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 | 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 2019-07-19

00:35:49***Saving seen data "./dancer.seen"
01:00:30foolsh__builtin: I bit of progress, if I change -O3 to -O2 optimization level I no long get a freeze and it actually gives me an error message "Unable to allocate wave buffer for sound 22 size 28478 9095 22050 7042!"
01:14:07__builtinAhh... it's probably out of memory
01:14:18__builtinHow much ram does the zip have?
01:16:52 Quit Acou_Bass (Ping timeout: 272 seconds)
01:18:58 Join Acou_Bass [0] (
01:19:48foolshDon't know can't find any reliable sources online but PLUGIN_BUFFER_SIZE is 0x80000
01:20:05foolshWhat target do you run it on ipodv6?
01:22:26Bilgus I wanna say the clipzip has around 16 mb
01:22:28__builtinAh, it looks like it has 8 MB (see tools/configure)
01:22:41__builtinProbably on the low end even for wolf
01:22:48__builtinNo way Duke will run on it
01:23:06Bilgusah yeah thats right because it stores two copies of the rom and only loads one
01:25:07foolshmemory=8 # not sure
01:25:44foolshhmm I'm gonna try increasing this and see what happens
01:26:12Bilgusfoolsh do you have the multiboot bootloader??
01:26:48foolshno but point me to it and I will
01:28:11foolshAh thank you
01:32:29 Quit Huntereb (Read error: Connection reset by peer)
01:32:38Bilgusnp I tried to get it added to the actual builds but I never heard back from any of the head mucky mucks and I've also no idea how to upload stuff to the web server for staging either
01:33:43 Join Huntereb [0] (~Huntereb@
01:34:16Bilgusalso it looks like jhMikeS is either gone forever or extended so I'll probably be changing the way the MB firmware works with filesystems but whats on gerrit works for the time being
01:43:10Bilgus__builtin >>>> hmm... I tried out my ipod6.0G and the lcd flashes are completely gone... I that from the patches fixing my reversal of lcd always on off or something else?
01:46:34__builtinI commented out the lcd_awake() that seemed to be causing the problem, but the flashes persisted on my 7G (slim)
01:46:52__builtinyet on the 6.0G (older model), the flashes were gone
01:47:04*__builtin suspects it's a hardware issue at the core
02:25:22foolshyeah nope clipzip definitely has only 8mb. Should ifdef out those targets with limited ram
02:35:52***Saving seen data "./dancer.seen"
02:52:14 Quit MrZeus (Ping timeout: 264 seconds)
03:30:34__builtinulmutul (logs): I think I may know the problem
03:31:04__builtinWith my assembly, that is. There's a corner case if r3=1 when calling the function
03:31:41__builtinafter decrementing it once at the start of the function, it will wrap around and cause big problems
04:12:15 Join dandels [0] (~dandels@unaffiliated/dandels)
04:35:55***Saving seen data "./dancer.seen"
04:45:00__builtinjfc, this is some coincidence I've run into here
04:45:58__builtinthe assembly bug I've been talking about was triggered by a sound exactly 7325 samples in length
04:46:43__builtinwhat's special about 7325, you ask? It's prime modulo *every single* power of two from 4 to 4096
