Previous day | Jump to hour: 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 | Next day

Seconds: Show Hide | Joins: Show Hide | View raw
Font: Serif Sans-Serif Monospace | Size: Small Medium Large

Click in the nick column to highlight everything a person has said.
The Logo icon identifies that the person is a core developer (has commit access).

Notice: Only Gecko based browsers prior to FF4 support the multipart/mixed "server push" method used by this log reader to auto-update. Since you do not appear to use such a browser, this page will simply show the current log, and not automatically update.

#rockbox log for 2017-01-10

00:04:09pamauryah... I feel like an idiot
00:04:14pamauryI had the hold switch
00:04:25lebelliumok working!
00:04:51lebelliumnever seen the 240x320 cabbie v2 on such a small screen :D
00:06:07pamaurylebellium: there is a small bug in the bootloader with respect to HOLD handling
00:06:17pamauryfor now make sure you boot with HOLD on off
00:06:23pamauryotherwise keys won't work
00:06:28pamauryI'll fix in my next gerrit update
00:08:28pamaurylebellium: can you confirm sound is not working ?
00:09:45lebelliumI confirm the sound is not working
00:09:52lebelliumbut playback doesn't crash
00:10:43lebelliumit's funny because album art does show up properly with cabbiev2 but not with my theme
00:10:44pamauryyes, it's "just" a problem with audio output apparently
00:11:24lebelliumoh and it believes the battery is empty
00:11:32pamaurylebellium: I think you guessed it, to access your files you need to go to the "contents" folder
00:11:46lebelliumI guessed it :)
00:11:47pamauryI need to figure out how to make it start there by default
00:12:04lebelliumon YP-R0 it's in mnt
00:12:04pamauryah ? It is supposed to read battery correctly
00:13:24lebelliumthe battery indicator just moves
00:13:30lebelliumnot stable
00:14:38pamaurylebellium: go to System > Debug > View battery
00:14:54pamaurywhat does it show ?
00:16:04pamauryand rockbox displays an empty battery bar ?
00:16:38lebellium4.153V now
00:18:43lebelliumno it's 97%
00:19:18pamauryI'm confused, you said it was thinking it's empty
00:19:35lebelliumyes, low battery message and shutting down
00:20:01lebelliumnow it seems to be stable at 97% but it wasn't
00:21:46pamaurywell I don't know, I'm reading the battery using Sony's power module
00:22:25lebelliumit just moves from 97 to 99 and back to 97%
00:22:29lebelliumwe'll see that later
00:24:15 Part robertd1
00:26:34pamaurywell there are several options to read the battery
00:26:55pamaurythe power modules does magic and converts it to a 5 level value (low, 1, 2, 3, 4, full)
00:27:12pamaurybut I found it a bit imprecise so I used the actual voltage
00:27:50lebelliumcongrats to pamaury
00:29:09pamaurythe sony logo is a bit ugly on the E580
00:29:41pamaurythe bootloader extract a 130x130 icon which is ok on the E450 and E460 but on the E580 they added a gradient that makes the cut very visible
00:29:50pamaurybut that's a minor detail
00:30:32lebelliumon E450 you don't see the blue/black difference?!
00:30:45pamaurythe bootloader also misses important features, like warning when the HOLD switch is on (instead of doing nothing), also it misses a timeout (it will happily drain the battery waiting for the user currently) for a default action/shutdown
00:30:56pamaurylebellium: no, the logo is a bit different
00:31:47lebelliumI guess we deserve to sleep now!
00:32:28pamauryyes :)
00:32:41pamaurytomorrow I'll add the A15
00:33:52lebelliumthanks for the efforts
00:33:54lebelliumgood night
00:38:28chrisjjpamaury: be68b6a7b crashed likewise, after 2-3hrs.
00:38:36***Saving seen data "./dancer.seen"
00:39:22pamauryit justs powers off ? are you use you have a long playlist and you have repeat ?
00:39:44 Quit lebellium (Quit: ChatZilla 0.9.93 [Firefox 50.1.0/20161208153507])
00:40:07chrisjjYes I am sure.
00:40:29chrisjjPlaylist has 500 3min tracks and Repeat = All.
00:40:46chrisjjNow I can't say /just/ powers off.
00:40:56pamaurythat's weird. The last commit related to imx233 (so zen indirect) dates from september
00:41:10pamauryhave you tried a dev build from more recently ?
00:41:30chrisjjI can say screen goes black, sound goes silent, keys go non-responsive... and power button press does a power-on operation suggesting power went off.
00:42:09pamaurypotentially sounds like a panic to me
00:42:12chrisjjNo I haven't because TMK the official dev builds do not include lcd_fix.
00:42:56chrisjjZEN panics I've seen don't go black screen.
00:43:31chrisjjEven if there's a backlight timeout, they show black screen on white background that stays indefinitely.
00:43:57chrisjjHowever, I may never have encountered a panic occurring when backlight is off.
00:44:05pamaurythat depends, a panic in some places may not be able to make the screen work
00:44:14chrisjjHow can I determine whether the black screen is hiding a panic display?
00:44:35pamauryyou can't
00:45:03pamauryI think bisecting is the only option
00:45:31chrisjjbisecting... what?
00:47:25__builtinbinary search through commits to isolate the introduction of a regression
00:48:27pamaurychrisjj: can I send you another build to try and narrow down the problem ?
00:49:21chrisjj__builtin: note: "official dev builds do not include lcd_fix" without which RB is unstable on ZEN, so I have very few commits to bisect :-)
00:51:11pamaurythe lcd fix only affects the bootloader
00:51:16pamauryat least that should be the case
00:54:22chrisjjGot it...
00:55:35 Quit xorly (Ping timeout: 240 seconds)
00:55:36chrisjjAnd I see a change in the start-up flashes.
00:55:38 Join lebellium_z3c [0] (
00:56:02lebellium_z3cpamaury: I have sound!!
00:56:54chrisjj7e0820fe2 crashes on the first backlight sleep.
00:56:54lebellium_z3cBut totally crappy
00:57:43lebellium_z3cLike the max output at any volume
00:59:28chrisjj7e0820fe2 - runs #1 and #2 crashed on the first backlight sleep. Runs #3 and #4 has survived backlight sleep.
00:59:44 Quit lebellium_z3c (Client Quit)
01:00:50 Quit shamus (Read error: Connection reset by peer)
01:04:36pamauryI thought the back sleep problems were long gone
01:06:43chrisjjCertainly on 45697a0bf-161212 (lcd_fix) I had no backlight sleep crashes.
01:07:01pamauryok, let's stop there and resume tomorrow, I'm tired
01:07:35chrisjjThey subsequently started only on be68b6a7bM-170108.
01:07:45pamauryI see two options: 1) I introduced a bug since 45697a0bf 2) somehow the lcd_fix also "fixes" the backlight sleep problem but I have no idea how it's possible
01:08:09pamaurybecause the lcd_fix is a one-time thing on boot
01:08:14chrisjjOne think before you go if you don't mind. Am I correct in saying the master has no lcd_fix version?
01:09:06chrisjjOK. "I introduced a bug since 45697a0bf " could be one of those changes that makes no difference to ZEN :-) but I think it unlikely.
01:09:15pamauryI really should push it, I just haven't tried it on the other imx target to make sure it introduces no regression. Because the way I understand it, the lcd_fix is only supposed to fix something in the bootloader
01:09:36chrisjjBut lcd_fix included .rockbox update
01:10:09pamauryyeah but it should make no difference, or at best it should only prevent a possible (but very very unlikely) white screen of death on boot
01:10:41pamauryI think we have to test 45697a0bf without lcd_fix to be sure
01:10:50pamaury(ie just 45697a0bf)
01:11:47chrisjjThe lcd_fix you gave me says "Version: 45697a0bf-161212"
01:12:04pamaurysorry, the commit right before it
01:12:18pamauryI'm tired
01:12:25pamaurygood night
01:13:09chrisjj"cd8b33327c9" Ah, got you. I'll see if have/can get that. Night-night.
