#qi-hardware IRC log for Saturday, 2017-01-14

wpwrakwhitequark: hmm, with current master, i still get "OpenGL context creation failed"00:52
wpwrak... in both GW and TW00:53
whitequarkexport GDK_DEBUG=OPENGL00:56
whitequark(and yeah, that is a completely different bug now)00:56
wpwrak(the DISPLAY is to be able to see where it stops, without having to interact with it first)01:03
whitequarkI haven't a slightest clue01:06
whitequarkexport GDK_DEBUG=ALL ?01:06
wpwrakthe only things that look fishy to me are those g_value_take_object: assertion 'G_IS_OBJECT (v_object)' failed01:09
whitequarkhmmm, I have an idea01:10
whitequarkwpwrak: lol01:11
whitequarkthis is a GTK bug01:11
whitequarkfixed innnnn https://git.gnome.org/browse/gtkmm/commit/?id=d1ae01f3dc14dfbadb3e6288c6bdab5f4a1332b501:11
wpwrakall this debug output looks as if it should be useful to figuring out why my redirected TW disappears, too. and easier to use than an event handler override ;-)01:12
wpwraknow what version would that be ... guess > 3.20 :-(01:13
whitequarkwpwrak: you can work around that01:17
wpwrakbump some refcount, i guess ? with something like #if GTKMM_VERSION_MINOR < 2001:18
wpwrakerr, < 21 or < 2201:18
whitequarktry this patch https://hastebin.com/hiqivovudu.cpp01:21
whitequarkno, hm01:24
wpwrakthat removes one of the two (solvespace:12851): GLib-GObject-CRITICAL **: g_value_take_object: assertion 'G_IS_OBJECT (v_object)' failed01:24
wpwrakopengl creation still fails (in both windows), though01:25
whitequarkwpwrak: https://hastebin.com/voqacunuyo.cpp01:35
whitequarkthis one01:35
wpwrakmakes opengl context creation work for GW, but with a quirk: the screen comes up with a white background, and only turns black when something happens01:41
wpwrakalso, TW still fails01:41
whitequarkyeah I didn't patch that01:41
wpwrak"happens" = e.g., hover over axis or icon01:42
whitequarkwpwrak: this one? https://hastebin.com/afiqahekeb.cpp01:47
wpwrakapplying on top of the previous one ... compiling ...01:51
wpwraknope. still exactly like before.01:52
whitequarkwpwrak: https://hastebin.com/sumaxofogi.hs ?02:35
whitequarkwpwrak: wait what03:52
whitequarkdid you seriously write lilo?!03:53
DocScrutinizer05didn't you know?04:07
whitequarknope, I only knew about pstree04:07
DocScrutinizer05wiki Werner Almesberger ;-)04:25
whitequarkeven initrd04:27
DocScrutinizer05aah you mentioned that04:28
DocScrutinizer05msdos tools04:34
wpwrakof those, initrd was probably the best. proper innovation, the stuff patents should be made of.05:27
wpwrakand it also solved a rather nasty situation that began to build up at that time: distributions needed more and more different boot floppies, to avoid driver X from irreparably (well, until reset) trashing registers hardware Y. the joy of ISA. we already had modules, but no way to control at boot time which to load.05:30
wpwrakwhitequark: applied sumaxofogi on top of voqacunuyo and afiqahekeb05:42
wpwrakGW now comes up cleanly (i.e., black background). TW still fails to bring up opengl05:43
wpwraklemme see if this goes away if i pick up the latest changes from git ...05:44
wpwrakbtw, there are two build warnings (gcc 6.2.0): https://0bin.net/paste/35w4kJQqZ4S6zFHj#rKiSYlanF4IYFDLhpKBgo5F-g2Cf+9C5WWZS1w5vT1X05:47
wpwrakseems that they added some python antidote ;-)05:48
wpwrakmaster @ 5c34b3f6ef86b94b9360023ddb5c6a35f935ade6 + sumaxofogi: GW and TW come up, background is black. however ... GW doesn't show the dashed lines until ther is some "action". also, slvs prints 2 x "glGetError() == 0x501"05:52
whitequarkwhat does MESA_DEBUG LIBGL_DEBUG bring out?05:57
whitequarkwait nevermind I can reproduce that05:57
whitequarkwpwrak: https://hastebin.com/nosuwoyona.php05:59