04:47:12__builtinhad the sound had just one more or less sample, it never would've triggered the bug and I never would've caught it
05:14:21 Quit TheSeven (Ping timeout: 250 seconds)
05:14:50 Join TheSeven [0] (~quassel@rockbox/developer/TheSeven)
05:35:44 Quit dandels (Ping timeout: 244 seconds)
06:36:00***Saving seen data "./dancer.seen"
06:42:40 Join amdj [0] (~aaron@freenode/staff/atheme.amdj)
06:45:57amdjI just want to say thankyou for such a great piece of software. I have a Sansa Clip+ glued to my car dashboard and plugged into the aux port and cigarette lighter socket for audio in the car, but it was continuing to play after turning off the car, needing me to turn it off manually or let it go on for hours, losing my mental place in the playlist until the battery died. I was
06:45:57amdjconsidering removing its internal battery, but Rockbox's "Car Adapter Mode" is a much better alternative. I've just flashed it and set that up, and it works. \o/
07:07:44 Quit Moarc (Quit: i znowu NADMUCHAŁ BALONA)
07:11:28 Join Moarc [0] (
07:47:57 Quit flabrus (Ping timeout: 245 seconds)
07:48:11 Join flabrus [0] (
07:48:11 Quit flabrus (Changing host)
07:48:11 Join flabrus [0] (~beard@unaffiliated/flabrus)
07:48:47 Quit bluebrother (Ping timeout: 245 seconds)
07:48:56 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
08:36:02***Saving seen data "./dancer.seen"
08:41:27 Quit TheSeven (Ping timeout: 264 seconds)
08:44:21 Join TheSeven [0] (~quassel@rockbox/developer/TheSeven)
08:53:22 Quit dys (Ping timeout: 245 seconds)
09:18:32 Join dys [0] (~dys@2003:5b:203b:100:a64c:c8ff:fef4:13a6)
10:29:42 Quit Huntereb (Ping timeout: 244 seconds)
10:36:03***Saving seen data "./dancer.seen"
10:54:12 Join Huntereb [0] (
11:04:59vuptrying to build rockbox manually for the clip+ results in a "undefined reference to `find_first_set_bit'" for me
11:05:33vupmanually adding build/firmware/asm/arm/ffs.o to the linker command make is work, at
11:06:17vupanybody have an idea for a proper fix?
11:10:54 Join ZincAlloy [0] (~Adium@2a02:8108:9440:dfc:2437:f781:4cd4:56c)
11:34:57 Quit foolsh (Ping timeout: 248 seconds)
11:40:39 Join foolsh [0] (~quassel@
11:43:25 Join pamaury [0] (
11:43:25 Quit pamaury (Changing host)
11:43:25 Join pamaury [0] (~pamaury@rockbox/developer/pamaury)
11:51:52 Quit ZincAlloy (Quit: Leaving.)
11:57:44 Join vmx [0] (
12:36:06***Saving seen data "./dancer.seen"
13:03:13 Join dandels [0] (~dandels@unaffiliated/dandels)
13:26:16 Quit dandels (Ping timeout: 245 seconds)
13:41:36Bilgusvup I just build for the clip+ like 5 minutes ago I'd say its your system..
13:42:20BilgusI'd try make clean
13:56:21 Join massiveH [0] (
14:02:05vuptried make clean already
14:02:29vupBilgus: which gcc version did you use?
14:09:53gevaertsIf it's about missing files during link, it's not going to be a difference in gcc version I think
14:10:28vupwell the strange thing is it links in libfirmware.a which according to `nm` contains the find_first_set_bit symbol
14:10:58vupso not really sure where its going wrong
14:12:16gevaertsA link order problem, maybe? That still doesn't say why it would go wrong for you and not for other people of course
14:16:52gevaertsAre you using -j or anything like that?
