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).

#rockbox log for 2021-07-17

00:00:16***Saving seen data "./dancer.seen"
02:00:19***No seen item changed, no save performed.
03:38:11 Join ZincAlloy [0] (
04:00:21***No seen item changed, no save performed.
04:36:57 Join massive_H [0] (
04:40:42 Quit massiveH (Ping timeout: 268 seconds)
05:59:17 Quit Maxdamantus (Ping timeout: 255 seconds)
06:00:23***Saving seen data "./dancer.seen"
06:14:29 Quit asabas (Quit: Relay server offline)
06:14:43 Join asaba [0] (~asabas@
07:10:13 Quit akaWolf (Ping timeout: 246 seconds)
07:12:11speachyamachronic: Is there any point in keeping the m3klinux port now?
07:13:18speachyI also find myself wondering if the spl+bootloader work you did for the x1000 would be (relatively easily) portable back to the jz47xx series
07:19:02 Quit Galois (Ping timeout: 252 seconds)
07:26:00 Join akaWolf [0] (
08:00:26***Saving seen data "./dancer.seen"
08:06:30 Join amachronic [0] (~amachroni@user/amachronic)
08:09:08 Quit akaWolf (Ping timeout: 268 seconds)
08:15:29speachybtw, I did a little rejiggering of the rockbox www site's sidebar menu; 'dev builds' now points at the daily build page, which in turn points at the post-commit build pages.
08:22:21speachythis is a bit friendlier for users, as the daily build pages include links to voice files etc too.
08:27:34rb-bluebotBuild Server message: New build round started. Revision cd17decd5a, 301 builds, 9 clients.
08:30:23 Join akaWolf [0] (
08:31:46speachywtf.. I thik gerrit completely lost its mind. somehow jammed up the server, load average peaked over 140 (!)
08:32:16amachronicthe forums have just become unavailable a few minutes ago too.
08:33:04speachyyeah, the forum software has a fuse of sorts to prevent things from going haywire.
08:33:29speachyforce-restarting gerrit un-wedged things.
08:34:05speachycaused a swap storm apparently
08:36:17rb-bluebotBuild Server message: New build round started. Revision cd17decd, 301 builds, 9 clients.
08:39:05amachronici guess the m3klinux port may eventually become bitrotted if nobody is actively using/testing it
08:39:08speachyshould have restarted the www server before re-starting the buildserver round.
08:39:22speachywell, one could argue it never actually _worked_ to begin with. :)
08:39:48amachronicbarrier to entry is high but it's not impossible.
08:40:10amachronicdrop it if you like, I've got no strong opinion either way.
08:40:11speachysince there was never a way to actually install the buildable-from-rb-source binaries.
08:40:23speachyplus the horrid jankiness of the m3k kernel etc
08:41:12amachronicprobably for the best to axe it then.
08:43:20 Join johnb2 [0] (
08:43:39amachronici am looking at the jz4760's PM now, and though the details differ the SPL boot looks similar to x1000.
08:44:22amachronicit would probably be possible to adapt the x1000 spl but I'm not sure if code sharing makes sense.
08:45:08amachronicthe x1000 code is already convoluted enough as it is to support the SPL
08:45:35speachyI didn't expect that level of serendipity but no longer having to rely on a hand-patched binary blob would be nice.
08:46:04speachy(plus the x3 code has seen a _lot_ of fixes since that bootloader binary was created)
08:46:34amachronicwhat's the current boot setup for those players anyway? I guess piggybacking the OF's SPL?
08:46:51amachronicand replace a kernel with rockbox?
08:47:43johnb2I would love to see a native port for the xduoox x3ii. It has the far better screen than the erosq and I prefer the physical buttons over the tochscreen of the m3k. I sold both my hifiwalker and m3k again ...
08:48:01amachronicjz4740's spl is also similar, but different
08:48:03speachyIIRC it's the stock OF SPL+bootloader but hand-patched to jump into our BL binary (and jump back if the hold switch is on)
08:48:29speachyand then a stock OF update image was hand-mangled to get that updated bootloader in.
08:51:08amachronicthe main missing piece is low-level NAND drivers and the architecture is much different from the x1000.
08:51:28rb-bluebotBuild Server message: Build round completed after 912 seconds.
