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 2015-10-09

00:00:58[Saint]The best I can do is assigning 0xABCD as the vendor ID and 0x1234 as the product ID.
00:01:20foolshwell HAVE_AGC built for the fuze+ lemme test it on target
00:01:49[Saint]For our purposes, it doesn't need to be a "real" (there's not really such a thing as a "fake" VID/PID) just needs to not be one that is recognized as an iPod.
00:03:00[Saint]TL;DR: most car head units, and all Apple Macs, see the USB VID/PID, see that it's an iPod, and flatly refuse to talk to the device in any other way.
00:03:32[Saint]Not having anywhere near complete IAP support makes that a very real problem.
00:03:38pamaurythere was a plan to claim a free usb ID some time ago
00:03:50[Saint]There was indeed.
00:04:07[Saint]It needed a Swede, or at least active coordination with a Swede, though.
00:04:43[Saint]And both of those things were difficult to get at the time, and they haven't been made any less difficult as time goes by.
00:05:34[Saint]Those of us interested in keeping the project alive need to have some rather important discussions about what to do with Rockbox.
00:05:52pamauryIt is not too late to apply
00:06:08pamauryyes I agree
00:06:17pamauryespecially about the infrastructure
00:06:33[Saint]No, indeed not. But we really need it to be one of the original founders, or at least have one of them in the chain of communication.
00:06:38[Saint]ANd, yes.
00:07:03[Saint]We need to separate ourselves from the total dependence on Haxx infrastructure.
00:07:23pamauryagree, the major question is how
00:07:55pamaurybasically I see either self-hosting or using free services like github or gitlab
00:08:04[Saint]I would be more than happy to host the build master.
00:08:42[Saint]And I think the only realistic option is moving to github for both versioning and regression/bug lodging.
00:09:32[Saint]The other issue, even though he swears it isn't an issue, is scorche hosting the forums entirely out of his own pocket.
00:09:41[Saint]That has bothered me for a loooooooooooooooong time.
00:10:40pamauryI am renting a server with a friend, we have plenty of power and space on it, I could run it I think
00:10:57[Franklin]the issue is whether it's long-term or not
00:12:00foolsh[Saint]: HAVE_AGC runs on fuze+ target, but I have to recode some music so I can test if it works
00:12:02pamaury[Franklin]: what exactly ?
00:12:14[Franklin]pamaury: is it a long term solution?
00:12:55pamauryI think so
00:13:18pamaury[Saint]: what was the reason to run the forum on a different server than the webserver by the way
00:13:29*[Franklin] thinks the project is too closely tied to haxx to break away completely
00:13:51[Franklin]maybe the website would continue to be hosted by them
00:13:51[Saint]pamaury: honestly, I have absolutely no idea.
00:14:21[Saint]Honestly. it's not tightly coupled to Haxx at all really.
00:14:22pamaurythere is nothing we cannot reproduce, we have all the scripts and so on, the only thing is the domain name really
00:16:07pamauryWe should probably discuss this on the mailing list ? Explain the plan, discuss the different possibitlies for each service
00:16:12[Saint]I would sincerely doubt that they would be unwilling to release the domain.
00:16:23pamauryI am sure the Swedes will be happy to help us
00:16:39[Saint]And, yes. I've had a mail regarding this partially drafted for a while now.
00:17:07pamauryI mean the gerrit situation made it obvious that we *need* to do something
00:17:24[Saint]Basically, I think that /anything/ that has the decoupling of the Swedes as an end goal would be eagerly accepted.
00:17:42pixelmaFor the Iaudios AGC doesn't make sense for mic recording at minimum because there are only two gain settings - and I believe for line-in the steps are still too coarse to have good results with AGC
00:17:52pamaurythere is will be technical questions on how exactly we migrate some services/bugs/patches and so on
00:18:01pixelmaSo it's not only software related
00:18:45pixelmaJust as an example
00:19:24pamaurypixelma: how fine grained is the gain control on the iriver ?
00:22:12pixelmaFor mic it's only 0dB and 20dB, line-in is in 1.5dB steps from -34.5 to 12 dB
00:23:31[Saint]IS there some technical reason as to why the mic gain isn't able to be user controlled more finely?
00:23:49[Saint]Or is it just some lazy-ass define that could be changed trivially?
00:24:27pamaury[Saint]: usually mic gain is hardware controlled and on most target is it very coarse
00:24:53pixelmaNo... as pamaury says
00:25:01pamauryon the fuze+ I think you have four settings like 0dB, 10dB, 20dB and 40dB
00:25:41pamaurydo we support different mic vs line-in gains ?
00:25:50[Saint]Doing it in software after the fact would be too expensive?
00:26:00[Saint]Not worth it?