01:16:50 Quit pamaury (Ping timeout: 240 seconds)
01:18:15chrisjj__builtin: Can you help me find this past ZEN cd8b33327c9 build on
01:22:26__builtinNo, I'm busy.
01:30:35 Quit Senji_ (Ping timeout: 240 seconds)
01:31:04chrisjjOK. Anyone else?
01:37:08 Quit WakiMiko (Max SendQ exceeded)
01:37:41 Join WakiMiko [0] (~WakiMiko@unaffiliated/wakimiko)
01:37:44 Join b-t` [0] (
01:38:02 Quit Bilgus (Remote host closed the connection)
01:38:04 Quit b-t (Write error: Broken pipe)
01:59:04 Quit elensil (Ping timeout: 256 seconds)
02:13:23 Join elensil [0] (~edhelas@2001:1c02:1903:d800:49e3:50f1:371b:c023)
02:14:45 Quit skapazzo (Quit: leaving)
02:23:44chrisjjIs WPS supposed to show the charge level when charging? Here (ZEN cabbieV2) it doens't.
02:38:38***Saving seen data "./dancer.seen"
03:19:30 Quit ZincAlloy (Quit: Leaving.)
03:23:35 Join o43 [0] (
03:38:10 Quit o43 (Ping timeout: 240 seconds)
03:47:32 Join o43 [0] (
03:54:06 Join Bilgus [0] (
04:28:55[Saint]what resolution is the ZEN?
04:28:58 Quit o43 (Ping timeout: 260 seconds)
04:29:35[Saint]No, lets do it a different way - what do you expect to happen, chrisjj?
04:30:16[Saint]it should show that a charger is connected, and the battery meter should reflect this. If you're expecting a numeric value, then, no.
04:31:08[Saint]this is handled by the .sbs and has nothing to do with cabbiev2 directly other than the fact that cabbiev2 uses the fallback statusbar.
04:37:03 Join o43 [0] (
04:37:59[Saint]Oh, hmmm, sorry, My mistake. I forgot cabbiev2 disabled the .sbs
04:38:42***Saving seen data "./dancer.seen"
04:38:58[Saint]So the answer is no.
04:40:09__builtin[Saint]: do you have a PortalPlayer iPod you can test something on?
04:40:52[Saint]It'll show if a charger is connected and not charging, is connected and charging, or the battery state.
04:41:04[Saint]__builtin: several, but none on my right this very second.
04:41:14[Saint]what do you need tested?
04:41:43__builtinhow slow an antialiased line function of mine is
04:42:12__builtinturns out the PP devices are the slowest devices (by cpu freq, at least) that have a color screen
04:44:49[Saint]MHz does not speed make.
04:45:11__builtinyes, I know
04:45:25__builtinhow else should I attempt to quantify speed then?
04:46:33[Saint]the only adequate and objective metric in my opinion would be the codec performance comparison benchmarks.
04:47:02[Saint]There you get a sample size of targets performing a known operation.
04:47:41__builtinwell, I can't easily run queries against those (i.e. which are the slowest ones with a color display)
04:48:19__builtinso cpu frequency will have to do ;)
04:48:22[Saint]Why not?
04:48:28[Saint]Oh. You said /easily/
04:49:00__builtinyeah −− with cpu frequency anything I need is just a grep and awk away
04:49:02[Saint]Meh. You know which targets have colour displays, and the output is standardized. You could form a nice little table with some grep magic.
04:50:41[Saint]you don't need to care about all the codec results as long as you pick one that is consistent amongst the targets you're interested in.
04:51:45[Saint]frequency falls over, I think, because on the surface it's 80MHz but you've also got the coprocessor to think about so it's realistically dual-core.
04:52:11__builtinwell, in this case it's a single thread doing all the work
04:52:14[Saint]though there's a lot of fundamental differences.
04:52:48__builtinbasically it doesn't *have* to be the slowest device, just slow enough to see if my algorithm is "too slow"
04:54:07__builtinI tried testing on one of lebellium's c200s, but horrible display makes it hard to really judge anything
04:55:17__builtinright now antialiased (Wu) lines take about 4x longer than Bresenham (or whatever rockbox uses)
04:59:23[Saint]serious question, at 220x176 to 240x320, does it matter?
04:59:36[Saint]is the difference even visible?
05:00:26__builtinactually, yes
05:00:38__builtinlet me get you some visual evidence
05:03:46[Saint]that's genuinely surprising.
05:06:34[Saint]I imagine 176x220 makes a significant difference.
05:09:41Bilguswow that is quite a difference
05:10:21[Saint]the problem is, though, it's in simulation and the pixel packing and DPI is completely different.
05:10:41[Saint]sim can't really tell you shit in this regard.
05:10:58 Quit o43 (Ping timeout: 260 seconds)
05:11:20*__builtin is waiting for [Saint] to test it on his device so he can see!
05:11:43Bilgus__builtin, what kind of algorithm are you using?
05:12:07__builtina fixed point adaptation of Wu's algorithm
05:12:07Bilgusis it simple interpolation?
05:13:28__builtinbasically it draws a 2-pixel wide line with different brightness values that add to unity
05:13:54__builtinand those values are dependent on how much "line" falls into each pixel
05:14:25__builtinwell, good night
05:15:07BilgusI'm a big fan of pre computed values
05:16:07BilgusI once did something similar with beizer curves for rounded controls
05:16:59 Join o43 [0] (
05:34:56 Quit o43 (Quit: Lost terminal)
06:30:23 Quit TheSeven (Ping timeout: 258 seconds)
06:31:02 Join TheSeven [0] (~quassel@rockbox/developer/TheSeven)
06:38:43***Saving seen data "./dancer.seen"
06:46:49 Join gs6 [0] (
06:46:54 Quit [Saint] (Remote host closed the connection)
06:47:06gs6Is Rockbox dead/dying?
06:47:43gs6There is no recent Android build, right?
06:48:20 Join [Saint] [0] (~sinner@rockbox/staff/saint)
06:59:45Bilgusthe wat I got it was there won't be a recent Android build either as they changed the threading model but there are new ports in progress
07:01:25Bilgus@ gs6
07:02:48[Saint]that first sentence is super awkward.
07:03:34[Saint]and, no, everything past Android 4.4.4 is dead.
07:04:01[Saint]you can compile for Android 5+, but it is an obscene hack.
07:04:18[Saint]honestly though, it's 2017.
07:04:32[Saint]wanting to run Rockbox on modern Android seems like masochism.
07:04:44[Saint]If you prefer, I could just come over and kick you in the balls.
07:04:50gs6Well, I don't know many decent music players for Android
07:05:02gs6Rockbox has features VLC doesn't have
07:05:10gs6And it's faster
07:05:43Bilguswat = way :)
07:06:14[Saint]But, yeah - perhaps you're looking at it through rose tinted glasses.
07:06:33gs6It would be cool if there was a new sort of project that's just a music player without support for embedded devices
07:06:52gs6'Cause all the audio features of Rockbox are really cool
07:06:58[Saint]well, we do run as an SDL app my man.
07:07:23[Saint]You'll need to compile it yourself, however.
07:07:58[Saint]for context, the takeaway from that is that it is exactly what you're asking for.
07:08:06gs6For Android??
07:08:20gs6Or just desktop?
07:08:24[Saint]Rockbox on generic x86/64, MIPS, ARM
07:09:00gs6Oh, I don't expect SDL Android support to be very good
07:09:18gs6Well, it's not in the Play store to try
07:09:24[Saint]There's no such thing, so, yes - you're quite right.
07:09:29[Saint]it wouldn't be very good.
07:09:53gs6I tried a Qt Android app and it was pretty garbage
07:10:08BilgusVLC for android has plugins now
07:10:12gs6Sadly, Android features are so tightly knitted to the UI
07:10:19[Saint]The reason why Rockbox never made it to the play store is multi faceted.