08:51:42rb-bluebotBuild Server message: Revision cd17decd result: All green
08:54:58speachykicking the www server too (I think this whole mess is once again attributable to bots suddently slamming the wiki)
08:59:42 Quit johnb2 (Ping timeout: 268 seconds)
09:03:49amachronici was thinking, since bootloader installation on x1000 ports & the sansa connect requires libusb,
09:03:58speachythere. that's most of the mess cleaned up.
09:04:04amachronicwould it make more sense to let rockbox utility handle libusb directly,
09:04:18amachronicor keep everything contained in the external binaries that we invoke?
09:04:40speachyI think bluebrother is wanting the latter
09:04:55speachyso those binaries can get "blessed" to run with elevated privliges
09:05:22speachyinstead of root like it currently needs
09:05:38amachronicI think libusb should not need root except on linux though.
09:06:11amachronicon windows the root step is installing the libusb drivers, which is currently pushed off to zadig.
09:06:19 Join dconrad [0] (~dconrad@
09:06:22amachronicmac apparently doesn't need any permissions.
09:06:53amachronicfor linux I think we just need a udev rule to make the device accessible to non-root users.
09:07:43Arsenhow common is it for ipod 6gs connector to start failing intermittently? I keep trying to rsync some music to my ipod and I keep getting 'usb 1-3: reset high-speed USB device number 25 using xhci_hcd', besides, finding the right port on my PC for the ipod to get past setup is also difficult (and I can replicate this on at least two machines, so it's probably the ipods fault)
09:08:28Arsenamachronic: that's quite common, for instance:
09:09:32speachyArsen: the connector itself? that's relatively rare, but it's quite common for the _cables_ to fail.
09:09:48bluebrotherit's not just the root issue. It's also a dependency issue.
09:10:18bluebrotherback in the days we had a dependency on libusb on mac, which was always a hassle. Getting rid of that really helped to simplify the build process.
09:10:38bluebrotherbut then again we only use libusb to enumerate the connected usb devices, so that wasn't too hard to replace.
09:10:53Arsenspeachy: could be the cable, yeah
09:11:09speachyso I'd try another cable first...
09:11:21ArsenI have two cables, though, both are intermittent, and I don't have a different apple device to test them with
09:11:30bluebrotherbut we still have that sudo / elevate issue on all OSes. So the best solution would be to offload the elevated stuff to external binaries.
09:15:19amachronicthe point is with libusb, the external binary does not need elevation, at least in the x1000's case.
09:15:27amachronicour elevated binary could just be a 'libusb enabler'
09:15:46amachronicrather than each port-specific install utility
09:16:52amachronicI guess some non libusb utilities may still require elevation
09:17:20amachronicipod maybe?
09:17:29speachyeg stuff that needs raw filesystem access, eg the ipodpatcher
09:19:32amachronicat minimum we need a way to designate raw USB devices in rbutil rather than just a mountpoint / drive letter.
09:19:40Arsenmaybe polkit can help? if something needs raw access to the block devices, one could instruct polkit to open that file, and it'd consult itself with its rules and the user, and hand it back if it is possible, I believe
09:19:58Arsenlibusb should be able to enumerate devices on Linux without any special privileges (see: lsusb on the cli)
09:20:23speachypolkit is great under linux, but most rbutil folks are on $OtherOS
09:20:56Arsenhm, good point, not sure how that would work on macos (windows presumably just lets you do anything as any process though :^)
09:21:49amachronicrbutil already does enumeration; we just need a way to pass the right device to the helper binary.
09:22:30amachronicthe current rbutil setup is linking all the helpers into rbutil itself as libraries, so if rbutil used libusb everywhere it would be easy to just make an API accepting a libusb device handle in all the helpers.
09:29:20 Join Maxdamantus [0] (~Maxdamant@user/maxdamantus)
09:29:53bluebrotherI don't think that would work. If we pass a handle it already is open ...
09:30:42bluebrotheralso, this is a generic issue. For the Ipods we don't need libusb, but raw disk access. So the elevation issue exists for non-libusb stuff, and on macos there's no polkit.
09:31:22bluebrotherI've started looking into moving that stuff into separate binaries some time ago, but some communication issues came up (and this not-having-time-issue)
09:31:59bluebrotheron Windows we also don't use libusb. We did in the past, but as with macos that complicated things quite a bit
09:32:04amachronicI agree elevation is a generic issue; but I need a way to pass _USB_ devices and not just _disks_ to the helpers.
09:32:46bluebrotherwell, is there a way to uniquely identify the device?