00:27:02pamauryI guess not worth it but I'm no audio expert
00:27:07pixelmapamaury: from what I see on my M5 - yes, and see above
00:27:32 Quit pamaury (Remote host closed the connection)
00:27:47 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
00:31:04[Saint]prof_wolfff: debug builds on ipod6g appear to be broken
00:31:47[Saint]prof_wolfff: a bunch of errors in file_internal.c and file.c|o
00:32:35[Saint]just looks like type missmatches.
00:34:28[Saint]prof_wolfff: (logs)
00:37:22 Join ulmutul [0] (
00:37:39[Saint]Hmmmm...another odd recording related define that is lacking for no obvious reason:
00:37:56[Saint]pixelma: any thoughts here?
00:38:21[Saint]again this only seems to exist for the iriver h*
00:38:43***Saving seen data "./dancer.seen"
00:41:00pixelmaI believe screen estate or so, if it's for the H1x0 though then it may "just" not have been adapted to other screens. IIRC it was from some big recording patches that existed for the Irivers in unsupported builds for a while until petur committed them to main Rockbox
00:41:27[Saint]H100, H120, H300
00:42:22[Saint]I guess I'll do the same thing here and just blindly enable it.
00:42:25[Saint]See what happens.
00:42:50[Saint]screen realestate is no issue on the non-nano iPods.
00:45:00 Quit ZincAlloy (Quit: Leaving.)
00:45:02pixelmanon-Mini I guess. The Nano's screen has already more pixels than the H1x0s
00:45:42ulmutulIMO AGC sucks - at least for music recording; you always end up with a lot of compression and pumping artefacts that you can't remove. You get much better results if you record with plenty of headroom and compress afterwards (during playback). So it's not worth the effort (Just my 2 cents :) )
00:46:01scorche|sh[Saint]: the forums thing isnt an issue =)
00:46:21pixelmaBut then there are other targets than Ipods
00:47:54scorche|shfor reference, i have offered to host the main site and so on in the past - there were concerns brought up about hosting everything in the US though
00:49:30scorche|shthe forums have always been a separate site - it initially started when Jeff from Misticriver dedicated a section to Rockbox - that eventually turned into a separate site
00:49:53scorche|shthere were many performance issues though and Jeff was really a separate entity
00:50:21scorche|shi simply stood up and made the effort to transition away from that site and to stand up my own to fix the issues
00:50:35scorche|shthat grew to also hosting translate and themes
00:55:40[Saint]Hmmmm...since the clickwheel iPod keymaps are identical, I guess there is legitimately zero reason for the iPods to not all have AB repeat/AB hotkey
00:55:41pamauryI have mixed feeling about hosting the site in the US too although I don't know if it's a real risk or not
00:56:06 Quit [Franklin] (Ping timeout: 252 seconds)
00:56:07[Saint]It makes shit VERY difficult to adhere to EU regulations.
00:56:17[Saint]data access, etc.
00:56:45[Saint]Though, like you say, not sure if that's actually a real concern or a meta concern.
00:57:15[Saint]For my part, with my services, I just turn away anyone based in the EU.
00:57:17scorche|shi believe the concerns were more around legality of codecs and such
00:58:22pamaury[Saint]: I'm not sure I understand what you mean, hosting the US is okay or not okay ?
00:58:32*[Saint] enabled AB repeat and AB repeat hotkey for all the iPods.
00:58:40[Saint]pamaury: not OK, generally speaking.
00:58:47pamauryok :)
00:59:15[Saint]data access and retention legislations are a royal PITA.
01:00:49[Saint]So much so that I ended up just cutting off the rest of the world from accessing my services about a year ago.
01:08:14 Join munch [0] (~munch@unaffiliated/munch)
01:08:46 Join [Franklin] [0] (~franklin@unaffiliated/franklin)
01:13:34user890104prof_wolfff: you can get it from
01:13:41 Quit ulmutul (Quit: ChatZilla 0.9.92 [Firefox 41.0.1/20150929144111])
01:13:50user890104my classic2g runs it
01:15:03 Quit munch (Quit: ZNC -
01:16:50[Saint]that was a point release for some odd hardware substitution they did, was it not?
01:17:19[Saint]I recall doing some investigation on this a while ago, but perhaps I'm mixing it up.
01:23:11 Join munch [0] (~munch@2601:98a:302:955d:126f:3fff:fed7:41)
01:23:11 Quit munch (Changing host)
01:23:11 Join munch [0] (~munch@unaffiliated/munch)
01:26:31foolsh[Saint]: HAVE_HISTOGRAM compiles for fuze+ but I have no idea what I'm looking for it on the player
01:40:32[Saint]Jesus Christ patch can be really stupid sometimes.
01:40:53[Saint]rejecting a patch with a fuzz factor of a single line.
01:41:45[Saint]Fuck me for wanting each patch to be independent of each other, right? :-S
01:43:19foolsh[Saint]: any particular program suggestions to analyze and set replay gain data?
