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 2021-09-03

01:20:34 Join lebellium [0] (~lebellium@2a01:cb10:2e:2000:3df2:389f:6929:6355)
01:35:55***No seen item changed, no save performed.
01:46:52 Join _bilgus [0] (~bilgus@
02:57:24 Quit lebellium (Quit: Leaving)
03:35:57***Saving seen data "./dancer.seen"
05:32:35 Quit blbro[m] (Ping timeout: 252 seconds)
05:32:47 Join blbro[m] [0] (~blbrostra@2001:470:69fc:105::8f7)
05:35:58***Saving seen data "./dancer.seen"
06:01:22 Quit hook54321 ()
06:01:52 Join hook54321 [0] (sid149355@user/hook54321)
06:04:21 Quit _bilgus (Quit: Leaving)
06:47:02 Join Moriar [0] (
07:21:03 Quit Moriar (Quit: Leaving.)
07:36:02***Saving seen data "./dancer.seen"
07:53:14 Join massiveH [0] (
09:04:52 Quit massiveH (Quit: Leaving)
09:36:04***Saving seen data "./dancer.seen"
11:36:05***No seen item changed, no save performed.
12:14:10 Join ZincAlloy [0] (~Adium@2a02:8108:943f:d824:442e:90cb:b9dc:2e1e)
13:36:08***No seen item changed, no save performed.
14:11:19speachya question for [semi-]active folks: Putting aside the question of migration, Any thoughts about moving from our mismash of services to a sourcehut instance?
14:13:23speachyI'm planning on setting up a private instance for my own use (to possibly replace gitolite+flyspray+mailman+dokuwiki+misc) and it occurred to me it might be a good fit for rockbox too
14:48:17__builtincan we really just "ignore" the issue of migration though?
14:48:56__builtinthere's a _lot_ of wiki pages and web miscellany that would have to get moved
14:49:34__builtinand in doing so, we'd (probably) break all the links pointing to everything we've ever touched
14:50:26__builtinjust my $0.02, of course
15:14:24 Quit tchan (Ping timeout: 256 seconds)
15:14:59 Join tchan [0] (
15:26:47braewoods__builtin: depends how it's done. if they move services as is it may preserve all that.
15:36:11***Saving seen data "./dancer.seen"
16:10:04_bilgus_my concern would be depending on it and they go belly up or price us out
16:11:02_bilgus_it says they are open source so does that mean we could host it ourselves if it had to be done or would it be a big ol mess?
16:11:52_bilgus_If priced out that would be the MO i'm sure −− big hassle to leave == more paying customers
16:13:21 Nick edward_ is now known as edward (~edward@user/edward)
16:16:28__builtinI (tangentially) know the guy that runs
16:16:48__builtinvery much a hardcore open-source guy like ourselves
16:19:06__builtinhe also created KnightOS, which is basically rockbox for graphing calculators
16:50:30braewoodsi would sooner trust someone that has actually contributed to the project over a random third party
16:51:47braewoodsspeaking of which i have available resources on my rented server
16:52:48braewoodsi can arrange something if asked to, a container or something perhaps
16:53:55braewoodsi have some public IPs available for container use
16:54:15braewoodswould be hosted on a Canadian SYS server
16:59:25munkismy 2 cent is that although I like I much prefer gerrit (although the js requirement is annoying.)
17:36:12***No seen item changed, no save performed.
19:34:09 Quit ZincAlloy (Quit: Leaving.)
19:36:13***Saving seen data "./dancer.seen"
20:00:03 Quit _bilgus_ (*.net *.split)
20:00:03 Quit Solanacean (*.net *.split)
20:00:03 Quit akaWolf (*.net *.split)
20:00:03 Quit dys (*.net *.split)
20:00:12 Join akaWolf [0] (
20:00:20 Join Solanacean [0] (solanacean@
20:01:55 Join _bilgus_ [0] (~bilgus@
20:05:29 Join massiveH [0] (
20:36:41speachy__builtin: oh, to be sure migration will be a real bitch regardless.
20:37:05speachybut my question was more about workflow & tools vs what we're using today
20:37:37speachyand self-hosting is eminently doable
20:37:59speachy(as I mentioned, I'm going to try and spin up an instance for my own use..)
21:13:12 Join borkitall [0] (
21:21:10speachy__builtin: it's probably more accurate to charactarize him as a hardcore free software guy. :D
21:21:44borkitallhey just wanna point out a bug introduced in daily 2021-08-13 (6de6e1459d) for the hifi walker h2
21:22:30borkitallthe changelog says it should remove clicking sounds, however it intoduces clicking sounds for me when using an external dac
21:23:04speachyuh.. rockbox doesn't support external dacs, unless I've really missed something
21:24:56borkitallokay maybe im calling it a different thing, im using a fiio e07k
21:27:16speachyrockbox opens the "default" ALSA device. If the DAC works at all, it's purely accidental. Rockbox has no ability to switch output devices and is hardcoded to the capabilities of the built-in DAC.
21:27:57speachyassuming you have this thing connected to the H2 via the USB port (rather than the analog AUX port)
21:30:27borkitallokay its an amp/dac, im using it through the aux port on the h2
21:36:16***Saving seen data "./dancer.seen"
21:37:20speachyok. so when do the clicks occur?
