarkthis is a very perceptible positive change in the UI, which does increase the resource requirements by orders of magnitude17:18
whitequarkand it should, and this change is welcome anyway.17:18
pcercueiopenVG :)17:19
mwcampbellThe standout quote from that article I linked is, "The GUI of my 4MHz Symbolics 3620 lisp machine is more responsive on average than that of my 3GHz office PC.  The former boots (into a graphical everything-visible-and-modifiable programming environment, the most expressive ever created) faster than the latter boots into its syrupy imponade hell."17:22
mwcampbellThat one didn't get truncated, did it?17:22
mwcampbellprobably did :(17:22
paul_boddieNo. You weren't affected by varchar(256) this time. ;-)17:23
mwcampbellon the surface, that quote seems like quite a criticism of software developers over the past few decades.17:23
mwcampbellBut whitequark makes a good point about actual advances in software features that necessarily require more resources.17:23
mwcampbellI wonder, what features would one have to sacrifice to have a fast-booting (defined as 10 seconds or less), responsive GUI on the relatively underpowered hardware of the Ben NanoNote?17:25
whitequarkmwcampbell: not that it is ever correct to compare cycles per second :p17:26
whitequark(unless it's exact same microarchitecture, I'd say.)17:26
pcercueidoes it use sysinit or systemd?17:27
mwcampbellDoes what? That blogger's office desktop?17:28
mwcampbellI have no clue17:28
whitequarkbesides, gui responsiveness doesn't have much to do to achieve17:29
whitequarkit's a matter of separating gui operations into their own thread and giving it the highest priority, as the iOS clearly demonstrates17:29
whitequarkI guess whatever that blogger runs (gtk?) cannot do that.17:30
mwcampbellI guess that blog post is more hot air than reasoned argument17:32
mwcampbellI'm sorry I took us so far off-topic with that17:32
mwcampbellSo, someone mentioned that the ubiboot bootloader can load a kernel from ubifs or SD with only about 8KB of code. What advantages does u-boot provide over ubiboot?17:33
paul_boddieI don't think it's off-topic to consider the properties of a system like fast boot, really. Everyone looks at microcomputers and asks why they are usable within a second of power on and yet today's machines need to "think about" starting up.17:34
whitequarkyou can achieve these speeds by tuning linux correctly17:34
paul_boddieYes, I know. I mean, you can boot into a Linux plus busybox environment very quickly without really thinking about tuning at all.17:35
whitequarkyou'd need an SSD, a custom kernel compiled for your hardware and parallelized boot process without anything extraneous. you could get to GUI very fast17:35
whitequarka matter of less than 10 seconds. I think I once did four.17:35
whitequarkit's a given fact that PCI enumeration takes very real and significant time17:36
whitequarkoh also like half of the boot time of my notebook is BIOS/EIF17:36
whitequarkargh, EFI17:36
paul_boddieA lot of the focus on parallel boot and the other stuff is necessitated by the amount of cruft in the boot process, however.17:36
paul_boddieArgh, EFI, indeed. ;-)17:36
whitequarkpaul_boddie: no, not really17:36
whitequarkit's only cruft if you lack perspective17:36
whitequarkis a network managing service "cruft"? is printer server "cruft"? etc.17:37
whitequarkpeople want everything to Just Work". Software abides.17:37
paul_boddieYes and yes. I have a desktop which doesn't need network management and I don't have a printer.17:37
whitequarkwell, if you care about that so much, go turn it off and get faster boot.17:37
paul_boddieAnd the printer doesn't need to be available immediately during the boot. Actually, I can be launching applications in about a minute after boot, so it's not important to me, but the continuation of the "telephone switch operating system" mindset means that a lot of people just accept a gazillion services running and people adding yet more without anyone asking whether they are all necessary.17:39
whitequarkwell that is the point of parallel boot17:40
paul_boddieThat's what I said! :-)17:40
whitequarkbesides, boot time doesn't matter17:40
whitequarkACPI S3 is usable even on Linux for almost a decade now I think17:41
paul_boddieI'll accept that I hate the alternative extreme where Windows boots (slowly) and you have to wait for it to become usable, the desktop icons appear then disappear then appear again, chugging noises are heard, the hourglass appears.17:41
paul_boddiemwcampbell: I imagine that if you deployed a minimal Linux plus busybox environment, you'd get to a shell within 10 seconds.17:45
mwcampbellMy Ben will arrive tomorrow (hopefully), so I'll soon find out.17:46
whitequarkalso funny how people complain that the new dynamic languages waste lots of cycles17:46
lekernel"The goal is to create something which can one day be seamlessly converted to a fully-asynchronous design, implemented using Muller C-gates. The latter is extremely difficult (though not entirely impossible) on available FPGAs, so the adult form of the design would have to be fabbed in actual silicon"17:47
whitequarkeven in cases when the workload is dominated by I/O time. even when I can modify one line in any given file of an equivalent of 100KLOC or 1MLOC C++ project and get to run it within four seconds.17:47