09:33:22amachroniceither by usb vendor/product ID, or by some more unique method (like bus/port numbers, I'm not sure)
09:33:43bluebrotherthe bus/port method might only exist on Linux
09:34:22amachronicthat's why I figured "open the device handle and pass it to the library" would be easier
09:35:05amachronicwith vid:pid we have to bail if there is more than 1 device plugged in.
09:35:14amachronic(arguably that is not a big limitation)
09:35:17bluebrotherbut if the problem is access permissions passing a handle won't work.
09:35:42bluebrothersince that handle had to be opened, and thus opened with the "wrong" permissions.
09:35:46amachronicpassing a handle is only sensible if the helpers stay linked into rbutil.
09:36:05amachronichigher permissions are not needed to make control requests once the drivers are installed
09:36:09bluebrotherwell, if we don't need elevation it's not much of an issue.
09:36:34bluebrotherthe build dependency might, at least on non-linux
09:37:23bluebrotheron Linux we have it anyway.
09:37:42amachronicwell, I've managed to release a static build of jztool for windows, linux, and mac using libusb.
09:38:05amachronicthe build system is not nice but surely we can just fix that.
09:38:10bluebrothersure, it's possible. As I said, we did use libusb in th e past.
09:38:21amachronicthe alternative is basically re-implementing libusb ourselves using every platform's native API, which seems silly.
09:38:23bluebrotherbut the dependencies on non-Linux were ... painful.
09:38:48amachronicwas this the 'old' libusb or new libusb1?
09:39:37bluebrotheriirc the old one
09:41:56bluebrotherwould be nice if we could dynload it.
10:00:30***Saving seen data "./dancer.seen"
10:12:06 Join johnb2 [0] (
10:14:25 Quit johnb2 (Client Quit)
10:18:56 Quit akaWolf (Ping timeout: 255 seconds)
10:22:44 Join akaWolf [0] (
10:28:46Arsenthe ipod is certainly getting data and the occasional "usb 1-2: reset high-speed USB device number 35 using xhci_hcd", but at 14kBps I won't be done any time soon
10:56:10 Join lebellium [0] (~lebellium@2a01:cb10:2e:2000:685e:32e1:f9f9:cd32)
10:56:16Arsenhm, is there any way I can see what the ipod thinks is going on with the usb connection?
10:56:25Arsenit's currently writing data at that speed from the kernels io cache
11:15:00 Quit ZincAlloy (Quit: Leaving.)
11:15:59 Quit mendel_munkis (Ping timeout: 268 seconds)
11:37:00 Join munkis [0] (
11:40:58 Join _amachronic [0] (~amachroni@user/amachronic)
11:44:21 Quit amachronic (Ping timeout: 268 seconds)
11:59:51Arsenyeah, this definitely makes it unusable, after some period of me trying to write to the drive (I'll try reading next), it fails
12:00:32***Saving seen data "./dancer.seen"
12:16:10Arsenyeah definitely doesn't happen on reads
12:16:35Arsentransferred way more data and it stayed stable
12:16:46Arsenmaybe I should mention at this point that I have an iFlash in this, so it may be related to taht
12:17:15ArsenI do recall a write bug related to ATA power management and iFlash adapters I think
12:24:59Arsenlooks like is down
12:25:23Arsenit's responding to pings
12:26:32 Join johnb2 [0] (
12:35:28 Quit johnb2 (Ping timeout: 252 seconds)
12:48:29 Quit _amachronic (Ping timeout: 268 seconds)
12:51:30 Join ZincAlloy [0] (
13:26:21braewoodsspeachy: is gerrit down?
13:29:41 Quit dconrad (Remote host closed the connection)
13:29:47ArsenI think everything is
13:31:04 Join johnb2 [0] (
13:48:18 Quit johnb2 (Ping timeout: 268 seconds)
13:53:21 Quit akaWolf (Ping timeout: 265 seconds)
14:00:36***Saving seen data "./dancer.seen"
14:03:44 Join akaWolf [0] (
15:01:53__builtinyep, can confirm
15:02:00__builtinspeachy: ^
15:06:44 Join dconrad [0] (~dconrad@
15:11:07 Quit dconrad (Read error: Connection reset by peer)
15:11:19 Join dconrad [0] (~dconrad@
15:18:01 Join cereal_eater [0] (
15:22:46 Quit cereal_eater (Client Quit)
15:23:45 Join cereal_eater [0] (
16:00:37***Saving seen data "./dancer.seen"
16:13:54cereal_eateris the fiio m3k an upgrade to a sansa fuze+? They seem very alike to me.