01:43:48[Saint]I use foobar2000 in WINE personally.
01:44:00[Saint]But that's solely a personal preference thing.
01:44:23[Saint]I just so happen to believe that foobar2000 is the best media player ever for a desktop environment.
01:44:34foolshas long as it works, I give it a try
01:44:52[Saint]MusicBrainz Picard is another. Clementine will do it as well IIRC.
01:45:22[Saint]If you don't use MusicBrainz Picard, do so.
01:45:44*foolsh makes it so
01:45:57[Saint]It's the best metadata management and tagging tool like....ever. IMO.
01:45:58 Quit pamaury (Ping timeout: 265 seconds)
01:46:46[Saint]It uses the audio signature and metadata databases.
01:47:08[Saint]makes tagging foolishly (foolsh-ishly?) easy.
01:47:50foolshright up my alley then
01:48:08[Saint]You seriously never had replaygain tags until now?
01:48:36*foolsh admits he thumbs the volume buttons every track
01:50:10foolshLmao, I gave him a conniption
02:01:22[Franklin]ooh... g#1221 looks really interesting
02:01:28prof_wolfffSaint: thanks, really never tried to build with (D)EBUG, i will try to solve it tomorrow, please tell me if you see any other problem, have you tried the bootloader?
02:01:34fs-bluebot_3Gerrit review #1221 at : 3iPod Classic: dual bootloader - v1 by Cástor Muñoz
02:01:41[Franklin]prof_wolfff: does it work (well)?
02:02:02[Saint]prof_wolfff: I have not, I am just going through now and checking off a couple of things that are disabled but have no reason to be.
02:02:28prof_wolfffyes, it is functional but needs manual install, it is tested on iPod 80 and 160slim
02:02:50[Franklin]that's awesome
02:03:13[Saint]prof_wolfff: as it stands now, there's AB repeat and AB repeat WPS single-action support, FM tuner, and FM tuner RDS.
02:03:22[Saint]But I suspect I'll find more.
02:03:34prof_wolfffuser890104: i have the firmware but need to insert it into the iPod to get the decrypted hash, it is a priority for me
02:03:43[Saint]Just checking to see if AGC and recording histogram adds anything of value to us.
02:04:02[Saint]Now that there's IAP and serial, FM support should "Just Work".
02:04:28[Saint]I can only check to see if it compiles presently, need to find my FM remote tuner.
02:04:49prof_wolfffi have not tried the FM support but a friend of me owns a radio remote so i could test it in the future
02:04:59*[Saint] nods
02:05:06user890104prof_wolfff: i can install it on mine, then sideload emcore and dump the beginning of the hdd, is this going to help you?
02:05:19[Saint]thank you for committing the patch sets for this by the way prof_wolfff
02:05:23[Saint]it is appreciated.
02:06:04[Saint]gerrit is largely useless now as it doesn't show dependencies anymore, and I didn't feel like going through the ~700+ possible iterations of the patch sets. :)
02:06:40prof_wolfffuser890104, i need the decripted NOR, offset 0x40 to 0x50, IIRC i could be done by spireader by TheSeven
02:07:18user890104prof_wolfff: i can dump the flash with emcore, too
02:07:45user890104i'll try it in a couple of minutes
02:08:08prof_wolfffuser890104: i you get a decripted copy, then the hash is at offset 0x40, length=16, if you add this hash to the bootloader then it should work
02:08:53 Join krabador [0] (~krabador@unaffiliated/krabador)
02:10:37prof_wolfffuser890104: the 0x40 offset if from start of NORBOOT, so it is 0x8000+0x40 from the start of the whole NOR
02:11:48 Quit munch (Quit: ZNC -
02:12:11 Join munch [0] (
02:12:11 Quit munch (Changing host)
02:12:11 Join munch [0] (~munch@unaffiliated/munch)
02:17:56user890104prof_wolfff: i remember that i have a backup of the whole NOR flash before any third-party changes
02:20:05prof_wolfffuser890104: it can serve if it is decrypted, encrypted copies are different on each device because each device uses its own UKEY
02:21:37prof_wolfffuser890104: if you cannot get it, i can send you tomorrow a .dfu to obtain it, is its more easy than decrypt it myself
02:22:14user890104prof_wolfff: ok, thanks
02:22:29user890104this is what i have at the moment
02:22:47user890104i guess it's encrypted
02:23:35[Saint]shit - I forget - how do we specify a custom version number/string at compile time?
02:23:58 Quit [Franklin] (Ping timeout: 272 seconds)
02:24:50prof_wolfffuser890104: the only way to verify is adding it to dualboot.c->of_sha[] and try to install the bootloader
02:25:03 Join jtdesigns01_ [0] (
02:25:32prof_wolfffuser890104: the bootloader will refuse to install if it don't find a known OF
02:26:11[Saint]make VERSION=EXPERIMENTAL -j 72
02:27:56[Saint]FM support needs more work than just enable CONFIG_FM_TUNER
02:28:10[Saint]Needs FM to be declared as a valid audio path too.