07:10:34[Saint]the primary reason is that we much target a specific resolution for the framebuffer.
07:10:46gs6Yeah, I remember reading that
07:10:55gs6Doesn't seem like a huge issue
07:11:00[Saint]So instead of a single app, there would be hundreds.
07:11:08[Saint]it's a very huge issue.
07:11:14gs6That's got to be like 5% of what Rockbox does
07:11:24gs6It's just the freakin' menu
07:11:59gs6But sorry if I'm way off base
07:12:06[Saint]Well, given that it _won't even boot_ if the application and host don't agree on the framebuffer in modern Android, it is a massive issue.
07:12:19[Saint]you are, it makes it sound like we're just lazy.
07:12:34[Saint]there's many very real and fundamental reasons this never happened.
07:12:59Bilgusmainly because saint is lazy :p i'm sure of it
07:13:20*[Saint] nods
07:13:51 Quit atsampson (Ping timeout: 240 seconds)
07:15:05[Saint]Technically speaking you could run a Debian armhf/el/eabi chroot in Android and run the sim or SDL app builds there.
07:15:19[Saint]but that's a hilarious chain of things just waiting to break.
07:24:33 Quit TorC (Ping timeout: 260 seconds)
07:35:28gs6It sounds like the core functionality and the UI are tightly coupled in Rockbox
07:38:23 Join TorC [0] (~TorC@fsf/member/TorC)
07:39:09gs6Another thing I'm curious about is how the decoding code compares to ffmpeg
07:40:01gs6I wonder if its smaller and more efficient since it had to run on embedded devices
07:42:10gs6I think it might be fun to refactor Rockbox and use the JNI to create an Android UI
07:42:40gs6But I don't know if it would be easier to start from scratch
07:42:52gs6Rockbox has a lot of cool features
07:43:08 Quit amiconn (Quit: - Chat comfortably. Anywhere.)
07:43:08 Quit pixelma (Quit: .)
07:43:19 Join pixelma [0] (~pixelma@rockbox/staff/pixelma)
07:43:20 Join amiconn [0] (~amiconn@rockbox/developer/amiconn)
07:44:35gs6I think there might be a licensing issue though since Rockbox is GPL, right?
07:44:59[Saint]It is, but no, there wouldn't be.
07:45:16gs6If you distributed an apk in the Play store, the user wouldn't be able to view the source
07:45:23gs6Or change it
07:45:27[Saint]So what?
07:45:38[Saint]These are not mutually exclusive things.
07:45:39gs6I think that might be against the GPL terms
07:45:58[Saint]Source msut be supplied to those distributed a binary in a reasonable format on request.
07:46:02[Saint]Not proactively.
07:46:07gs6Oh really
07:46:11[Saint]Really really.
07:46:18gs6Okay, that's good
07:46:48[Saint]fun fact: "reasonable format" is unreasonably (pun intended) vague.
07:47:00[Saint]Technically, you could print it out on A4 letterhead and mail it to them
07:47:21gs6Well, if it was on Github with compilation instructions I think that would be sufficient
07:47:37[Saint]Entirely so.
07:47:56[Saint]As for how Rockbox compares to ffmpeg in the coded sense - basically identically.
07:48:24gs6Does Rockbox predate ffmpeg?
07:49:13gs6Maybe it would be better to just port mpv to Android
07:49:28[Saint]Rockbox is about 16 years old now.
07:50:08gs6But Rockbox has playlist and audio features not in mpv
07:50:36[Saint]You might want to look at Warble/librockbox.
07:50:48gs6Oh, librockbox?
07:50:53gs6That sounds cool
07:50:57gs6Is that new?
07:50:57[Saint]It's in our source tree. It's a demo of a standalone playback library.
07:51:07gs6Very cool
07:51:08[Saint]No. It's about 6 years old.
07:51:21[Saint]Spawned from our last GSOC.
07:51:25gs6Was the Android Rockbox not built on that?
07:52:16gs6So maybe an Android UI and JNI bindings to librockbox could work
07:52:25[Saint]The Android port uses the same core code all the embedded ports do.
07:52:43[Saint]Honestly, you'd be better off just using android-ffmpeg.
07:53:14[Saint]it's infinitely more feature complete than librockbox. librockbox was just a proof of concept.
07:53:19gs6Does ffmpeg have equalizer stuff and other signal processing?
07:53:53gs6Yeah, I would love to have an android player based on ffmpeg
07:54:04gs6I guess VLC might do that
07:54:28gs6I just thought VLC for Android was crappy compared to Rockbox
07:54:35gs6Much slower and less intuitive
07:55:06[Saint]You're not wrong about the slower part.
07:55:13[Saint]Intuitive might be arguable.
07:55:27gs6I couldn't do some of the playlist stuff
07:55:31[Saint]I don't really think there's anything about our UI that's intuitive at all.
07:55:57gs6It has been a while since I tried VLC
07:56:00[Saint]It's the best we can do with the absurd amount of granularity we have though.
07:56:27[Saint]it's hard to make a UI that holds several hundred settings intuitive.
07:59:35gs6Just tried VLC
07:59:39gs6It's so garbage
07:59:53gs6I can't even explain the amount of shitty things going on
08:00:29gs6There's a tooltip that keeps popping up and it cover up everything
08:00:50gs6If I try to select VLC in the app switcher, it doesn't work and goes to my home screen
08:01:11gs6When I play one file in a directory, it doesn't add the rest to the playlist
08:01:21gs6It's laggy as af
08:01:28gs6laggy af
08:01:55gs6There are weird icons that don't say what they do
08:02:52gs6The Rockbox UI is just a straightforward list. Sure, there are a lot of submenus and you have to backtrack a bit, but at least it's not doing annoying or unpredictable stuff
08:03:48 Join ender` [0] (
08:05:04gs6Like "playback speed" is an icon of a dude running
08:06:13[Saint]The one obvious benefit with VLC is HW accelerated decoding on supported targets. So no matter what it's always going to be infinitely more efficient.