16:21:12 Quit dconrad (Remote host closed the connection)
16:32:20braewoodscereal_eater: what are you looking for?
16:33:38cereal_eaterI have an sansa fuze+ but the touchpad is somehow uneasy to use. The Fiio M3K also has a touchpad an d the same screen resolution. So my question is, if it is worth buying the Fiio M3K:
16:36:29braewoodsit's a newer port and still what i'd consider experimental
16:36:56 Join dconrad [0] (~dconrad@
16:37:02cereal_eaterthanks braewoods Basically I am looking for a modern mp3 player that supports rockbox, has a big screen and a microsd card slot.
16:37:28braewoodsthat eliminates most rockbox options
16:37:46braewoodsthere's still some viable older ones
16:37:59braewoodsmost of them don't use touch pads or screens
16:38:23braewoodsi own a number of older rockbox units
16:39:18braewoodshonestly the usefulness of sd cards depends on how often you plan on ejecting storage
16:39:39braewoodscereal_eater: how "big"?
16:40:17braewoodsthe one with the largest screen i've seen to date where gigabeat units
16:40:27cereal_eaterI'd like a screen > 320x240 and I need storage > 500 GB
16:40:46braewoodsoh you mean resolution
16:40:57braewoodsunfortunately the wiki is down right now
16:41:15 Quit dconrad (Ping timeout: 255 seconds)
16:41:15braewoodsas for storage...
16:41:25braewoodsdo they even make 512 SD cards yet?
16:41:39braewoodshuh so they do
16:41:42cereal_eatersorry I am talking about micro sdxc cards
16:41:43braewoodsbut pricey
16:41:53braewoodscereal_eater: i know, but eh.
16:42:53braewoodswhat kind of screen you looking for?
16:43:34braewoodsamong the older ports there's some coldfire and PP options. and a few higher end ARM ones.
16:44:11braewoodsi mainly know the HDD units from the old days, most of them can be modded with higher capacity storage
16:44:44braewoodsthe iriver H120 is one of my testing units
16:45:10braewoodsgrayscale screen but it has good visual contrast
16:45:27braewoodscan be modded with flash storage
16:46:36cereal_eaterI have some modded iPods, but I am looking for recent players.
16:46:44cereal_eaterIsn't the m3k a stabel port now?
16:46:55braewoodsnot to my knowledge
16:47:09braewoodsthe more recent "stable" is the agptek rocker...
16:47:22braewoodsxduoo or w/e
16:48:56cereal_eaterm3k is stable & functional:
16:50:38braewoodsso it is. i don't work with that hardware.
16:51:25braewoodsbut if you check the front page
16:51:36braewoodsfiio m3k is under unstable
16:51:47braewoodssame under device status
16:52:07braewoodsthe native port is still experimental last i heard from the chatter here seems to be down
16:52:33braewoodsit is
16:55:55 Join vitt13 [0] (
17:00:36 Quit Maxdamantus (Ping timeout: 255 seconds)
17:02:35 Join Maxdamantus [0] (~Maxdamant@user/maxdamantus)
17:36:43 Quit cereal_eater (Quit: Connection closed)
17:41:47 Quit lebellium (Quit: Leaving)
17:44:56 Quit vitt13 (Ping timeout: 252 seconds)
17:50:20Arsenis the git up?
17:50:24ArsenI'm not sure where it is
17:50:43Arsenno http frontend?
17:50:45Arsenseems down too
17:50:59braewoodsafaik the web site is entirely offline though everything else is still accessible
17:51:00Arsen(also consider upgrading that to a TLS protocol, specifically https, it's not hard to serve git repos with https)
17:51:18Arsenah yeah it's fetching actually
17:51:33braewoodsyou can use the github mirror if it bothers you so much
17:51:49Arsenit doesn't really, it's just a suggestion for something mid-tier priority
17:53:01Arsenis all the rb code on device under firmware/ (excluding bootloader)
17:53:28ArsenI want to bisect a bit to see if there was an ATA bug that could be causing the horrible unstability while writing to an iFlash adapter that was fixed
17:54:17Arsen(ipod6g, a build from like a year ago, I get ``usb 1-2: reset high-speed USB device number 35 using xhci_hcd'' while writing but not reading)
17:55:30 Quit Maxdamantus (Ping timeout: 255 seconds)
17:55:42Arsenand, indeed, aab72f96 sounds related
17:58:32Arsenyeah, this isn't in any releases, I'll give this build a try
17:59:58Arsencan the arm-elf-eabi-gcc cross compiler be one built by me (rather, by my distributions maintainers), or must it be built with the in-tree build
18:00:40***Saving seen data "./dancer.seen"
18:03:26braewoodsArsen: no idea, but only the scripted build is supported
18:07:08Arsenbraewoods: instructions from docs/README?