wpwrakwhee ! :)06:02
wpwraknow all looks good: GW and TW come up, background color is right, update happens in time, no complaints on the console.06:03
wpwrakline drawing is still as glacially slow as before. but i guess that's unrelated.06:04
whitequarkwait, really?06:05
whitequarkthat... makes no sense06:05
wpwrakdid that ever stop computers from misbehaving ? :)06:07
whitequarkwpwrak: pushed that to master06:10
whitequarkwpwrak: can you do: cmake . -DOPENGL=1 and rebuild?06:11
whitequarkwpwrak: nevermind, that's not actually possible to do on GTK3.06:11
Action: whitequark stares06:11
whitequarkwpwrak: are you *absolutely sure* your X server / graphics driver combo isn't totally fucked up?06:12
wpwrakbuilt master ... works (just with the slowness)06:12
whitequarkcan you show me GDK_DEBUG=ALL output and drag something a bit06:13
whitequarkon my machine, dragging a line in solvespace eats about 25% of a core plus 10% for xorg06:14
whitequarkand it's incredibly snappy06:14
wpwrakx server is generally a bit sluggish, because if memory is always runnings low and there's a background load of junk in browser tabs, but nothing that would explain lines being so slow06:14
whitequarkwhat does the fps counter say?06:15
whitequarkagain on my machine with intel graphics it takes 1ms to draw a frame06:15
whitequarkeven with DocScrutinizer05's 1200ppi mouse it would probably not have much difference coping with the load!06:15
wpwrakwhoa ! GDK_DEBUG=ALL -> https://0bin.net/paste/tTlNPpLxPQ3UFTIf#En1+HvbdKYB1WoYK1yX6z0+A6U2TN18xVFxfXK5gAa606:16
whitequarkGdk-Message:     0: layout_start=0.0  paint_start=18.7 frame_end=467.2 predicted=16.706:17
whitequarkwhat the *hell*06:17
whitequark>paint_start=18.7 frame_end=467.206:17
wpwrakfps pretty consistently around 5906:18
whitequarkso it says "59 1/s" ?06:19
wpwrakactually, the timing has changed06:20
wpwrakbefore, it was lagging. now, the line isn't updated at all until the mouse stops moving06:20
wpwrakwhile, say, highlighting an axis, still happens, even while the line is "stalled"06:21
whitequarkwhat about this https://hastebin.com/najelucosu.lua06:21
wpwraktrying ...06:22
wpwraknope, still the same06:23
whitequarkwell, it's not actually *slow*06:24
whitequarkwhat happens is it waits for vsync.06:24
whitequarkI have no idea why, to be honest06:25
wpwrakyeah, i think before it was genuinely slow/laggy. but now, it's some update prevention06:25
whitequarkthatmight have been06:25
wpwrakwhat's interesting is that highlighting of axes does not move a "stalled" line.06:25
whitequarkwhat do you mean?06:26
wpwraktime for a video :)06:26
whitequark... hang on I think I know what miiiight be happening06:29
whitequarkand it's rather disturbing06:29
whitequarkcan you do: ulimit -s 8192 then run solvespace until it (presumably) crashes?06:30
wpwrakwait .. having a system problem ...06:32
wpwraksoemthing is filling my disk :(06:34
wpwrakhere we go: http://downloads.qi-hardware.com/people/werner/slvs/drag-noupdate.mp406:43
wpwrak"special", isn't it ? :)06:44
whitequarkalso uhm06:44
whitequarklook at the fps counter06:44
whitequarkdoesn't that say 1/2ms every time?06:44
wpwrakah yes, that has gotten faster. could be because i killed the background activities.06:45
whitequarkok so it has nothing to do with vsync06:46
whitequarkor glFlush06:46
wpwrakulimit so far hasn't caused any trouble. been running it for a couple of minutes, moving the mouse around06:46
wpwrakwhat's also fun is that, when moving the mouse slowly over the horizontal/vertical position, the H/V appears and follows the mouse, while the line remains frozen06:47
whitequarkthat one I know of...06:50