08:08:08gs6Well, I must be lacking it lol
08:08:22gs6I have a rather old phone and rockbox is much smoother
08:09:03gs6The UI of VLC is not so bad as I mess with it
08:09:14gs6Starting to figure it out
08:12:51gs6I don't see an option to set the file browser root
08:15:15 Quit Moarc (Quit: i znowu NADMUCHAŁ BALONA)
08:25:26 Join Moarc [0] (
08:38:47***Saving seen data "./dancer.seen"
08:39:34 Join wodz [0] (
08:39:37 Quit wodz (Client Quit)
08:39:56 Join wodz [0] (
08:46:32 Join petur [0] (~petur@
08:46:32 Quit petur (Changing host)
08:46:32 Join petur [0] (~petur@rockbox/developer/petur)
08:51:30 Join JdGordon_ [0] (
08:51:30 Quit JdGordon_ (Changing host)
08:51:30 Join JdGordon_ [0] (~jonno@rockbox/developer/JdGordon)
09:13:05 Quit TorC (Ping timeout: 240 seconds)
09:14:08 Join TorC [0] (~TorC@fsf/member/TorC)
09:39:21 Quit TorC (Ping timeout: 245 seconds)
09:50:24 Join xorly [0] (
09:58:59 Join pamaury [0] (~pamaury@rockbox/developer/pamaury)
10:06:31pamaurychrisjj: I'll send you a build for cd8b33327c9
10:19:36 Quit krnlyng (Ping timeout: 258 seconds)
10:30:51 Quit elensil (Ping timeout: 240 seconds)
10:32:16 Quit pamaury (Ping timeout: 245 seconds)
10:32:52 Join krnlyng [0] (
10:37:17 Quit Tristitia (*.net *.split)
10:37:17 Quit benedikt93 (*.net *.split)
10:37:34 Join benedikt93 [0] (~quassel@unaffiliated/benedikt93)
10:38:46 Quit scorche (*.net *.split)
10:38:47 Quit mc2739 (*.net *.split)
10:38:47 Quit K1773R (*.net *.split)
10:38:49***Saving seen data "./dancer.seen"
10:39:06 Quit prg318 (*.net *.split)
10:39:06 Quit GodEater (*.net *.split)
10:39:06 Quit rogeliodh (*.net *.split)
10:39:09 Join mc2739 [0] (~mc2739@rockbox/developer/mc2739)
10:39:19 Join GodEater [0] (~whoknows@
10:39:19 Quit GodEater (Changing host)
10:39:19 Join GodEater [0] (~whoknows@rockbox/staff/GodEater)
10:39:29 Join scorche [0] (~scorche@rockbox/administrator/scorche)
10:39:48 Join prg318 [0] (~prg318@deadcodersociety/prg318)
10:40:12 Join Tristitia [0] (
10:40:25 Join K1773R [0] (~K1773R@unaffiliated/k1773r)
10:41:26 Join rogeliodh [0] (
10:48:58 Join elensil [0] (~edhelas@2001:1c02:1903:d800:611e:2ad2:8fd2:c667)
10:49:56 Quit xorly (Ping timeout: 252 seconds)
10:54:00 Join xorly [0] (
10:54:30wodzpamaury: beep_test opens /dev/icx_beep and then calls a few ioctl() ICX_SOUND_BEEP (0x620b4004) and ICX_BEEP_STATUS (0x6201c004)
11:10:51 Join parchd [0] (~parchd@unaffiliated/parchd)
11:21:50 Join TorC [0] (~TorC@fsf/member/TorC)
11:29:32 Quit xorly (Ping timeout: 252 seconds)
11:36:15 Join Piece_Maker [0] (
11:38:18 Quit Acou_Bass (Ping timeout: 260 seconds)
11:38:18 Nick Piece_Maker is now known as Acou_Bass (
11:41:53 Join xorly [0] (
11:42:21 Join robertd1 [0] (~root@
12:20:33 Join pamaury [0] (
12:20:33 Quit pamaury (Changing host)
12:20:33 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
12:23:31 Join Mihail [0] (25d4b372@gateway/web/cgi-irc/
12:26:18Mihailwodz: I want add scaling for CVDD2 on AMSv2 (,51616.0.html). But at low values of CVDD2 internal flash stop working. If I revert voltages to normal - it accept CMD0, CMD8 (response 0x1AA) but stop with timeout on ACMD41 (response 0x40ff8000). I try power off (MCI_PWREN) but internal flash and sd card always
12:26:18Mihailwork, so probably both powered directly to CVDD2. Are you have idea how reset internal flash controller or what should I check/try?
12:26:42wodzI am not amsv2 guy, sorry
12:26:47wodzMihail: ^
12:29:31Mihailbut you know well sd spec :) way to reset sd card without power off?
12:30:21 Quit holyheels (Read error: Connection reset by peer)
12:31:47wodzMihail: Other then usual cmd0 and 74+ clock cycles - no
12:35:59 Quit gs6 (Read error: No route to host)
12:36:01MihailI already try it, also try CMD0 with argument of 0xFFFFFFFA and 0xF0F0F0F0 (from emmc spec) - no success :(
12:37:40MihailIf you have time can you review ?
12:38:51***Saving seen data "./dancer.seen"
12:39:07 Quit xorly (Ping timeout: 258 seconds)
12:40:34wodzMihail: Are you sure you want to panic on failed sd transfer?
12:42:56MihailI want to remove this checking at all - count > 0 have no sense. But first check if we have this case.
12:43:18Mihailcount < 0
12:43:55wodzMihail: I mean line 776 of gerrit task
12:45:05wodzaaa, it is just sanity check so it may make sense
12:48:27Mihailprobably I should change it to "count < 1" as count == 0 also have no sense
12:49:16 Quit Ruhan (Quit: Connection closed for inactivity)
12:53:11 Join atsampson [0] (
13:02:10wodzpamaury: radio_test binary uses some 'private' IOCTLs. According to the strings this are to access tuner registers, tuner GPIO2 ???, and perform reset.
13:03:06wodzpamaury: There are some standard IOCTLs as well, namely VIDIOC_G_TUNER and VIDIOC_S_TUNER
13:04:19wodzpamaury: Ah, and VIDIOC_QUERYCAP
13:08:20 Join holyheels [0] (
13:09:50 Join xorly [0] (
13:15:22 Join ZincAlloy [0] (~Adium@2a02:8108:8b80:1700:6580:4f45:dca9:4816)
13:17:39wodzpamaury: 'private' IOCTLs are of type 'v' instead of v4l2 'V'
13:27:41 Quit fulon (Ping timeout: 245 seconds)
13:29:18chrisjjpamaury: FYI, the last run of 7e0820fe2 (on the unit (Unit Q) where some previous runs crashed) has gone 10h with no crash.
13:30:28chrisjjAlso be68b6a7bM-170108 (memDFS) on a second unit (Unit G) has run with no crashes - 8h so far.
13:31:17chrisjjSo, mysterious variation.
13:35:36 Quit parchd (Ping timeout: 245 seconds)
13:49:21 Join mutnai [0] (6db90a3e@gateway/web/freenode/ip.
14:00:09 Join skapazzo [0] (~skapazzo@
14:01:35 Join Ruhan [0] (uid76353@gateway/web/
14:11:37 Join paulk-blaze [0] (
14:26:01 Quit xorly (Ping timeout: 245 seconds)
14:32:50pamaurychrisjj: that's weird
14:33:13pamauryit would interesting to bisect on the unit where it crashes though, to see if we can find a particular commit responsible
14:34:03wodzpamaury: sony's kernel module definitely supports 'v' IOCTLs but I can't find definition for this in sources
14:34:22pamaurywodz: not all ioctls are in the headers unfortunately
14:34:39pamauryit's good to know there exists ioctl to access registers directly though
14:34:59pamaurythis means that we can either try to use v4l2 interface or directly the tuner and reuse some of our drivers, depending on what is best
14:35:18wodzpamaury: as well as it seems registers are exposed through /proc
14:35:29pamauryah yes I remember that
14:35:43pamaurywodz: can you try to document this somewhere on the wiki ? Like SonyNWLinuxTuner ?
14:36:23wodzdirect register access will have the benefit of bypassing region and dependence on nvp thing
14:37:45pamaurygood point
14:38:02pamauryhowever we are not sure what tuners are in use
14:38:15pamauryI think it would be a good idea to gather a list of components for each player
14:38:40pamaurythe soc is easy but for dac and tuner, it's
14:38:43mutnaiyou could go for sony nw-s10 series instead of nwz-e580. it supports FLAC , NC headphones included and 77 hours of battery, but no microSD slot.
14:38:54***Saving seen data "./dancer.seen"
14:38:56mutnaifor Yogurt
14:39:22wodzThe clean way would be to use standard v4l2 interface and let sony handle the burden
14:39:47pamaurymutnai: there is no 'instead', I'll support as many as I can. However I don't have a nw-s10, I would need at least someone with the device to run things on it for me
14:39:53pamaurywodz: agreed
14:40:29wodzpamaury: The only thing I am not sure is how 'compliant' sony's driver is
14:43:43chrisjjpamaury: Weird agreed.
14:46:31chrisjjCan we make PANIC send info somewhere apart from the screen? E.g. to \panic.txt? To USB? Even to audio out!
14:46:55pamaurymutnai: also the S10 is quite expensive
14:47:28pamaurychrisjj: not really, but we don't know for sure if it's a panic
14:47:37pamaurydoes the WEN have a LED ?
14:47:44mutnaisorry... this was for Yogurt, for yesterday discuss )
14:48:01chrisjjThe ZEN does have a LED.