18:07:36braewoodsi'd have to refer to my old notes... the wiki is offline as i said.
18:07:43braewoodsand i don't know where those are right now
18:08:16braewoodsthere's one script you run for building it under tools or so
18:08:24braewoodsbut it expects some stuff to be installed on the host
18:08:26braewoodsi don't hae that list
18:09:15Arsenyeah, that's why I'm asking
18:09:30Arsenthe wiki being down that is
18:23:24Arsenalright, building with the rbdevwhatever script
18:23:39Arsenconfigure informed me that gcc 11 is, in fact, unsupported
18:25:22braewoodsi use debian 10 containers
18:26:22Arsennot even debian 10 uses gcc 4 though :^)
18:28:22braewoodsyea, we're still on 4.9
18:28:28braewoodsthere's a good reason for that all
18:31:02Arsenoh I know
18:31:50Arsenalright, building normal for the 6g
18:37:10braewoodsi've rewritten minor parts of rockbox to make use of new compiler features
18:37:18braewoodslike C11 support for utf16 strings
18:37:25braewoodsstring literals
18:38:45Arsenah, of course: ``usb 1-2: reset high-speed USB device number 41 using xhci_hcd``
18:44:26Arsenoh yeah that's some weird bug
18:44:35Arsenthe device becomes entirely unresponsive
18:44:53 Quit ZincAlloy (Quit: Leaving.)
18:45:18 Join ZincAlloy [0] (~Adium@2a02:8108:943f:d824:6943:c0ff:45ee:e90b)
18:45:18 Quit ZincAlloy (Client Quit)
18:47:44Arsenoh yeah, any attempt to write to a file called rockbox.ipod seems to fail
18:48:38Arsenthis is quite inconvenient
18:49:20braewoods_bilgus: slowly but surely i'm working on inflate. it does occur to me though this could be used to grant more space on bootloader constrained targets, assuming decompression of the rest of the code is somehow viable.
18:50:03Arsenmaybe windowses drivers have more luck
18:50:29braewoodsArsen: sometimes there's incompatibilities with usb hosts or ports
18:50:55ArsenI've tried xhci and ehci on the same machine
18:51:01ArsenI'm not sure whether it's the same controller on this machine
18:51:10Arsenah, windows update, of course
18:56:16_bilgusbraewoods, we already use ucl compression −− similar to upx on those targets
18:59:08braewoods_bilgus: ah, ok.
19:01:01braewoodsi wonder if lz4 would be any good
19:13:58Arsenthis is so weird, every time I try to write rockbox.ipod it just fails miserably (under any name or path)
19:16:28Arsenthere's some other boot mode iirc
19:17:05Arsenone that allows me to write/read the drive but doesnt actually boot rockbox
19:17:12Arseniirc it's a part of the bootloader
19:22:18 Join Maxdamantus [0] (~Maxdamant@user/maxdamantus)
19:24:38 Join dconrad [0] (~dconrad@
19:33:30Arsenthe ipod can copy files looks like
19:33:58 Join cockroach [0] (~blattodea@user/cockroach)
19:40:39braewoods_bilgus: it's kinda weird. i've never used a bitstream system before.
19:40:49braewoodsinflate works with bit streams instead of bytes
19:40:55Arsenthis is quite inconvenient
19:40:56braewoodsfor reading anyway
19:41:10braewoodsArsen: could it be a hardware malfunction?
19:41:13Arsencan I force the bootloader to go into USB mass storage mode? or something similar
19:41:35braewoodswell, if there is a way to do that, it'd be triggered by connecting to USB prior to booting up.