14:17:40gevaertsVersions here (according to the configure output):
14:18:46foolshvup: do you have ccache installed, I had some strangeness recently even after running "ccache -c" I had to uninstall it to compile correctly
14:19:07vupnope don't have ccache installed
14:19:23vupi am using -j$(nproc)
14:19:34vupbut manually running the link command also runs into this issue
14:20:18gevaertsWhat's your output for ar t ./firmware/libfirmware.a ? Mine is
14:21:16vupalso thought about order, but `-lfirmware` is specified before `-lrbcodec` which is where it complains about the missing symbol
14:21:36vupok i am using gcc 9.1.0 ;)
14:21:40vupthats quite a bit newer
14:21:58gevaertsMy toolchain hasn't been updated in a while :)
14:22:31vupwell configure also tells me that 9.1.0 isn't of the recommended 4.4.4 version
14:23:37gevaertsThat may or may not be relevant. configure recommends a specific version because we know it works, but it's been a while since rockbox development was at its peak, so people haven't tested newer compilers extensively recently
14:25:04vupwell fixing it manually by adding build/firmware/asm/arm/ffs.o to the linker command lets in run through and seems to work fine on the actual hardware
14:25:27vupthis is my ar -t firmware/libfirmware.a output
14:26:10gevaertsThat seems to be the wrong link. That's my list :)
14:26:57vupoh yeah
14:27:31gevaertsNot that it matters. They're exactly the same
14:29:36vupah figured it out
14:29:51vupusing −−thumb makes this error occur
14:30:06vupwithout that it works fine
14:30:53vupwhich is kinda sad, because −−thumb brings rockbox.sansa down to 497k from 634k
14:35:28*gevaerts tries
14:36:08***Saving seen data "./dancer.seen"
14:36:11gevaertsYes, same problem here then
14:39:07vupinteresting so nothing todo with my gcc version
14:42:53gevaertsIt works if I put -lfirmware between -lrbcodec and -lunwarminder
14:43:00gevaertsIt fails if it's anywhere else
14:46:16vupcan confirm
14:49:39gevaerts"You are in a maze of twisty little makefiles, all alike"
14:54:14 Join ZincAlloy [0] (~Adium@2a02:8108:9440:dfc:79ea:9da:aac8:ee99)
14:54:49vupwell the linker invokation is exactly the same between a −−thumb and a normal build (apart from using the ../tools/ wrapper)
14:55:12 Quit massiveH (Quit: Leaving)
14:55:27gevaertsYes, I don't actually understand why that makes a difference
14:56:17vupyeah me neither
14:57:14gevaertsFor clip+, a non-thumb build also works with -lfirmware between -lrbcodec and -lunwarminder, so if that also works for all other targets, changing the order might be possible
14:57:28gevaertsAt least if someone manages to understand the makefiles enough :)
15:01:53vuphmm putting `-lfirmware` twice with the second one being after `-lrbcodec` also works
15:02:01vupthat seems a little easier to add in the makefile
15:02:35gevaertsGetting this wrong makes the list of link erros longer than my scroll buffer :)
15:06:05gevaerts makes it work for me for clip+ thumb and non-thumb
15:06:22gevaertsI'm going to try some other targets to see if it breaks anything obvious
15:11:17vupawesome, works for me aswell
15:18:28fs-bluebotBuild Server message: New build round started. Revision b0de98a, 280 builds, 9 clients.
15:18:34gevaertsOK, let's hope nothing breaks :)
15:24:07 Join speachy [0] (40eebded@
15:27:24speachyFWIW I was building my clip+ firmware with GCC 4.9 for a while.
15:27:37speachygenerated a smaller image IIRC.
15:28:01speachynot that I really banged on anything but at least audio playback (with various codecs) was stable.
15:30:10speachyand IIRC the linker issue is caused by a change of defaults in binutils. Not really a GCC bug per se, but rather with how the overall toolchain was configured.
15:33:27gevaertsMaybe, but that's still an old change then
15:34:09gevaertsI mean, my linker is
15:35:10speachyhas to do with aggressively discarding unused symbols.
15:35:14 Quit deevious (Quit: deevious)
15:35:46speachyforget the linker flag that enables/disables that behavior.
15:36:45speachybut I ran into this in the 2011-2012 timeframe; without that aggressive unreachable/uncalled code culling (and -Os, and LTO) the code simply didn't fit.
15:37:12speachyran into a lot of LTO bugs too. :)
15:37:19gevaertsI'm not *against* linkers that produce efficient binaries :)
15:37:40gevaertsSo if changing the order fixes things, I'm happy
15:37:47speachyI'd wager rockbox would greatly benefit from LTO
15:38:04speachybut we should upgrade the toolchains to at least 4.9.x first.