02:28:12 Quit jtdesigns01 (Quit: jtdesigns01)
02:29:26prof_wolfffSaint: IIRC when doing the IAP patch there was some things to do for the FM tunner, i wil look at this in the future
02:29:26 Join [Franklin] [0] (~franklin@unaffiliated/franklin)
02:32:22[Saint]prof_wolfff: there's a slim chance I might beat you to it, but, I'd have to get you to overlook it anyway.
02:33:07[Saint]needs added to firmware/target/arm/s5l8702/audio-ipod6g.c
02:34:38prof_wolfffi will borrow the remote FM from my friend, should be easy to test an fix it
02:34:54[Saint]what audio driver does the classic use?
02:35:11[Saint]bah - I shouldn't be so lazy.
02:36:41[Saint]Oh lord...
02:36:51prof_wolfffIIRC i leave things almost prepared for the FM
02:37:00[Saint]and of course nothing else uses it, so, the driver has no idea about AUDIO_SRC_FMRADIO
02:38:12[Saint]cs42l55 driver is really quite elegant, in a way. rather simplistic.
02:38:47***Saving seen data "./dancer.seen"
02:39:44[Saint]I suppose we don't have a datasheet for cs42l55 either, just to make things extra wonderful?
02:40:36[Saint] /* Ask Cirrus or maybe Apple what the hell this means */
02:40:44[Saint]I guess that answers that question.
02:42:14prof_wolfffthere is a datasheet for a similar CODEC, IIRC that stuff is internal initialization that depends on power voltages
02:42:20[Saint]Oh. Hmmm....apparently someone has one.
02:42:25foolshHAVE_AGC is confirmed to work on fuze+, I would test the e2x0 but I gave mine away
02:42:37*foolsh looks at [Franklin]
02:43:06[Saint]foolsh: yeah, offhand I'm not seeing any reason for it to not be abled globally for anything with HAVE_RECORDING
02:43:30[Saint]It kinda looks like the define just got forgotten about.
02:44:02[Franklin]what's up?
02:44:14[Saint]Looking through the source, one can see in various places, forgotten or far-too-similar, even occasionally identical but differently named, defines.
02:44:59foolsh"#define HAVE_AGC" in your firmware/export/config/$target will enable automatic replay gain
02:45:00prof_wolfffSaint, user890104: i must go to bed now, i have to get up very early tomorrow, user890104 tell me if you need the .dfu, if so i will prepare it and send to you tomorrow
02:45:03[Saint]a list of all defines and their functions would be rather handy. I'll need to do that to do what I want to do anyway (template generation)
02:45:15[Saint]goodnight prof_wolfff.
02:45:42[Franklin]foolsh: how should I test it?
02:46:02user890104prof_wolfff: gn, yes please upload it somewhere and PM me a link
02:46:22prof_wolfffuser890104: ok
02:51:34 Join adnap [0] (
02:53:01foolsh[Franklin]: check your gmail
02:54:12foolshuse them with per track replay gain enabled should play them exactly the same volume, without it one is much softer
02:55:12adnapCan the database_changelog.txt be changed to UTF-8? Unicode filepaths are messed up when exporting this file.
02:55:44adnapIt would be backwards compatible.
02:56:38adnapOr rather, old database_changelog.txt files would be forward-compatible.
02:59:13[Franklin]foolsh: building
03:02:16foolsh[Franklin]: the setting is in Settings->Playback Settings->Replaygain
03:02:45foolshchange type to Track Gain
03:04:12foolshI cranked preamp all the way but I don't think it matters, and there's Prevent Clipping duh why wouldn't you enable it
03:04:58*[Franklin] has some partial work on ducky
03:05:11[Franklin]jumping is partially there
03:05:20 Quit jtdesigns01_ (Quit: ZNC - 1.6.0 -
03:05:59foolshAh sweet, I was thinking, should relative jumps be used instead of explicit, it seems it maybe "easier" to code
03:07:01[Saint]foolsh: I think there's some fundamental misunderstanding here.
03:07:28*foolsh isn't surprise
03:07:36[Saint]replaygain "Just Works", always has.
03:07:44[Saint]This is gain control for recording.
03:08:01foolshAH thats a horse of a different color!
03:08:07[Saint]'tis indeed. :)
03:08:24*foolsh sets off to record some things
03:08:44[Franklin]foolsh: that's a good point
03:08:57[Saint]Sorry I didn't click prior, I thought you wanted to view/modify gain after recording.