14:48:02pamaurychrisjj: what I could do, for debug purposes, is to make backlight blink on panic
14:48:09pamauryand/or the LED
14:48:14pamaurythat way it's easy to tell
14:48:21wodzhell, ioctl code in this sony's radio driver is convoluted :/
14:48:52chrisjjI'd guess backlight blink on panic will be visible regardless of how messed up the LCD is.
14:49:05chrisjjLED blink on panic would certainly be visible.
14:49:54 Quit paulk-blaze (Remote host closed the connection)
14:50:06pamaurychrisjj: ok, I'll look into it tonight. LED blinking on panic sounds like a good idea in general on target that have a LED
14:50:18 Join paulk-blaze [0] (
14:51:46chrisjjIs the current LED behaviour due to .rockbox? Or some hardware process?
14:53:02pamauryit's controllable by software
14:53:30pamauryon the ZEN X-Fi it's basically a red-green led controlled by two PWN. On LED I think it's a (blue ?) LEd controlled by a PWM
14:53:48pamaury*On ZEN I think it's a blue LED
14:54:03 Quit fs-bluebot (Ping timeout: 258 seconds)
14:55:05chrisjjIndeed I have only even seen the ZEN LED show blue.
14:55:28 Quit bluebrother (Ping timeout: 248 seconds)
14:56:42chrisjjI know it is controllable by software, but I'm interested to if it is controll/ed/ by software when working under RB. I.e. it is RB itself that is turning the LED from dim to bright on start-up, and off on shut-down?
14:56:55pamauryCurrently rockbox doesn't use LEDs, because I don't know what to do with them ^^
14:57:18 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
14:57:20pamauryI think Creative's bootloader send them to some state on boot and we don't touch them but we could
14:57:24pamaury*set them
14:57:38chrisjjOK, as I suspected. So I take it the LED behaviour under RB is actually due to hardware e.g. the PWM running on auto.
14:58:19pamauryyeah but PWM can be changed, the point of PWN is that you setting a brightness setting and you let the hardware handle it
14:58:59pamauryhowever if you want blinking led, that requires software
14:59:14pamaurywhat would you use the LED for ?
14:59:51chrisjjI understand PWM.
15:00:26chrisjjWhat I don't know is what is causing the LED to go from dim to bright when RB starts.
15:01:01 Quit jhMikeS (Ping timeout: 245 seconds)
15:01:19wodzpamaury: unless you can control PWM freq. With low freq you can set fill factor to 50% and voila - you have nice blinking led
15:02:06pamaurychrisjj: Creative's bootloader as I said
15:02:59chrisjjThen what is causing the LED to go from bright to off (as I see) upon this mystery crash to power-off??
15:03:02pamaurywodz: ah yeah good point, I was more thinking about a slowing diming LED, indeed if you just want on/off, PWM can do it
15:03:45pamaurychrisjj: maybe the device simply powers off completely
15:04:03wodzon panic you don't need to be fancy with dimming
15:04:28chrisjjpamaury: OK, then that leaves the question of how a RB crash could power off the device completely.
15:04:49pamaurythe watchdog can do that
15:05:29pamauryOn imx233 I enable the watchdog. If the device crashes really hard (typically in interrupt context), the watchdog could trigger and reset the device
15:05:38chrisjjAh, right...
15:06:00chrisjjAnd "reset the device" == put into the power-off state?
15:08:26 Join fs-bluebot [0] (
15:09:14pamaurychrisjj: I don't remember, I need to read the datasheet. I think it's a reset, so it reboots but then the bootloader notices that the user is not holding the power button so it disregards the boot and powers off
15:09:53chrisjjA real PITA for debugging!
15:10:25chrisjjCan I ask: why is reset-on-watchdog in use?
15:10:53wodzgood practice?
15:12:18pamaurygood practice, if you reach this your device in unrecoverable anyway
15:14:00chrisjjGood practice /on stable commercial software/.
15:14:11chrisjjCan you think of any advantage on RB currently?
15:14:51pamaurycan you think of any disadvantage ?
15:15:08pamauryit the reset button is not accessible that's useful
15:17:55 Quit paulk-blaze (Remote host closed the connection)
15:18:15chrisjjI can think of a big disadvantage.
15:18:20 Join paulk-blaze [0] (
15:19:07chrisjj"your device in unrecoverable anyway" but the PANIC info on screen is not unrecoverable. It stays... unless a watchdog reset wipes it.
15:20:30chrisjjreset on watchdog can make sense in release software on devices without accessible reset button, but on ZEN the reset button is accessible by paper clip.
15:22:08 Join pamaury_ [0] (~pamaury@rockbox/developer/pamaury)
15:22:23pamaurychrisjj: you don't understand
15:22:31pamaurywatchdog won't trigger on panic
15:22:40chrisjjPlease clarify :-)
15:22:41pamaurywatchdog triggers if the entire system freezes
15:23:08pamauryif watchdog wasn't enabled, your device would go unresponsive, no panic, no thing
15:23:46chrisjjOnly because if the software so chooses.
15:24:28wodzpamaury: How do you 'feed' watchdog?
15:24:28pamaurythe software doesn't choose to crash
15:24:46chrisjjSoftware can choose to use the watchdog to trigger reset... or choose not to.
15:25:11wodzchrisjj: That is not the purpose of watchdog. The reset is side effect.
15:25:16pamaurywodz: I reset the watchdog on every timer IRQ iirc
15:25:28pamaurychrisjj: a watchdog only does reset
15:25:55chrisjjUnderstood. Still it is the software's choice.
15:25:57pamauryyou cannot detect the watchdog triggering because by definition it means your software is malfunctioning
15:26:09chrisjjUnderstood. Anyway, I'm not trying to persuade you to change the RB choice.
15:26:22wodzpamaury: That what I thought. You shouldn't do that in IRQ. It may happen that firmware enters infinite loop with irqs on. This will not trigger watchdog.
15:26:23pamaurychrisjj: let it put differently: how does software detects that it is not working ?
15:26:57chrisjjHey, you don't want to ask that of me. I'm an honours Computer Scientist :-)
15:27:13pamaurywodz: not sure what you mean. If the softwarre gets stuck in IRQ context, it will trigger because it is only reset when IRQ is entered
15:27:23chrisjjI'm trying to understand if there's some advantage in RB choosing to engage a watchdog to give reset on freeze.
15:27:33pamaurychrisjj: what else can you do ?
15:27:34 Quit pamaury_ (Ping timeout: 240 seconds)
15:27:46pamaurythe only other option to let it freeze
15:28:08wodzpamaury: I mean if software is stuck in main loop BUT irqs are unmasked it will fire indefinitely and hence watchdog will be happy
15:28:11chrisjjYou can do nothing, let the s/w freeze and (hopefully) read the valuable PANIC info from the surviving LED display.
15:28:48pamaurywodz: I know but there is no easy way to detect that the 'userland' is frowen
15:29:01pamaurychrisjj: THERE IS NO PANIC ON FRICKING FREEZE
15:29:21chrisjjUnderstood. But is there a freeze after PANIC??
15:29:29wodzpamaury: Feed watchog in some thread (UI?)
15:30:04chrisjjpamaury: recall what we're seeing. The device powered down after (what you suggested might be) a PANIC.
15:30:45pamaurychrisjj: then there is no panic, software gets stuckm, watchdog triggers and powers off. The panic screen disabled the watchdog obviously
15:31:56chrisjjAh, if you're sure panic screen disables the watchdog, then I see no way the power-down state can be hiding a PANIC as (I think) you suggested.
15:32:24pamaurywodz: I knwo but that means we need so place in apps/ that would do that
15:32:51chrisjjFAOD, this is assuming "panic screen disabled the watchdog obviously" means disabled such that there will be no watchdog expiry and hence no reset cause by watchdog.