wpwraki'll go for a nap. no new bugs or nagging about features for a while ;)07:02
wpwrakwhitequark: joyful news: seems that the opengl fixes also solved the text window dying when opening issue. at least i couldn't reproduce it in a number of tests now. updated patch: http://downloads.qi-hardware.com/people/werner/slvs/display-textwnd.patch12:48
wpwrakone thing i noticed is that it seems that the window gets mapped before receiving its position, i.e., it / its frame very briefly appears at a different position, then the window is drawn at the final position. not sure who does what in that sequence.12:49
whitequarkwpwrak: that's "normal"12:49
wpwrakperfect then ;-)12:49
whitequarkit's slightly inelegant but really not worth fixing12:50
whitequarkmaybe once I rewrite all of the platform code to be sane12:50
wpwrakyeah, doesn't cause problems12:51
whitequarkpatch looks better now12:52
whitequarkwell, I'm not happy with how it's structured and formatted but at least the semantics makes sense12:52
whitequarkwill apply in a few hours12:53
wpwrakthanks !12:55
DocScrutinizer05((very briefly appears at a different position)) dosn't cause problems as long as no stupid windows desktop re-arranges all icons on screen permanently, like it's done when you switch to a different display resolution14:27
DocScrutinizer05...on windoze14:27
DocScrutinizer05same effect happens as well when windows reloads display driver, on NVidia updating the graca driver crap14:30
DocScrutinizer050,3s black, then deskzop comes back re-arranged14:31
DocScrutinizer05rather boring video of "interesting snowflakes" http://maemo.cloud-7.de/share-service/DSCF2275.MOV - what's *really* highly interesting is the shake stabilizer of that Fuji camera, conider I took that video one_handed and the original shaking was X and Y just as bad as in Z-rotation, but the stabilizer totally removed it, except the latter14:55
strawberyargonDocScrutinizer05, yea its good. i wonder what it would be like with even less compression :) example: http://blog.planet5d.com/2015/10/magic-lantern-shows-power-of-canon/18:44
strawberyargoni feel "yea its good" didnt do it justice, its most acceptable to the point i dont think about shakyness. so my focus then went to the next thing, which was compression18:47
DocScrutinizer05well, 5D mk-II (or III) yeah, I wish I could at least remote trigger this Fuji XQ1, but then it's a completely different class than Canon 5D and alas no magic lantern or _anything_ available19:44
DocScrutinizer05a 5D would be overkill for my purposes19:46
DocScrutinizer05but something with at least 100mm (better 150), focus distance <40cm, magnif >= 1.0 would be much appreciated19:48
DocScrutinizer05and I actually started looking for cameras that allow hacked firmware and have a decent versatile hacked firmware19:49
DocScrutinizer05a used 5D-Mk1 with primary 100mm macro might match all my criteria19:50
DocScrutinizer05then OTOH a DL24-500 (with magn=4) is a nice telescope too ;-) I just wonder if the macro properties it has are any good, also if it ever might see hacked firmware19:54
DocScrutinizer05>> Magnification Up to 4 x (angle of view equivalent to that of approx. 2000 mm lens in 35 mm [135] format)<<19:55
DocScrutinizer05>>Focus range: Macro close-up mode: Approx. 1 cm (0.4 in.) to infinity (wide-angle position), or 70 cm (2 ft 4 in.) to infinity (telephoto position) <<19:57
DocScrutinizer05not quite sure if the latter really means I could do decent macro shots of areas 10*10mm, with lens min 30cm from object19:58
DocScrutinizer05if "telephoto position" actually means "effective 2000mm lens" then I guess a distance of 70cm is quite good still, for effective magnification20:00