03:08:57[Franklin]maybe labels would be worth implementing
03:09:26foolsh[Saint]: so what does HAVE_HISTOGRAM do?
03:10:01[Saint]it should add a histogram to the recscreen.
03:10:12 Join jtdesigns01 [0] (
03:10:20 Part jtdesigns01
03:10:31foolsh[Franklin]: maybe, but jumping relative lines and checking for out of bounds might be all it needs
03:14:31[Franklin]that makes it too hard to maintain
03:14:54[Franklin]if you want to insert one line, you have to change all the offsets around it
03:15:04 Join jtdesigns01 [0] (
03:15:12foolshokay, I leave it in your capable hands
03:15:32*[Franklin] senses some sarcasm, but OK :P
03:15:45foolshno no, you read me wrong
03:16:16[Franklin]anyway, enabling AGC seems to have messed up the language strings
03:16:43[Franklin]where there should be "Games", there's "Only Unknown Types" instead
03:16:50foolshreally? thats interesting
03:17:10*[Franklin] tries a reboot
03:17:52[Franklin]still messed up
03:19:49foolsh[Saint]: A histogram does not show up in the fuze+ rec screen, maybe there's something else to that define
03:20:01foolsh[Franklin]: thats weird try a clean build
03:20:40[Saint]foolsh: going over the source a couple of times now it seems as though it has been either forgotten about or deliberately removed.
03:20:49 Join Google [0] (442bb28d@gateway/web/freenode/ip.
03:21:03 Quit Google (Client Quit)
03:21:05[Saint]unless I'm repeatedly missing it.
03:21:28[Saint]There's bits of it left in a couple of places, but, it doesn't actually seem to get used.
03:21:49 Join Google [0] (442bb28d@gateway/web/freenode/ip.
03:22:07GoogleHello Everyone
03:22:13 Nick Google is now known as Guest31597 (442bb28d@gateway/web/freenode/ip.
03:22:26[Franklin]hello Guest31597
03:22:29jtdesigns01seriuosly augustine?
03:22:46jtdesigns01(thats my brother messing with me)
03:22:58jtdesigns01i know the ip address :P
03:23:54Guest31597you finally bothered to check the userstring good job ROFL
03:24:27krabadori'm waiting a newer rockbooox!!!!
03:24:28Guest31597u spellt seriously wrong
03:24:38jtdesigns01fool me once shame on you, fool me twice shame on me
03:24:52*foolsh asks the children to go play in #rockbox-community
03:25:14jtdesigns01yeah about to say that myself
03:26:24krabadorfoolsh, ok, stop joking.
03:27:15Guest31597foolsh: sorry
03:28:06Guest31597jtdesigns01: how is the DF coding going???
03:28:55jtdesigns01Guest31597: i am supposed to be writing a paper right now, you are too. bye.
03:29:06Guest31597oops its not DF what is it?
03:29:12Guest31597im done
03:29:20 Quit Guest31597 (Quit: Page closed)
03:32:07 Quit krabador (Quit: Take The Time)
03:34:08[Franklin]good night guys
03:34:22[Franklin]foolsh: will test it tomorrow
03:34:24 Quit [Franklin] (Quit: Lost terminal)
03:37:02 Join test [0] (442bb28d@gateway/web/freenode/ip.
03:37:25 Nick test is now known as Guest31452 (442bb28d@gateway/web/freenode/ip.
03:44:09 Quit Guest31452 (Quit: Page closed)
03:55:43 Quit amiconn (Ping timeout: 246 seconds)
03:55:43 Quit pixelma (Ping timeout: 246 seconds)
04:02:07 Join FSanches [0] (~felipe@2804:14c:37:268b:487e:bfe2:30c2:3a2a)
04:02:27 Join amiconn [0] (~amiconn@rockbox/developer/amiconn)
04:02:37 Join pixelma [0] (~pixelma@rockbox/staff/pixelma)
04:03:08 Quit foolsh (Quit: - Chat comfortably. Anywhere.)
04:04:11 Join foolsh [0] (
04:15:25 Quit adnap (Remote host closed the connection)
04:16:20 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
04:19:35 Quit JdGordon_ (Ping timeout: 255 seconds)
04:38:49***Saving seen data "./dancer.seen"
04:49:35 Quit jtdesigns01 (Quit: ZNC - 1.6.0 -
04:50:20 Join jtdesigns01 [0] (442bb28d@gateway/web/freenode/ip.