19:42:11Arsenbraewoods: certainly, but I don't think it is due to the frequency of iFlash related commits and due to this not happening when I copy files around on rockbox itself (rather, it only happens when an external device is using usbms to write to it)
19:42:53braewoodsi don't own any ipods so i can't really sy
19:43:04Arsenfound it: /* Enter USB mode if SELECT+RIGHT are pressed */
19:43:24braewoodsi see. most units i've used just automatically enter it if USB is connected during boot
19:44:51Arsenoh, god, it was going somewhat well, and it reset now, damn it
19:58:02Arsenyeah these attempts are futile, I'll disassemble it tomorrow and put a newer build on the sd card manually
19:58:08ArsenI could also somehow test the sd card too
20:00:44***Saving seen data "./dancer.seen"
20:02:33Arsenor not tomorrow, as the old saying goes, nights are made for screwing with hardware
20:14:57 Quit akaWolf (Ping timeout: 268 seconds)
20:16:24 Join akaWolf [0] (
20:21:08speachythe wiki wedged the www site again. first time in a while though
20:21:57speachyfigures, the first time I'm unavailable for a while it goes titsup
20:22:16speachythe foswiki cache db was over 110MB
20:22:27speachywhich it had to load/parse on every...single...hit...
20:26:05speachyArsen: The final fix for the iflash issues was pre-6g, and due to differences in how the 6g+ did its ATA stuff, it wasn't relevant.
20:26:13Arsenspeachy: ah
20:26:15speachyI'm not aware of any 6g+ ata issues
20:26:15braewoodsspeachy: if we were to change wikis, what would we need from the new software?
20:26:23speachybraewoods: migration
20:26:41braewoodsafaik the existing wiki has no official migration path
20:26:42speachyI have a preliminary script to migrate to the dokuwiki syntax
20:26:58braewoodsah, i see.
20:27:00speachyfoswiki 1.x -> 2.x is also convoluted enough to not be easy
20:27:20braewoodsyou'd think they'd have developed a guide or script or so for migration
20:27:21speachyI also did an experiment to see about doing a static export of the existing wiki
20:27:38Arsenhm, the SD cards sector size is 512 and the FS is defined (and operated with) 4K per sector
20:27:40braewoodsin the end it may be easier just to hand port them over
20:28:06braewoodsspeachy: how many pages does it have?
20:28:27braewoodslike 1000 or so?
20:28:35speachyit's one of those giant hairballs that I've not been motivated to deal with.
20:28:44speachythere are currently 2267 pages
20:28:54braewoodswow. that's a lot to move by hand...
20:29:02speachyIf I had to wager I'd say 3/4 are the auto-created user pages
20:29:06Arsenoh wow that's quite a bit
20:29:08speachyonly some of which are of any relevance
20:29:26braewoodsindeed, it may be better to archive the old wiki and gradually port over the useful data
20:29:32speachywhat I'd prefer to do is move everythign to a static site generator instead.
20:29:51braewoodsno objections. can we put updates to it in gerrit?
20:30:01speachyit'll be git-backed, so "of course" :D
20:30:14braewoodsit should be its own repo
20:30:21speachyactually I wanted to do that to dokuwiki, but it turns out its git integration was... not as sane as first thought.
20:30:41braewoodswhat static site generator did you have in mind?
20:31:08speachy(putting aside the obvious downtime for *, the wiki wedging things keeps taking out my personal stuff too, and that's the only reason I noticed things were down again)
20:31:33braewoodsif we went to do this, it might be a good idea to use something compatible with github pages, if that's realistic
20:31:43braewoodsreason being it could serve as a backup copy if we have an outage
20:31:44speachythe options are many, and they all suck in their own ways.
20:32:02speachyI'm _really_ not motivated to try and build new templates, do all the integration work, etc etc
20:32:09ArsenI've had good experiences with hakyll and jekyll, former of which currently does my needs but has the drawbacks of being haskell and being a library for building an ssg
20:32:22braewoodsspeachy: i could give it a spin.
20:32:23speachyI use nikola for my personal www site
20:32:24Arsenyeah building a template would be work
20:32:42Arsenheard good things about it, haven't had to give it a tr
20:32:43braewoodsi've been also doing amateur web development
20:33:21braewoodsi'd just need some idea of requirements so i pick a suitable option
20:33:21speachyrebranding and revamping all around! need a focus group, A/B testing, a full responsive design
20:33:50braewoodsi assume integration with the current rockbox site theme is what we'd want
20:34:33speachyanother thing that could work too is making the existing wiki completely private, login required to do anything
20:34:39braewoodsspeachy: incidently, we might be able to move some stuff like dynamic updates of external content into client-side calls of some kind
20:34:54speachyand export the whole thing staticly for the public view
20:36:02speachythat's what I was leaning towards, where I left off on that effort was fighting the exporter to use the rockbox theme. or rather, trying to port the theme to the exporter's needs.