DocScrutinizer05but too much guessing from my side in all that20:02
wpwrakcanon powershot powershow sd880is aka ixus 870 is goes down to 20 mm and runs chdk. and you've already seen many pictures taken by it ;-)20:04
DocScrutinizer05ooh, that 4 times magnification is digital, meh20:05
DocScrutinizer05I don't want "to go down to..." - I need a certain minimum distance of 30cm from lens to object ;-)20:06
wpwrakat 20 mm, your main problem is to get light to the object, because the lens tends to shadow it pretty well, especially if you don't have a lot of ambient light.20:06
wpwrakthat nikon has a nice exposure range. the sd880is only goes -3/+3 EV, which can sometimes be insufficient.20:08
DocScrutinizer05with DL24-500 in tele mode, minimum distance is 70cm which is maybe a tiny bit clumsy since I need to stand up to reach the camera, but as long as it makes crisp pictures (actually video, via HDMI-out) of an area as 'large' as 10*10mm20:08
wpwraknaw, for getting close, you want macro mode.20:09
wpwraktele may sound useful, but it isn't :)20:09
DocScrutinizer05I have three requirements: format filling 10*10mm macros, video to large monitor, *minimum* 30cm distance object<->lens20:09
DocScrutinizer05if the thing can serve "dual-use" as mice "telescope" too, I won't complain20:11
wpwrak30 cm or mm ?20:12
DocScrutinizer052000mm effective lens sounds like the latter is granted20:12
wpwrakyou want a 10x10 mm field from 30 cm ?20:12
wpwraki hope you're good at grinding your own lenses ;-)20:13
wpwraki doubt anything even remotely like this is commercially available20:13
DocScrutinizer05aiui a 150mm lens has object size == picture_on_sensor size when object has 150mm distance from sensor20:14
DocScrutinizer05aka "magnicication = 1.0"20:15
DocScrutinizer05if it could get focused to 15cm distance20:15
wpwrakyes, that's the problem20:15
DocScrutinizer05macro lenses of 100mm or 150mm start at 30cm20:16
DocScrutinizer05but I don't need a 25mm sensor20:16
DocScrutinizer05generally a macro lens is defined as "magnification >=1" which is just a question of lens focal length / min focus distance20:17
DocScrutinizer05of course a 20mm lens can do magnification=1 in a focus distance of 20mm20:18
DocScrutinizer05s/object has 150mm distance from sensor/object has 150mm distance from LENS/20:20
DocScrutinizer05focal length == focus distance20:20
wpwrakhmm, there do indeed semm to be lenses of the type you want: http://www.the-digital-picture.com/Reviews/Sigma-150mm-f-2.8-EX-DG-HSM-Macro-Lens-Review.aspx20:21
DocScrutinizer05of course there are, I skimmed through them today :-)20:21
wpwrakah :)20:22
DocScrutinizer05note that I'm massively guessing all that stuff, started to learn about it just 3h ago20:24
wpwrakheh, focal range = lens length :)20:24
wpwrakerr, no, wrong parameter20:25
wpwrakfocal distance is what you want. and that's 30-44 cm in the ones you picked20:25
DocScrutinizer05the Tokina looks actually reasonably well spec'ed and priced for the purpose20:28
DocScrutinizer05but then I stumbled over http://cameradecision.com/review/Nikon-DL24-50020:30
DocScrutinizer05then (about same time) I started considering firmware hacks20:30
wpwrak(dl24) that 10 mm focus range is probably at 24 mm. at tele, the ratio may not be so good. but i guess you'd have to try it. that's the sort of stuff you don't find in specs, and even reviewers don't do much testing of macro.20:35
wpwrakfor firmware hacks, i think canon (with chdk) still rules20:35
DocScrutinizer05>>Focus range: Macro close-up mode: Approx. 1 cm (0.4 in.) to infinity (wide-angle position), or 70 cm (2 ft 4 in.) to infinity (telephoto position) <<20:35
wpwrakwell, not too bad20:36
DocScrutinizer05so yes, "wide angle position" is ,ost likely "24mm zoom"20:36