05:08:14 Quit jtdesigns01 (Ping timeout: 246 seconds)
05:32:15 Join nick_p [0] (
05:34:53 Quit TheSeven (Disconnected by services)
05:35:06 Join [7] [0] (~quassel@rockbox/developer/TheSeven)
05:35:11 Quit nick_p (Client Quit)
05:35:32 Join nick_p [0] (
05:36:58 Quit nick_p (Client Quit)
05:48:25 Quit n1cky (Quit: leaving)
06:35:14 Quit ParkerR (Quit: Leaving)
06:38:52***Saving seen data "./dancer.seen"
06:44:43 Join ParkerR [0] (~ParkerR@unaffiliated/parkerr)
06:55:14 Quit ParkerR (Remote host closed the connection)
06:59:59 Join ParkerR [0] (~ParkerR@unaffiliated/parkerr)
08:12:16 Join wodz [0] (
08:15:07wodz[Saint]: AFAIK AGC needs sufficiently fine grained gain steps which IS hardware depended. Putting this aside irivers Hxx are the only targets where someone cared enough to enable this.
08:22:37 Join ender` [0] (
08:24:50wodzAbout moving ifrastructure - I am definitely against hosting in US. We need to host in a country which does allow reverse engineering really.
08:26:12wodzSpeaking about build system - it works but is mainly unmaintained. It lacks a few features like test builds and release builds. There are alternatives.
08:38:54***Saving seen data "./dancer.seen"
08:45:50 Quit amiconn (Remote host closed the connection)
08:45:50 Quit pixelma (Remote host closed the connection)
08:47:52 Join pixelma [0] (~pixelma@rockbox/staff/pixelma)
08:47:54 Join amiconn [0] (~amiconn@rockbox/developer/amiconn)
09:02:33 Join petur [0] (~petur@rockbox/developer/petur)
09:36:38 Quit amiconn (Remote host closed the connection)
09:36:38 Quit pixelma (Remote host closed the connection)
09:37:00 Join amiconn [0] (~amiconn@rockbox/developer/amiconn)
09:37:00 Join pixelma [0] (~pixelma@rockbox/staff/pixelma)
09:52:50 Join b0hoon [0] (~quassel@
09:55:25 Join TheLemonMan [0] (~lemonboy@unaffiliated/thelemonman)
09:57:39 Join soap_ [0] (~soap@rockbox/staff/soap)
09:59:45 Quit soap (Ping timeout: 260 seconds)
10:22:08 Quit munch (Quit: ZNC -
10:22:33 Join munch [0] (~munch@2601:98a:302:955d:126f:3fff:fed7:41)
10:22:33 Quit munch (Changing host)
10:22:33 Join munch [0] (~munch@unaffiliated/munch)
10:38:56***Saving seen data "./dancer.seen"
10:57:30 Join krabador [0] (~krabador@unaffiliated/krabador)
11:11:38 Join JdGordon_ [0] (~jonno@rockbox/developer/JdGordon)
11:14:43 Quit JdGordon (Ping timeout: 272 seconds)
11:34:22 Quit williamtdr (Ping timeout: 264 seconds)
11:59:16 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
12:13:56 Join williamtdr [0] (uid27909@gateway/web/
12:33:45 Quit kvieta (Ping timeout: 260 seconds)
12:38:20 Join kvieta [0] (~kvieta@
12:39:00***Saving seen data "./dancer.seen"
13:28:27 Join pablo_pi_ [0] (~pablo@
13:31:17 Quit pablo_pi (Ping timeout: 255 seconds)
13:40:14 Join Aldem [0] (~Aldem@unaffiliated/aldem)
13:44:14 Quit krabador (Quit: Take The Time)
14:03:13 Join ZincAlloy [0] (
14:03:59 Join stickyb1t [0] (
14:16:22 Quit shamus (Ping timeout: 264 seconds)
14:17:12 Quit stickyb1t (Ping timeout: 246 seconds)
14:17:22 Join shamus [0] (
14:27:47 Part b0hoon ("GTG... Bye.")
14:39:01***Saving seen data "./dancer.seen"
15:19:41wodzpamaury: I tried hard to simplify the logic in hwstub_server. I think there is still small glitch in HWSERVER_OPEN_DEV but I need to think harder about it.
15:22:11 Quit wodz (Quit: Leaving)
15:35:39 Join amayer [0] (
16:01:54 Quit JanC (Ping timeout: 240 seconds)
16:15:49 Join JanC [0] (~janc@lugwv/member/JanC)
16:39:06***Saving seen data "./dancer.seen"
16:50:43 Quit petur (Quit: Leaving)
17:10:37 Quit amayer (Ping timeout: 240 seconds)
17:42:28 Join krabador [0] (~krabador@unaffiliated/krabador)
17:53:50 Quit Aldem (Read error: Connection reset by peer)
17:54:15 Quit rela (Quit: Leaving)
17:58:28 Join dfkt [0] (~dfkt@unaffiliated/dfkt)
17:59:08 Quit TheLemonMan (Quit: "It's now safe to turn off your computer.")