20:36:51Arsenwhat markup is the wiki using? bbcodes?
20:37:08braewoodsno, some wiki syntax
20:38:32speachyit's using a syntax unique to foswiki.
20:39:33Arsenoh, it's written in perl
20:39:44braewoods(insert perl joke here)
20:40:32speachynothing inherently wrong with perl; it's just the thing is a complete clusterfuck internally. 2.x is considerably improved
20:40:56braewoodsspeachy: in the future i'd like to see if i can do anything with a static site generator
20:41:13braewoodsbut i'd need some way to experiment with rockbox's site theme
20:41:22Arsennothing wrong, yeah, it just shows the age of the thing
20:41:23speachydoing "something" is easy. migrating the whole thing is a _lot_ harder.
20:41:40braewoodsi'm aware, it's going to be tedious no matter what.
20:41:42speachyit's a lot of work, period.
20:41:57speachyand then there's keeping the old URLs working. :D
20:42:06braewoodswhich may not be practical
20:42:40braewoodsbeing able to go through gerrit makes it easier to effect change
20:42:43Arsenmaybe it'd be easiest to rewrite old URLs
20:42:43speachythere's not a lot of deeplink references from within the main site, but there's an awful lot of forum posts and other legacy.
20:43:14braewoodsi'll do some field research.
20:43:18braewoodssee what i can find.
20:43:34braewoodswhat interpreters does the server already support?
20:43:41braewoodspython? python3?
20:44:06braewoodstrying to think of what site generators would integrate with minimal chanes
20:44:52speachysite generators are trivial to integrate. it's static content, after all.
20:45:07braewoodsstill gotta run it server side
20:45:17braewoodsin any case
20:45:18speachybut the server currently has active php and mod_perl running
20:45:39braewoodsah, i meant for running something in a git hook after pushes or so
20:46:08braewoodsa website repo could regenerate stuff after pushes
20:47:39braewoodsspeachy: could you review my crc32 addition? it's been sitting in gerrit for 4+ days
20:48:02speachyagain, trivial to enable. gerrit already has commit hooks running, and beyond that there's always cron.
20:48:14braewoods g#3575
20:48:17rb-bluebotGerrit review #3575 at : crc32: add fast variant of regular crc_32r function by James Buren
20:49:42Arsenalright I'm trying to write to the sdcard to see if it fails, or if its something happening on the ipod side of things
20:50:02ArsenI have to use a loop device in order to emulate a logical sector size of 4096, though
20:50:24Arsenthat is slightly suspect to me, but it could be that the adapter emulates this in hardware too, since it did work
20:51:10braewoodsArsen: i remember reading that the original ipod drive is something weird like 2048.
20:51:41Arsenthis is an sd card, though, through an iflash adapter, and I know that doesn't have a very great reputation but I don't have any better option :P
20:51:48 Quit dconrad (Remote host closed the connection)
20:53:04 Join dconrad [0] (~dconrad@
20:55:16Arsenhm, `.opt/rockbox/firmware/export/config/ipod6g.h:#define SECTOR_SIZE 4096`
20:55:32Arsendoes rb assume a sector size? or am I not tracing enough code to be able to see the contrary
20:56:26braewoodsit defaults to 512 if nothing else defines it
20:56:52braewoodsit's hard coded in any case
20:57:19Arsenoh huh
20:57:27Arsencould it be not 512 on the iflash adapter?
20:57:38braewoodsno idea.
20:57:46braewoodsbut given it's set by the port
20:57:50braewoodsit's probably 4K
20:58:07 Quit dconrad (Ping timeout: 268 seconds)
20:58:13braewoodskeep in mind there's multiple kinds of sector sizes
20:58:29braewoodsfilesystems are free to group the physical / logical sectors into larger sectors
20:58:42braewoodsit won't necessarily be 1 to 1
20:58:54Arsenthe FS I have is formatted for 4K, so it could be the case, that would mean that I was using more of the drive than 1/8th capacity (since sectors were 1/8th the size they were supposed to be), resulting in write errors
20:59:23braewoodsas long as it is exposed correctly it shouldn't matter but i'm not familiar with ipods
20:59:29braewoodsipods do some really weird shit
20:59:52Arsen``mformat -T 60866434 -h 255 -s 63 -H 0 -S 5 -i "/dev/sdc1" ::'' this is how it was formatted
21:00:09braewoodseventually i plan to add MTP support which may help workaround cases where UMS is somehow less desirable
21:00:25Arsenhow complicated is MTP to implement?