15:32:56pamaurychrisjj: you mentioned the screen was dark, it was not clear if it was a power-down
15:33:31pamaurychrisjj: you claim the LED is off but you only said that after I suggested it. And inddeed the LED being off seems to indicate it's a real power down
15:33:36chrisjjI actually said also that the state reponds to power press by entering power-up.
15:33:52pamauryyes but that's no indication
15:34:02pamaurybecause the panic screen will also reboot if you hit power
15:34:16chrisjjHow so? I know no other state the reponds such.
15:34:40pamaurywodz: maybe we could have a 'software watchdog' that resets on thread switch ?
15:35:00wodzpamaury: ?
15:35:19pamauryto detect when a thread is stuck
15:36:02pamauryyou setup a timer of expire after 10sec, and reset the timer on every thread switch. If the timer expires, panic
15:36:22wodzshould work
15:36:29chrisjjpamaury: And does the panic screen turn off the LED?? I thought not.
15:37:56wodzpamaury: The canonical way to do that actually is that each thread sets bit in guard variable. When timer expires guard variable is checked if all bits are set. If not trigger some action (panic, reset, whatever)
15:38:35pamaurythat would mean nontrivial changes in the threading code though no ?
15:39:00wodzWho said this is trivial change :-)
15:39:02 Quit mutnai (Quit: Page closed)
15:39:10wodzembedded development is non trivial
15:39:31pamaurygenerally speaking having such a mecanism in rockbox in general would be nice
15:39:49pamauryit can use the tick task as a timer
15:42:29chrisjjpamaury: PANIC screen caused by the ZEN USB issue does /not/ turn off the LED. I'll continue to assume no PANIC turns off the LED, unless you advise otherwise.
15:44:42wodzpamaury: We do check stack guard on context switch. This shouldn't be too hard to extend to check/set guard bit as well.
15:45:17pamaurywodz: how does this work exactly ? Each thread has a variable and what/when set/checks the bit
15:46:01pamauryeach thread sets its bit when it is switched to and regularly one checks if the bits are set and clear them, panic if on bit is not set ?
15:46:48wodzpamaury: Yes. There is one guard variable however. Like bitfield.
15:54:27 Quit wodz (Quit: Ex-Chat)
15:56:26 Quit rela (Ping timeout: 245 seconds)
15:58:39chrisjjpamaury: I confirm that the memDFS fail had LED off. Re your "you claim the LED is off but you only said that after I suggested it.", if you doubt my report, no problem. You should be able to confirm it on the device you have. If you need another device from me, just ask.
16:01:12chrisjjSo, the fail state looks to me 100% like a power-off. And you've said the angered watchdog does reset to power-off. So I think it reasonable to conclude that this fail state is due to watchdog angered by e.g. freeze.
16:05:28chrisjjRe your 'what would you use the LED for ?', in PANIC I would blink it ... /if/ that could be done in hardware and no continued software execution, but suggests to me not (<< wodz).
16:06:54pamaurystmp35xx does not apply, you want to look for the stmp37xx datasheet, and yes it can be done
16:07:00 Join fulon [0] (~vshyba@
16:08:53chrisjjI looked an could not find the datasheet for the ZEN's STMP3760. I did find your wiki note of differences, and saw no mention of any difference in PWM.
16:09:40 Quit MrZeus (Ping timeout: 256 seconds)
16:10:19pamaurychrisjj: stmp3760 falls into the scope of stmp3700
16:14:16chrisjjThanks. Same applies. I don't see enough division for hardware LED blink. But if you say it can be done in hardware, I'm sure you're right :)
16:15:39chrisjjIf you can add LED blink during PANIC screen, that would be great. Perhaps encode some sub-info in the blink rate? Shame the hardware can't do morse code to spell out the PANIC text! :)
16:15:41pamauryOne uses PWM
16:16:21pamaurythe output of the PWN goes to the gate of a transistor that controls the LED input. By changing the frequency and duty cycle, one can dim and blink the LED
16:18:25chrisjjUnderstood 100%. But to make it /blink/, one needs a long divider... which I don't see on this chip.
16:21:10pamaurychrisjj: there is plenty of room. The click is 24MHz, each channel has a local divider (CDIV) in range 1-1024 and then the PERIOD (measured in divided clock) can range from 1 to 65525
16:21:49pamauryso you can basically reach blinking ~1Hz, if my math is correct
16:22:59chrisjjI make it 0.3Hz. Great!
16:25:35chrisjjThen seriously I think you could get three eye-resolvable bits of PANIC info in the pulse rate and width.
16:29:20 Join Senji_ [0] (~Senji@
16:34:37 Join xorly [0] (
16:35:05chrisjjIf LED panic is going to go in core RB for users generally, perhaps the easiest encoding would be just rate, measured by eye by counting the number of pulses over e.g. 10secs.
16:35:59pamaurythat seems way overkill to be honest. A blinking led is a good start
16:37:56chrisjjFari enough. Esp. since I predict we're not going to see a blink in this case. We'll see dark - because the device is powered down.
16:38:11chrisjjBTW, is there a PNIC counter stored anywhere?
16:38:56***Saving seen data "./dancer.seen"
16:43:44 Quit paulk-blaze (Ping timeout: 248 seconds)
16:48:58 Join paulk-blaze [0] (
16:50:34 Quit paulk-blaze (Remote host closed the connection)
16:54:19 Join paulk-blaze [0] (
17:26:54 Join aude_ [0] (6df78c7b@gateway/web/freenode/ip.
17:28:31 Join n3m9 [0] (
17:28:35aude_hi. I recently lost my Clip Zip in a jacket that was stolen (not surprising, there was a Clip Zip with Rockbox in it after all). I'm looking for a new device. any recommendations? :)
17:32:57 Join pamaury_ [0] (~pamaury@rockbox/developer/pamaury)
17:38:16 Join MrZeus [0] (~MrZeus@
17:40:01 Quit petur (Read error: Connection reset by peer)
17:43:12 Quit pamaury (Remote host closed the connection)
17:46:10 Join chrisb [0] (
17:46:51 Quit pamaury_ (Ping timeout: 240 seconds)
17:54:52 Quit paulk-blaze (Quit: Leaving)
18:38:35 Join JanC_ [0] (~janc@lugwv/member/JanC)
18:38:57***Saving seen data "./dancer.seen"
18:39:48 Quit JanC (Killed ( (Nickname regained by services)))
18:39:48 Nick JanC_ is now known as JanC (~janc@lugwv/member/JanC)
18:47:48 Quit aude_ (Quit: Page closed)
18:51:40 Join Senji [0] (~Senji@
18:53:31 Quit Senji_ (Ping timeout: 240 seconds)
19:16:01 Join lebellium [0] (
19:17:09 Join parchd [0] (~parchd@unaffiliated/parchd)
19:18:53 Join petur [0] (~petur@rockbox/developer/petur)
19:19:03[Saint]It really bugs that shit out of me that chrisjj chooses to argue with highly experienced embedded software developers based on nothing but the notion of how he believes things should work in fanciful chrisjj-land.
19:19:20[Saint]So much so, I'm very highly considering just removing the problem entirely.
19:19:48[Saint]Take that as fair warning my man.
19:21:22comradekinguThere is a difference between walking in nature and being out to feed the trolls
19:22:52chrisbwhat's happening?
19:23:06__builtinchrisb: not you
19:23:40chrisbchrisjj != chrisb
19:24:09__builtin[Saint] and a bunch of others are rather annoyed by chrisjj's, uh, "behavior"
19:24:23*chrisb just updated two old sansa e250's to the latest release
19:24:43 Part __builtin (" - Chat comfortably. Anywhere.")
19:24:47 Join __builtin [0] (~xray@rockbox/developer/builtin)
19:25:18[Saint]At this point I find it basically impossible to believe that chrisjj isn't going out of his way to antagonize.
19:25:57[Saint]It's either that, or being entirely socially inept, and needing to take a step backwards and reevaluate things drastically.