18:01:28 Join Aldem [0] (~Aldem@unaffiliated/aldem)
18:20:51pamaurywodz (logs): I need to study your patch carefully, I think there is a problem in the way you handle device and client disconnect
18:24:22 Join amayer [0] (
18:26:38 Quit pamaury (Ping timeout: 272 seconds)
18:29:54 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
18:39:09***Saving seen data "./dancer.seen"
18:45:58 Join Strife89 [0] (~Strife89@2602:306:bce1:8c20:189d:ae30:aac:4f68)
18:46:21 Quit pamaury (Ping timeout: 246 seconds)
18:47:18Strife89When did the Rockbox project get a time machine? :)
18:51:31Strife89That doesn't make any tense.
18:56:53gevaertsSorry. That wasn't in-tense-ional
19:01:45 Join wodz [0] (
19:32:31 Join lebellium [0] (
19:48:55 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
20:03:11 Quit krabador (Read error: No route to host)
20:03:23 Join krabador [0] (
20:03:32 Quit krabador (Changing host)
20:03:32 Join krabador [0] (~krabador@unaffiliated/krabador)
20:14:27 Join nlogex [0] (
20:14:29 Quit Strife89 (Quit: Leaving)
20:30:46 Join petur [0] (~petur@rockbox/developer/petur)
20:33:16wodzpamaury: ping
20:39:05pamaurywodz: pong
20:39:10***Saving seen data "./dancer.seen"
20:40:03wodzpamaury: You mean that reflist is not freed and hwstub device not released when necessary, right?
20:42:04pamaurynot sure what you mean. What I have in mind is the following:
20:42:04pamaury- devices in reflist should not disappear unless the client has closed them
20:42:04pamaury- reflist must be destroyed and devices closed (or unref) when client quits
20:42:04DBUGEnqueued KICK pamaury
20:42:04pamaury- devices that disappear from global list but are still in reflist could just be marked as "disconnected" and operations on them would always fail and the server would not actually perform the usb operation
20:42:36pamauryin other words, it is the responsability of the client to close a disconnected device
20:44:11wodzwith reflist not destroyed on disconnect I agree, clearly needs to be done
20:44:39wodzbut I don't see the point of holding reflist entry when device got disconnected
20:44:51pamaurybecause client still has a reference to it
20:45:34wodzbut calling any hwstub command after removing entry from reflist will fail with HWSTUB_INVALID_REF
20:45:50wodzthats the sign for the client that device is not valid anymore
20:45:51pamauryexcept close
20:46:16wodzmaybe close should not fail but is bogus anyway
20:47:30wodzif you get HWSTUB_INVALID_REF that means device is dead and you should clean on your side
20:48:17wodzcalling HWSERVER_GET_DEV_LIST is probably the most appropriate
20:48:35wodzI mean to refresh list of devices available
20:48:41wodzon client side
20:48:58pamauryyeah but client should still call close and thus making the server do some kind of 'auto-close' seems wrong. And I think this will actually make the design of server simpler
20:50:18wodzI can't see how this will make server simpler.
20:51:15 Join mirak [0] (
20:53:11pamauryyou keep one global list of devices (some of them may have been disconnected). Each device has a reference counter: number of clients that hold a reference to it. Each client thread has a local list: this is a list of opened devices. Whener a client opens a device, the thread updates the ref count in the global list and adds it to its local list. whener it closes the device, it unref the global one and delete it from the local list. Now each
20:53:11pamauryglobal device has a ref count of 1 initially. One device disconnect (hotplug), you simply decrement this number. In all cases, when a device ref goes to 0, it is removed from the global list
20:54:53wodzOk but who is responsible for checking if ref count is 0?
20:57:20pamaurythe function that unref
20:57:34pamaury(could be hotplug or client)
21:00:13wodzok but now look at such scenario: two clients opened the same device (so ref count is 3 in your scheme), the device got disconnected (ref count drops to 2), now client wants to do something and gets error. It would need to know why it is so to perform internal cleanup and close()
21:00:49wodzwhy bother to issue usb transfer which will fail to tell client it failed?
21:01:20pamauryon disconnect, set a flag to mark the dev as disconnected
21:02:19wodzyou can as well check if the device with specific ref is present on devlist, no need for a flag
21:03:00wodzI see the point however that reflist doesn't have to contain whole copy of dev info
21:03:59pamauryyeah, you have less copies, less allocation, less potential for bugs
21:04:15pamaurythe only allocated data (dev list) is ref count so resource management is obvious
21:04:38pamauryI agree it is conceptually a bit weird at first but in the end I think it's just simpler
21:08:05wodzI see it like this: 1) devlist is as now. It is mutex protected and updated by hotplug callback. 2) client builds up a list on opened refs (only refs, no structure copy) and operates on data from devlist (it is mutex protected anyway). 3) ref count is for holding device hwstub_open() ed only when needed.