21:00:25braewoodsyou can't use mkfs.vfat?
21:00:32braewoodsit's modest difficulty.
21:00:46braewoodsit's a fairly complicated protocol
21:00:46ArsenI can't remember why I needed to do that
21:01:04Arsenprobably due to the sector size differing between what the ipod acted as and what the sd card reader acted as
21:01:06braewoodsbut it basically means the rockbox firmware does all the writes
21:01:29braewoodsfor the moment though i'm focused on another project
21:01:44braewoodsi was planning to look at MTP again when I finish with the gigabeat S
21:01:57braewoodsas far as pressing issues goes
21:01:58Arsenmaybe mtp would have helped here :P I noticed that writes are mostly/only failing while using UMS
21:02:13braewoodsUMS will still be necessary
21:02:24braewoodsif nothing else it's useful for low level stuff
21:02:40Arsenoh naturally,
21:02:55braewoodsjust i think for most everyday stuff MTP should be preferred
21:03:00braewoodsbut it has to be implemented first
21:03:25braewoodsright now i'm working on ZIP support
21:06:08 Quit Maxdamantus (Ping timeout: 268 seconds)
21:08:52Arsenalright got trunk rockbox running
21:09:06Arsen(which also means I managed to get this ipod to boot again)
21:10:26Arsenphys sector size: 512
21:10:41ArsenI don't see the logical one, probably because it's about four times after I usually sleep
21:11:04Arsenno, it's just not there, but power mgmt: unsupported is :^)
21:11:37 Join dconrad [0] (~dconrad@
21:13:40Arsenwait, is the logical sector size picked by the ata drive or the actual host?
21:14:15speachytbhe drive
21:14:30Arsenwould it be visible in the identifying info dump?
21:25:05Arsenthis spec is "fun" to follow
21:30:02braewoodsthe drive defines the actual sector size
21:30:31braewoodsthough the host is free to cluster drive sectors to form a higher level sector of sorts
21:30:56braewoodsi've used filesystems that can group the physical / logic sectors into a higher level unit of sorts
21:31:28braewoodsuseful sometimes for alignment
21:39:55ArsenI'm decently sure the SD card is fine, writing the same data to it using my card reader works, copying data in rockbox works (I'll test this even further now), but writing over usbms does not
21:42:02speachyI don't think rockbox is hardcoded to a specific sector size but if the player was shipped with a 2K/4K sector drive and you shove an iflash into it, it'll most likely never work with the original firmware ever again.
21:42:44ArsenI don't really care about apples OF so that's fine by me
21:43:15Arsenbut I think iflash tests for that usecase specifically, idk
21:49:57Arsenwrite test concluded (I think): duplicating about 4gb of music worked on the device itself
21:54:52Arsenmeanwhile copying a 700K rockbox.ipod with serial over usb fails
22:00:47***Saving seen data "./dancer.seen"
22:05:54 Join Maxdamantus [0] (~Maxdamant@user/maxdamantus)
22:10:05Arsen"cdc_acm 1-2:1.1: ttyACM0: USB ACM device" yay
22:15:15 Nick massive_H is now known as massiveH (
22:15:17 Quit dconrad (Remote host closed the connection)
22:16:57 Quit cockroach (Quit: leaving)
22:18:20 Join dconrad [0] (~dconrad@
22:23:13 Quit dconrad (Ping timeout: 268 seconds)
22:25:47ArsenI pcap'd the usb traffic and it sends a scsi write(10) command followed by a scsi data out, the data out contains data of the file i tried to cp (I matched it with the data in the file to confirm), and it never receives a response, after a while, the kernel decides to try to reset the device and send the command again, which ultimately results in another write, data out, and no response
22:26:49Arsennow I need to find where scsi is handled presumably
22:27:39ArsenI have almost the entire conversation logged
22:28:19ArsenI'm missing the initial handshake and partition scan, everything past that (mount; cp ~/music/...flac .; reset) is in this sca
22:28:42Arsenat times like these I wish I didn't delegate messing with usb to someone else
22:33:37 Join dconrad [0] (~dconrad@
22:49:29 Quit dconrad ()
23:07:06Arsenhm, I'm seeing nothing in the debug views Show Log File, and I see no log file anywhere on the disk, I must have missed something, is enabling logging just done by enabling debug and logf in the advanced build? (btw i also changed rbdir and checked both the old and the new one)

Previous day | Next day