lekernelit's easier to do it with fpgas than silicon17:47
lekernelthere are the same problems, only you don't need a slow and expensive respin with the fpga when you fuck up17:48
lekernel(from http://www.loper-os.org/?p=846)17:48
mwcampbellwhitequark: Do you use any dynamic languages on the Ben?17:51
paul_boddieI have used Python on the Ben.17:51
mwcampbellMaybe this is just a psychological effect, but I assumedf that I would need to avoid dynamic languages on the Ben, because after all, it's only a 336 MHz MIPS processor, so it'll probably be hard work to write responsive applications17:52
paul_boddieWell, I was using Python on early 1990s workstations with 25MHz or so processors, so the CPU speed is not really an issue, even if Python has become less lean over the years.17:55
paul_boddieOf course, you might want to use a systems programming language to get more horsepower. On my long list of things to do, I aim to compile some Python stuff using Shedskin for the Ben and see how that turns out.17:56
paul_boddieIn my experience, the more significant problem is likely to be the RAM footprint, which for an untuned version of Python is not too bad, but if you start importing modules that need other modules like numpy, then you can easily exhaust the RAM.17:57
paul_boddieThat's arguably a problem with numpy and the way they bundle absolutely everything into the package so that you're loading a ton of Fortran-related stuff when you (or rather pygame) imports numpy.17:58
mwcampbellGood point about RAM usage.17:59
mwcampbellIn a language that's compiled to native code ahead-of-time, the OS can mmap in just the pages of code that it needs.18:00
mwcampbellWhereas in an interpreted language, all of the code that you just imported is just data as far as the kernel is concerned, so it has to be kept in memory.18:00
mwcampbellAnd of course JIT compilation makes RAM usage worse.18:01
mwcampbellIs there an explanation somewhere of why the Ben's hardware has such modest specs in the first place? That has to be a common question.18:02
mwcampbellMy best guess is that it was such necessary to use low-spec hardware in order to reach an acceptable price even for a small-scale manufacturing run.18:03
mwcampbells/such necessary/necessary/18:03
mwcampbelloh, nice :)18:04
pcercueinever heard about shedskin, looks intetesting18:05
paul_boddieWith the Ben you always have to remember that the starting point was someone else's device at a particular price point. That in turn would have influenced and been influenced by the availability of SoCs known to and understood by the manufacturer.18:06
paul_boddieI don't use Shedskin that much, but I have used it for prototyping some stuff that others would perhaps want to rewrite in numpy-based technologies, and it was a pretty good fit and gave a nice speed-up.18:07
paul_boddiePeople get upset about it because there are limitations, but people tend to overstate them and want all their ultra-dynamic sugar for their programs and yet not have those programs get fat.18:08
Action: paul_boddie has to go.18:09
mwcampbellThe Ben was based on another device?18:17
whitequark(ultra-dynamic sugar) well, that's why we use high-level languages.18:20
whitequarkproperly used metaprogramming can reduce the size of program 10x, or even 100x, in extreme cases.18:20
wpwrak(cruft) hmm, ubuntu 12.10: comes up, network manager finds Ethernet, Bens, configures each as possible internet access. then rotates between them, causing rather interesting failure patterns.18:28
wpwrakin this case, there would be two choices: 1) keep it simple, or 2) only use what actually works.18:28
whitequarkwpwrak: that is a bug, not cruft ;)18:29
wpwrak(cruft II) on to cups. as usual, printing didn't work. nota bene, this is a fresh install. turns out it somehow mis-autodetected the printer and it only figured out what it was after deleting and re-creating it. the printer is a postscript printer and all i want is postscript to be sent to it untouched. ancient lpd could handle this perfectly, without ever breaking. now we have cups that has some regression on what's the same set of t18:31
wpwrakasks on each release.18:31
wpwrak(cruft III) xorg. each time i have to install a new ububtu, xorg regresses in my triple-head setup, usually requiring a day of searching for returning to the status quo antes.18:33
viricthere you go, with ubuntu18:34
wpwraksometimes it gets worse: once, we had to debug the radeon driver. this time, i had to install a patched non-official xorg server package to get at least mousing out of the first screen. 3rd screen is still gone, without a clue why.18:34
mwcampbellI'm sorry I started us on that tangent about boot time and cruft.18:35
wpwrakwhitequark: it's software complexity growing until the thing collapses under its own weight. in all three case, these are things that keep coming back. they're no isolated bugs. xorg is particularly bad. once you have to find some magical kernel command line option, once you need to fix the driver, once you need a very hackish downgrade to an earlier server version, etc.18:37
wpwrakshe sheer amount of trouble also means that community fora gets crowded with yesteryear's fixes for similar problems and it's increasingly hard to find sonething that actually helps to move forward18:38