21:08:11 Quit bluebrother (Disconnected by services)
21:08:16 Join bluebrother^ [0] (~dom@rockbox/developer/bluebrother)
21:08:16 Join fs-bluebot [0] (
21:09:42wodzWhen device is disconnected the entry from devlist is removed by hotplug callback. Next action from client will return HWERR_INVALID_REF since there is no such ref in list.
21:10:43 Quit fs-bluebot_ (Ping timeout: 255 seconds)
21:11:43pamauryhow do you know the ref is not valid anymore ?
21:12:04wodzsearch for node with this ref in devlist will fail
21:20:46 Join TheLemonMan [0] (~lemonboy@unaffiliated/thelemonman)
21:24:48pamauryI see your point, I just don't like it because it involves searches in two lists for each operation, one of them being mutex protected (and you need the mutex everytime)
21:25:29pamauryand it works only because you implement with lists, any other implementation (hash map for ref list for example) would basically not work
21:25:58pamaurybut you are the one doing it so if you disagree with me just do it the way you want :)
21:29:38pamauryYou might as well think I'm overthinking this ;)
21:29:54pamaury(which is kind of true since thinking all day is my job ^^)
21:32:47wodzYou need to protect global list (or any other book keeping system) with mutex anyway
21:33:42pamauryonly for a few number of operations: open, close and hotplug
21:33:56wodzthe lists are short (realistically how many devices will be connected at once) and how many clients
21:35:03pamaurywith my scheme, a device cannot disappear from the list as long as a client has opened it thus, thus any other operation is safe. In other words, the list is race-free for reading if you know the device is opened
21:35:19pamauryyeah I know the lists are short ^^
21:36:34wodzat the cost that you need to check if the node is not in 'disconnected' state on every operation
21:36:37 Quit mirak (Quit: Ex-Chat)
21:36:57pamauryyeah but that costs nothing
21:37:21pamaury(compared to the cost of doing a usb operation)
21:38:00wodzIn my scheme there is no usb operation on disconnected device either
21:38:56pamauryI think we will not agree on this one :)
21:40:26wodzbasically the difference is that you want to signal disconnection with flag and I want to signal it with failed search.
21:41:08 Join cmhobbs [0] (~cmhobbs@fsf/member/cmhobbs)
21:41:18wodzsince you need to perform search anyway I prefer my solution :-)
21:41:50pamaurybut I only have one search, you have two :-p
21:43:18wodzyou still need to keep ref list per client, don't you?
21:43:36pamauryyeah but the list contains pointer to the global list
21:44:09wodzok, now I understand your idea
21:44:37pamauryyeah I should have said that, it was not obvious from my writing ^^
21:45:03wodzso you want to search only local list, then check validity of pointed node
21:46:42 Nick chardy is now known as chxr (
21:52:02wodzthat makes sense
21:57:10 Quit nlogex (Quit: WeeChat 1.3)
22:02:06 Join jtdesigns01 [0] (442bb28d@gateway/web/freenode/ip.
22:06:46 Quit jtdesigns01 (Client Quit)
22:08:02 Quit FSanches (Quit: Leaving.)
22:08:23 Join jtdesigns01 [0] (~jonathan@2601:400:8000:2669:cc2b:a290:f8f2:ea8a)
22:11:58 Nick krabador is now known as crabador (~krabador@unaffiliated/krabador)
22:22:00 Quit jtdesigns01 (Quit: Leaving)
22:22:23 Join jtdesigns01 [0] (~quassel@2601:400:8000:2669:230:bdff:fe71:cebd)
22:23:20 Quit jtdesigns01 (Client Quit)
22:24:28 Join jtdesigns01 [0] (~ggrt7@2601:400:8000:2669:230:bdff:fe71:cebd)
22:25:01 Quit jtdesigns01 (Client Quit)
22:27:19 Join jtdesigns01 [0] (~quassel@2601:400:8000:2669:230:bdff:fe71:cebd)
22:28:21 Quit petur (Quit: Leaving)
22:37:22 Join michaelni_ [0] (
22:39:13***Saving seen data "./dancer.seen"
22:41:14 Quit michaelni (Ping timeout: 272 seconds)
22:48:07 Join suYin [0] (
22:53:49 Nick crabador is now known as krabador (~krabador@unaffiliated/krabador)
22:59:38 Quit wodz (Quit: Leaving)
23:01:19 Quit Rower (Ping timeout: 250 seconds)
23:01:51 Join Rower [0] (
23:03:01 Quit krabador (Quit: Take The Time)
23:25:25 Quit ZincAlloy (Quit: Leaving.)
23:31:01 Quit amayer (Quit: Leaving)
23:44:29 Quit user890104 (Quit: .)
23:44:39 Join user890104 [0] (Venci@unaffiliated/user890104)
23:52:52 Quit ender` (Quit: There are two types of people in the world: those who can extrapolate from incomplete data.)

Previous day | Next day