19:30:40 Quit Mihail (Quit: - A hand crafted IRC client)
19:40:13 Join paulk-collins [0] (
19:41:39comradekinguSeems he is just making conversation, what is the issue?
19:41:59 Quit n3m9 (Read error: Connection reset by peer)
19:44:02[Saint]I find it pretty hard to believe anyone could read the backlog and say that, but so be it.
19:44:36[Saint]One supposes a lot of it is contextually sensitive with reference to historical action.
19:52:49comradekinguIf you dont entertain the thought of being wrong, maybe you put too much faith in being right
19:54:11*gevaerts wonders how that statement is in any way relevant for #rockbox
20:12:55 Join pamaury [0] (~pamaury@rockbox/developer/pamaury)
20:27:22 Join johnb2 [0] (
20:32:35johnb2Database->Autoupdate: am I right assuming, that this is done only once after boot or unplug from USb?
20:35:13johnb2Mihail: as the forum is down again, would you mind giving the link to the clip+ build here on IRC?
20:38:58***Saving seen data "./dancer.seen"
20:40:57[Saint]johnb2: you are both correct and incorrect.
20:41:08[Saint]There's another case, insertion of removable media.
20:44:14chrisbhow can i determine the max removable memory the sansa e250 can address?
20:45:49[Saint]By looking at the sd specification and knowing that it is 4TB and we'll literally never hit that density.
20:46:39[Saint]which is a long winded way of saying that any capacity you throw at it will work under Rockbox.
20:46:59chrisb[Saint]: ok, thanks
20:47:01[Saint]The OF may or may not have arbitrary limitations.
20:49:55[Saint]Original Firmware
20:51:22johnb2Mihail: never mind, Firefox remembered the URL:
20:55:46lebelliumpamaury: because of the standby mode, you can't get the boot selection screen once you used the OF -_- Same problem on yp-r0. We solved it by adding a reset function when you press power off several seconds. That way you don't need a pencil everytime to press the reset hole... I don't know if that would be possible on NWZ
20:55:52 Join wodz [0] (~wodz@
20:56:00johnb2thanks, and this is nothing relevant regarding power consumption on a flash based device, right?
20:56:13 Quit alucryd (Remote host closed the connection)
20:56:51pamaurylebellium: you can plug usb, no need for reset
20:57:08pamaurylebellium: I'm not sure it's possible to do that but I'll have a look
20:57:14lebelliumpamaury: I mean when you're not at home
20:57:18 Join alucryd [0] (~quassel@archlinux/developer/alucryd)
20:58:16pamaurylebellium: I know, I was teasing you ;)
20:58:26johnb2relevant = significant
20:58:28pamauryand mentionning that usb "solves it" if you can
20:58:38lebelliumpamaury: also I would display "Rockbox" first instead of "walkman" and make it the default selection after a short timeout
20:58:48pamaurywodz: ok, fmp doesn't seem very useful at first glance...
21:00:01pamaurylebellium: the bootloader will (when I implement it) remember the last user choice
21:00:09pamauryand only show the menu when switch is on
21:00:14pamauryat least that's my current plan
21:00:15lebelliumnice :)
21:03:00pamauryI think it is possible install another program on the device to monitor the power button and start it from sysv init
21:05:03chrisjjpamaury: FWIW, 7e0820fe2 (on the unit (Unit Q) where some previous runs crashed) has now gone 17h with no crash. And be68b6a7bM-170108 (memDFS) on a second unit (Unit G) has gone 15h with no crash.
21:06:54chrisjjSo even though resuls are no solid repeatable, they should give us enough battery_bench data to determine the memDFS saving, and perhaps enough crashed to determine the casue on a more-debuggable build.
21:07:24chrisjjpamaury: Is there a PANIC counter?
21:07:42pamauryno, why would there be ?
21:08:45chrisjjOK. Why? So after a reset we can see whether there had been a PANIC hidden by the failed LCD.
21:09:47pamauryI think you are a bit obsessed with the panic screen. May I remind you that the goal is to avoid panic and not counting how many of them you manage to achieve ?
21:10:18[Saint]Gotta catch 'em all, PANICmon.
21:10:24chrisjjThanks, I am duly reminded :)
21:10:38chrisjjAnd avoiding this panic is exactly what I'd like to do.
21:11:12chrisjjYou suggested this black screen is hiding a panic, but we don't know if it is. A panic counter would answer that.
21:11:35chrisjjHowever your blinking LED will answer it.
21:11:49[Saint]If you PANIC, there's no guarantee you can iterate the counter.
21:11:53[Saint]It's irrelevant.
21:11:54pamaurythe LED blinking would answer that. I think we first need to try to run the commit before the lcd_fix
21:12:05[Saint]You'd have to know you were going to PANIC before you did.
21:12:11[Saint]And if you knew that, you wouldn't...
21:12:26pamaurywe know that some recentish builds panic, and that the lcd_fix didn't, so the natural commit to try is the parent of the lcd_fix
21:13:14[Saint]the guts of what I'm saying is that there's no way to reliable implement a PANIC counter, even in a world where it was a rational thing to do.
21:14:09chrisjjThough it seems the panic cause is /subsequent/ for the lCD fix, no?
21:14:16 Join Bray90820_ [0] (
21:14:19chrisjjs/for the/to the/
21:14:40 Quit Bray90820 (Ping timeout: 248 seconds)
21:15:02 Quit johnb2 (Quit: Nettalk6 -
21:18:41pamaurychrisjj: I want to rule out the possibility that lcd_fix is actually fixing some crash (elsewhere than in the bootloader)
21:19:01pamaurywhen we confirm that, we can properly bisect between this commit (known working) and HEAD
21:19:31 Quit Bray90820_ (Ping timeout: 240 seconds)
21:20:23chrisjjAre you thinking that lcd_fix could have fixed a crash /and/ then that fix got busted by a subsequent commit?
21:23:50 Quit alexweissman (Remote host closed the connection)
21:23:57 Join Bray90820 [0] (
21:25:39pamaurythat's unlikely but you are the one that want to rigorously test everything usually
21:28:18[Saint]you mean to say you're not interested in the frequency and distribution of a PANIC that shouldn't happen anyway?
21:28:22[Saint] could you!
21:29:06*gevaerts objects to this meaningless capitalisation!
21:29:15gevaertspanic is not an acronym!
21:29:37__builtinPower ANd Interrupts Cancelled
21:30:02pamaurychrisjj: ok, so let's assume that cd8b3332 work (the parent of lcd_fix)
21:30:21chrisjjOK (I'd guess it will)...
21:30:52pamauryyour tests show that 7e0820fe2 and be68b6a7b have a problem right ?
21:31:47chrisjj(I'm still the one who wants to test everything rigourously! :-) )
21:33:12chrisjj(PAmauryNotInControl ;-) )
21:41:51 Quit parchd (Ping timeout: 245 seconds)
21:47:38chrisjjpamaury, I'll be happy to test the build before the lcd_fix, but I'm interested to know how this helps pinpoint this crash that's appeared /after/ lcd_fix.
21:50:12pamauryI'm looking at commits in the range cd8b3332..7e0820fe2 and none of them seems relevant for such a crash :-/
21:51:27pamaurylet me build 8e82839f and send it to you, just to be sure
22:01:42chrisjjThanks. Running OK so far on Unit Q.
22:02:56wodzpamaury: Do you have any idea what icx_sound_refclk0_change() can do?
22:02:59chrisjjFWIW, this reverts the start-up flash pattern to that of lcd_fix.
22:04:51chrisjjIn particular, the second appearance of the CREATIVE word again has a 1D screendoor effect in white. I.e. every second pixel row is solid white, with the CREATIVE word in white on black showing one the other rows.
22:06:07chrisjjThe flash pattern differs between units which appear identical hardware-wise, so goodness only knows the cause. This somewhat erodes precision in testing.