21:41:35 Join dconrad [0] (~dconrad@
21:44:04dconradI'm curious when clicking might occur, I definitely thought I had eliminated it in my testing
21:44:54dconradheadphone out I would think is quite a bit more susceptible than line out
21:57:41borkitalljust when playing music normally
21:58:38dconradnot play/pause actions, but /while/ music is playing?
22:01:29dconradis it a specific track?
22:03:17borkitallyeah, while playing music
22:03:59borkitallit happens on most songs i have, heavier ones like metal are more affected
22:04:25speachyvolume affects it at all?
22:04:55dconradok, do me a favor, go into sound settings -> maximum volume limit and turn it to -14 dB, does it go away?
22:05:18rb-bluebotBuild Server message: New build round started. Revision 3e52c7c734, 303 builds, 10 clients.
22:05:50speachyyeah, the H2's line out is _very_ hot
22:06:24dconradI'm wondering if it's tracks that are so hot a 1 sample offset could clip them
22:06:36dconrader, 1 "quantization unit"
22:07:37dconradquestion is, if that is indeed the case, are we overloading the input to the amp or is it the track is clipping the 32-bit range
22:08:57borkitallclicking is still present when changing volume
22:09:15dconradvia the maximum volume limit?
22:09:16borkitallchanging maximum volume limit to -14dB however seems to work
22:09:24dconradok, that's what I wondered
22:09:47dconradnormal volume doesn't have any effect when using line out
22:09:53dconradbut maximum volume does
22:10:07dconradok, so now increase it one step at a time. when does the problem occur?
22:10:50 Quit j-r (Ping timeout: 250 seconds)
22:12:35borkitallsound is fine up to -2dB, changing it to 0dB introduces clicking
22:12:44dconradyep, that's what I figured
22:13:21dconradso, I would say for the time being -2dB would be the solution, but I suppose we'll have to fine-tune it so it doesn't occur at 0dB
22:13:46dconradjust not quite sure how to do that
22:14:02borkitallyeah thanks for the help
22:14:14 Join Moriar [0] (
22:14:26dconradfor sure
22:16:56dconradis there any chance you could specify a track it occurs for you on? I thought I tested it with a full-scale sine wave at 0dB, but maybe a dynamic "actual music" track would let me reproduce it here
22:17:34rb-bluebotBuild Server message: Build round completed after 736 seconds.
22:17:36rb-bluebotBuild Server message: Revision 3e52c7c734 result: All green
22:24:32dconradlooks like we don't have a 32-bit version of clip_sample_16(), that would have been the easy solution
22:25:10 Quit _bilgus_ (Ping timeout: 240 seconds)
22:25:28 Join j-r [0] (
22:26:11borkitalltrack 3 from this album
22:26:28borkitallim using flac
22:26:39dconradok sweet, thanks
22:31:34dconradwow, yeah that track uses ALL the pcm range
22:33:28dconradpretty dope too, ngl
22:33:29speachyhmm. Given that we only actually care about the DC offset for "silent" samples, perhaps the DC offset hack should get pushed deeper into the PCM layer (eg the mixer?) so that it only affects "muted" output.
22:34:33dconradso like, limit it to samples below a certain amplitude?
22:34:58speachywe should only need it at a literal zero sample.
22:35:12dconradmakes sense, though I'd be wary that the affected/not affected transition would cause a problem
22:35:15speachyanything non-zero won't trigger the DAC's auto-mute/shutdown
22:35:37speachywell, a 0-1/64K transition ought to not be audible anyway?
22:35:49speachyer, 1/32K
22:36:24dconradhopefully, but I can't say I know one way or another
22:38:18dconradwhy would it be beneficial to move it to the mixer? we could, I would think, do the same logic in the volume code
22:38:39dconradif 0 do -1, otherwise unaffected
22:39:06 Join _bilgus [0] (~bilgus@
22:39:29speachyso a bandaid for the bandaid
22:39:39dconradthe other side of the coin is, we could try to handle overflow gracefully and clip it
22:39:58dconradwell, yeah, but the first bandaid is for something we don't have control over :-P
22:40:10speachy..or just make the max vol -1
22:40:36dconradhaha yeah that too
22:40:47dconradwell, -2 given the step size
22:44:08dconradthat does remind me though, we probably don't have any guarantee of 64-bit integer support on multiple targets, right? for example, the 16-bit volume code had a 32-bit integer to put the result into in case it needed clipping
22:45:29dconradnewer stuff I imagine is fine, but older, I could see it not being supported
22:47:37 Quit massiveH (Ping timeout: 245 seconds)
23:10:06 Join massiveH [0] (
23:10:22speachythe toolchain has int64_t (aka long long) but that's obviously going to be emulated
23:11:00dconradprobably not something you want performance-critical code using, in other words?
23:12:10dconradnot an issue at the moment, but I think it might be if we push high-bit-depth farther up into the mixer like you were thinking, because the volume code will (I think?) have to more resemble the original 16-bit method
23:14:40dconradunless we drop the samples down to 16-bit before scaling, which... defeats the whole purpose
23:31:53dconradwe could do "if already at max range, don't add offset", though that has the potential for artifacts, I think... or maybe it's so small nobody would realistically hear it
23:32:12dconradeh limiting the volume is probably the best way
23:36:17***Saving seen data "./dancer.seen"
23:48:22 Quit Moriar (Quit: Leaving.)
23:51:55_bilgusfigure this was done on the supported devices long ago using an oscilloscope I'm sure geaverts or saratoga could spill the gest I always get the two mixed up

Previous day | Next day