DocScrutinizer05I'm however more interested in the other "option"20:37
DocScrutinizer05[2017-01-14 Sat 21:00:25] <DocScrutinizer05> if "telephoto position" actually means "effective 2000mm lens" then I guess a distance of 70cm is quite good still, for effective magnification20:37
DocScrutinizer05aiui "effective 2000mm lens" means a 35/4mm object would fill the sensor's diagonal?20:39
DocScrutinizer05well, nope wait20:40
DocScrutinizer05meh, that math20:40
DocScrutinizer05let's start with 500mm lens, which means a single lens 500mm from sensor focuses parallel ("infinity") light to the sensor20:42
DocScrutinizer05so if you can focus this lens to an object 500mm away, you get a magnification of 1.0, right? IOW an object size of the sensor will fill the picture20:43
DocScrutinizer05now we're roughtly with my flawed math above: the cam does 4 times digital zoom, making the sensor smaller. and the focus distance isn't 500mm but 700mm20:44
DocScrutinizer05sensor size 1" (I assume that's diagonal?), so when magnification is "area" (is it?) we have an effective 1/2" sensor with digital 4x zoom, so a 0.5" object should fill the sensor diagonal, which is just within my requirement specs20:48
DocScrutinizer05but then lens length is usually standardized to 35mm 'sensor', not 1"20:49
DocScrutinizer05puzzling shit, makes me dizzy20:49
DocScrutinizer05on the bright side, it seems the DL24-500 is waiting to suffer some thorough tests in next Saturn market20:51
wpwrakwell, they usually specify the equivalent20:51
DocScrutinizer05yep, that's what I meant20:52
DocScrutinizer05so when I use a 'fake' 500mm lens length value, I'd also need to use a 'fake' 35mm sensor, right?20:52
DocScrutinizer05not 1"20:53
DocScrutinizer05so an object of 35mm size would fill the sensor diagonal of DL24-500 when placed 700mm from lens surface in tele macro mode20:55
DocScrutinizer05unless we apply digital 4x zoom20:55
DocScrutinizer05well, actually 700/500 * 35mm20:56
wpwrakisn't it 700 from "film" (sensor) surface ?20:56
DocScrutinizer05nope, I think that's from lens surface20:57
DocScrutinizer05the focus distance20:57
DocScrutinizer05http://www.europe-nikon.com/en_GB/product/digital-cameras/premium-compact/dl24-500-f-2-8-5-6#tech_specs  >>[W]: Approx. 30 cm (1 ft) to infinity, [T]: Approx. 2.0 m (6 ft 7 in.) to infinity, Macro close-up mode: Approx. 1 cm (0.4 in.) to infinity (wide-angle position), or 70 cm (2 ft 4 in.) to infinity (telephoto position) (All distances measured from center of front surface of lens)<<20:59
DocScrutinizer05so I guess it's possibly correct to assume an object of 35mm *7/5 size would fill the diagonale of the 20.8 million px DL24 sensor, when placed a 700mm in front of lens surface in tele macro mode21:02
DocScrutinizer05zooming in digitally might still yield somewhat useful video (for work) at 10 times that, so a 3 to 4mm would fill the monitor diagonal21:06
DocScrutinizer05et a ~600*400 resolution21:07
DocScrutinizer05and 70cm is more than enough headroom for all tooling like soldering irons etc ;-)21:20
DocScrutinizer05WOW, Saturn makes that DL24 a sent from Starship Enterprise: viewfinder 2359 Megapixel21:32
DocScrutinizer05should suffice for proper holograms21:33
DocScrutinizer05wtf? http://cameradecision.com/review/Canon-PowerShot-G1-X-Mark-II  >>Pro: Electronic (Optional) Optional Viewfinder //  Con: Electronic (Optional) No External Viewfinder Option << what now?21:53
DocScrutinizer05a pity but canon seems doesn't cut it22:00
paul_boddieI didn't think that any of the Nikon DL models were actually available yet.23:01
--- Sun Jan 15 201700:00

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