22:07:51 Join froggymana [0] (~frogs@unaffiliated/froggyman)
22:08:02 Quit froggyman (Ping timeout: 248 seconds)
22:08:06 Nick froggymana is now known as froggyman (~frogs@unaffiliated/froggyman)
22:10:49[Saint]My understanding is that there's at least two distinct LCDs used.
22:11:58pamaurywodz: no clue, but part of the icx sound driver is available in the linux code iirc
22:13:00wodzpamaury: anyway there is private ioctl which radio_test associate with SETFREQ which calls this function
22:13:31pamaurywell you could try to look in the em1 manuals if it matches something
22:14:13pamaurymaybe it enables/changes the frequency used by the tuner (some tuners don't use a crystal but let the soc generate the clock for them)
22:15:04 Quit chrisb (Ping timeout: 240 seconds)
22:15:05wodzpamaury: e47x don't use em1 AFAIK but something newer
22:16:36pamauryyeah but it's the evolution, that can still be interesting
22:18:02pamaurywodz: in arch/arm/mach-emxx/include/mach/icx_sound_fchg.h there is the function you mention
22:18:32pamauryand it is implemented in sound/arm/icx-beep.c
22:18:49wodzpamaury: yeah, just reading it
22:19:02pamauryit's not really obvious from the code what it does :D
22:20:52pamaurythe header has this comment:
22:20:53pamauryREFCLKO(MCLK) : 11.2896 [MHz] еее default
22:20:53pamauryREFCLKO(MCLK) : 10.7987 [MHz]
22:21:13pamauryand it definitely looks tied to tuner
22:22:28wodzpamaury: does si407x output digital or analog signal?
22:22:56wodzah, line level analog output
22:24:23wodzRCLK like of tuner needs 32.768kHz signal - maybe thats related
22:25:07chrisjjpamaury: 8e82839f has done bad on Unit Q. Screen black, audio silent, keys unresponsive, LED off. Any other observations you need?
22:26:03 Quit tchan (Ping timeout: 246 seconds)
22:30:26pamaurychrisjj: no, that's weird
22:30:36wodzpamaury: Here I documented somewhat my findings about radio interface
22:30:42pamauryso let me rebuild 8e82839f+lcd_fix
22:31:34 Quit robertd1 (Ping timeout: 240 seconds)
22:32:07 Join robertd1 [0] (~root@
22:35:27lebelliumwodz: may the FM tuner have some unused RDS capabilities?
22:37:06wodzlebellium: That depends which version of si407x they used actually. If this is si4072 then no, if it is si4073 then it supports RDS. Judging from driver disasm I think it may be si4073
22:38:53lebelliumis it Si470x like the Sansa Clip or really Si407x? I can't find much about the latter
22:38:56pamaurywodz: it could be that this rclk0 is internally multiplied (in the soc) by say 3
22:38:59***Saving seen data "./dancer.seen"
22:39:08pamaurygiving a very good estimate of the 32768kHz
22:39:45wodzlebellium: It is easy to check actually r1 has chip id field.
22:40:24wodzlebellium: what you mean about real Si407x?
22:41:04lebelliumI mean is Si407x a typo?
22:41:21lebelliumI can't find anything about Si4072 on google
22:41:34lebelliumSi470x is famous
22:42:47wodzah, yes a typo. si470x
22:43:28lebelliumclearer to me now
22:43:38lebelliumso it may have the same tuner as the Clip Zip
22:44:52wodzpamaury: Do you know if there is a way to trigger running radio_test with OF actually?
22:45:04lebelliumSony never offered RDS in OF but I never knew if it was an hardware issue or only a software matter
22:45:11wodzpamaury: like debug menu or something
22:45:22pamaurywodz: my guess is that radio_test is run by the service menu (which is a sort of debug menu)
22:45:33pamauryyou can reach via a magic key combination described in the service manual
22:45:33wodzpamaury: how to enter this?
22:45:47pamauryalternatively, the rockbox bootloader gives you an option to run it
22:46:13pamaurymanual list here:
22:46:26pamaurychrisjj: can you test the build I listed above ? (
22:46:26*wodz reading
22:49:38wodzHeh, I am too slow touching buttons. I need some practice :P
22:50:24lebelliumwodz: I've done it so much time on E580 and A10 that I almost know the key combo by heart now :D
22:50:30lebelliummany times*
22:51:07lebelliumnote that there was a mistake in the key combo of the E580 service manual
22:51:15lebelliumhopefully it's not wrong in the E470 one
22:51:54wodzhaha, how fancy
22:52:10wodzfloating bubbles made my day :P
22:53:44chrisjjpamaury: That latest (8e82839feM-170110) on Unit Q shows the same fail.
22:54:55pamaurychrisjj: unless I made a mistake, this is the same code as the original lcd_fix
22:57:21 Quit wodz (Quit: Leaving)
22:59:24chrisjjHow "same"? They aren't binary identical:
23:03:09pamaurywe don't have reproducible builds, so that's expected, but it's the same source code
23:03:31pamauryunless the original lcd_fix is not the commit that I think it is
23:03:48pamaurycan you just try and report ?
23:07:28 Quit Senji (Ping timeout: 252 seconds)
23:09:31 Join StaticAmbience_ [0] (
23:10:09 Quit StaticAmbience (Ping timeout: 246 seconds)
23:11:41 Join StaticAmbience [0] (
23:13:44 Quit petur (Remote host closed the connection)
23:13:48 Join jhMikeS [0] (
23:13:52 Quit StaticAmbience_ (Ping timeout: 258 seconds)
23:16:24 Quit lebellium (Quit: ChatZilla 0.9.93 [Firefox 50.1.0/20161208153507])
23:18:32chrisjjHang on. That's the same one.
23:19:43chrisjjI did try the build of your [21:46] and my [21:53] was the report (same fail).
23:20:27 Quit holyheels (Read error: Connection reset by peer)
23:21:17chrisjjBTW, that 'original lcd_fix' (the first successful LCD fix I know) is Version: 45697a0bf-161212.
23:22:52pamaurychrisjj: unless I am mistaken, 45697a0bf is 8e82839f + lcd fix (as an extra commit). Thus 45697a0bf is the (or should be the) same as
23:23:06chrisjjI hear you.
23:23:12pamaurynow I am worried because you are claiming that 45697a0bf doesn't have this backlight sleep crash but 8e82839f_lcdfix does...
23:23:49chrisjjYour [22:03] asks me to try the build that I already tried from [21:46].
23:24:54pamaurychrisjj: yes, I was unsure if you saw my message
23:25:07pamaurythen I realized I miss your "Doing..." message
23:25:12chrisjjOK, got you. I always check your messages before I write mine :-)
23:25:38chrisjjYes, I am claiming that 45697a0bf doesn't have this fail that 8e82839f_lcdfix does.
23:27:02chrisjjI wouldn't like to call it a backlight sleep crash because I don't know it is really that. I know only that the device goes to what appears to be power-off state (so obviously backlight is off) without good cause.
23:27:39chrisjjBTW, '45697a0bf doesn't have this fail' is based on hundreds of sessions on six devices.
23:34:11pamaurychrisjj: can you retest
23:34:16pamaurythis is the original lcd fix, afaik
23:37:12chrisjj(The zip is binary-identical to the original of 2016-12-12.)
23:40:07chrisjjYour [22:34] on Unit Q is has run OK for 52secs. Continuing...
23:44:56 Quit StaticAmbience (Quit: No Ping reply in 180 seconds.)
23:46:22 Join StaticAmbience [0] (
23:52:45 Part robertd1
23:53:39chrisjjNow 14min.
23:56:02 Quit xorly (Ping timeout: 256 seconds)
23:56:25chrisjjpamaury, I'm thinking angry watchdog is the only way we know of RB putting the device into power-off, barring power key press.
23:57:32chrisjjYou've said *PANIC* screen disables the watchdog, meaning I presume disables the reset effect of angry watchdog.
23:58:53 Quit chrisjj (Quit: Page closed)

Previous day | Next day