wpwrak(just venting about my experience of the last few hours :)18:39
whitequarkthis is the price of complexity, combined with diversity, combined with proprietary devices.18:40
wpwrakone could of course say that all this is an epic QA failure. maybe that's the case.18:40
wpwrakwhitequark: in my case, the hardware is the same for a very long time, and not exotic at all. and of course i don't use any proprietary drivers. yet each new version runs into trouble.18:41
mwcampbellMy Ben just arrived, a day early. I know what I'll be doing tonight. :)18:48
wpwrakenjoy an uncrufted system :)18:51
mwcampbelland I'll hopefully develop some uncrufted software to run on it18:52
mwcampbellpaul_boddie said earlier that the starting point for the Ben was someone else's device. Does this mean that the Ben is an unauthorized clone of some other product, or merely that it used the components of some other product as a starting point?19:00
wpwrakit's not a clone :) it's a variant (customized keyboard, etc.) of a dictionary, made by the company that made the dictionary19:03
mwcampbellah, OK19:04
mwcampbellAnd presumably that company authorized the copylefting of the hardware design?19:05
mwcampbellAnyway, that origin explains why the specs are so modest. One surely doesn't need a very fast processor for a dictionary.19:05
wpwrakthe hardware design isn't fully copylefted. i suppose they authorized the re-making of the schematics, though.19:06
wpwrakthings we don't have are layout and mechanical design19:06
mwcampbellah OK19:07
mwcampbellanyway, the Ben seems to be much more hacker-friendly than many mainstream devices19:08
wpwrakwell, there are partial scans of the case, which have some use if you want to make mechanical add-ons: http://www.almesberger.net/misc/ben/scan/19:08
wpwrak(the partial scans never got finished because one day the pc on which i had a VM with windows that was running the scanning application died)19:09
mwcampbellI'll probably only be doing stuff at the software level. I'm no electrical engineer.19:09
mwcampbelljust a programmer19:09
wpwrakactually, this is a better url. less traffic on my site :) http://downloads.qi-hardware.com/people/werner/ben-scans/19:12
mwcampbellWhat kind of keyboard, if any, did the original dictionary product have? just curious19:13
wpwraki suppose it was somewhat similar19:14
mwcampbellIn the stock software installation, is the OpenWrt logo displayed by the kernel itself, or by a user-space process early in the boot procedure?20:31
mwcampbellAny advice on which GUI toolkit to use if I'm writing a new app that I want to be responsive on the Ben?21:07
mwcampbellI won't be writing any games; the display will be used primarily for text and UI controls.21:07
mwcampbellIs GTK reasonably snappy on the Ben?21:08
kristianpaulby default there is not tk21:08
kristianpaulall is framebuffer21:08
kristianpaulyou need compile your own openwrt for using gtk21:08
mwcampbelloh, I thought there was a framebuffer version of GTK included.21:09
kristianpaulbetter try allegro or why not pyallegro21:09
kristianpaulah there is somwhere yes21:09
mwcampbellI'll check out Allegro.21:09
kristianpaulpyallegro was already introcuded to the nanonote by david k so it should worth a look as well21:16
mwcampbellNo vector graphics or TrueType fonts in Allegro though, right?21:21
mwcampbellso a UI made with Allegro would probably look retro, IIUC21:22
mwcampbellthen again, any UI that is responsive on a 336 MHz processor without the assistance of a GPU might look retro21:24
wpwrakor very futuristic ;-)21:26
mwcampbellBTW, does the Qi Hardware community need a faster download mirror, particularly in the US? I'm getting ~20 KB/s from downloads.qi-hardware.com21:26
mwcampbellI'm not complaining; in fact, I'm offering to help21:27
mwcampbellkristianpaul: Looks like the latest Ben NanoNote image does have a DirectFB-based build of GTK.21:32
kristianpaulgood :)21:33
kristianpaulbtw if you need gtk on a xfdev, there is jlime port as well21:34
wpwraknot sure if one can really recommend jlime anymore. as far as i know, it'21:35
wpwraks been unmaintained for years.21:35
mwcampbellX seems quite superfluous on a device like this anyway21:35
mwcampbella screen this small practically requires that you only run one app at a time, and it seems to me that X offers nothing at all if one can run GTK, Pango, Cairo, etc. directly on the framebuffer device.21:36
mwcampbellSo which is better on a device like the Ben: GTK or Qt?21:37
kristianpaulwpwrak: well yes but it still smooth for some tasks no ;)21:38
mwcampbellOne appealing thing about GTK is that if I use GTK, then I can write apps in the Vala language, so I'd get ahead-of-time compilation with the benefits of a more high-level language than C/C++21:38
wpwrakkristianpaul: oh, it's very smooth ... until you need a cross-development environment, and all you find is an ancient toolchain (a great one though, much easier to use than the one from openwrt)21:40
wpwrakfor all my little bits of development on the ben, i used SDL. but that's quite low-level. pixels and lines :)21:41
mwcampbellAllegro is equally low-level, right?21:42
kristianpaulwpwrak: ;)21:43
kristianpaulyeap easy to use it is21:43
kristianpaulnot get get made i guess :)21:43
--- Thu Mar 21 201300:00

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