15:40:19gevaertsMaybe, although I seem to have vague memories of people playing with lto back in the day and it not making a huge difference. Could be wrong though
15:41:00gevaertsMostly due to code already being fairly aggressively #ifdeffed
15:41:45speachyEarly GCC4 is when it first became generally usable, but modern GCC+binutils is considerably better.
15:44:37speachyokay, I'm finally respinning the talk-in-plugins patch, rearranging some of the other stuff to be in more consistent places.
15:54:53gevaertsSomething odd is going on with the manual build
15:54:55speachyheh, whoops, I just rebased prior to my gerrit push and now I'm hit by the linker thing.
15:57:26speachydid a clean, make dep, and now it's all peachy again
16:00:12gevaertsok, looks like the build finished and nothing broke :)
16:00:18vuphey looks like nothung broke
16:00:37gevaertsCelebrations are in order! Let's get coffee :)
16:10:15*speachy will have a nice cup of herbal tea instead.
16:10:57gevaertsCoffee is a herb :)
16:15:45speachyhmm, are herbs technically leaves, or does any of the plant qualify?
16:16:02speachycoffee is strange. how can something that smells so good taste so vile?
16:17:08speachybut in any case, I cannot tolerate caffeine, so I can't even have a cup of earl grey. "good decaffinated tea" is an oxymoron, unfortunately.
16:22:05BilgusjHMikes and I played with enabling Thumb for ARM a year or so ago and didn't have those issues maybe its in the logs what we had to do to make it work
16:23:34vupI didn't get to test it a lot yet but with this fix it seems to work fine
16:23:53vup(on clip+ with thumb enabled)
16:26:32Bilgus../tools/configure −−thumb 62 b
16:26:39Bilgus../tools/configure −−thumb 62 n
16:26:48Bilgusb is for bootloader
16:27:18Bilgusill try it real quick
16:27:37gevaertsWell, it should work *now* :)
16:29:23 Join robertd1 [0] (~robertd@
16:33:41Bilgusnot being that I haven't rebased..
16:36:09***Saving seen data "./dancer.seen"
16:36:14speachywhen I get home I can dust off my failing-internal-flash clip+ and give it that a try.
16:36:33speachy(as opposed to my failed-internal-flash clip+)
16:36:34Bilgusyeah same failure weird wonder what actually changes since 2017
16:36:52Bilgusoh and I guess its ../tools/configure −−thumb −−target=62 −−type=n
16:37:00speachy(as opposed to the mass grave of fuzes..)
16:37:51BilgusI've killed 1 clip+ since ~2010ish
16:38:40Bilgusand it still has an external pheonix connector so I can do load measurements just the sd slot died
16:38:44amdjThis Clip+ cost me £0 so I don't mind if it dies eventually.
16:38:52BilgusI would
16:39:27foolshAnyone? Is there a patch to allow rockbox to boot on usb power? I have a fuze+ I'd like to test on but it's battery is long gone
16:39:47BilgusIve been running all my other sansas off the sdcard with the multibootloader patch and I hope that makes the flash last forever
16:40:20Bilgusfoolsh johnb had one for the clip+ and the idea should be the same
16:41:06BilgusI can't remember if he did it I did it or someone else did it but Ill try to find it
16:42:09speachyonly had .rockbox on my internal flash. most likely that being baked daily in Florida sunlight took its toll..
16:42:57*speachy curses at Vivado for the tenth time this morning.
16:43:57vupvivado as in xilinx vivado?
16:44:14vupif yes, i feel your pain
16:48:43foolshBilgus: Ah I found #g899 (after wrangling with gerrit's search syntax)
16:48:45speachymore spefically, the reference design can't be recreated due to 17 missing IP blocks and two more that are present, but require an additional license.
16:49:23Bilgusdid better than me lol I was still looking
16:49:24speachyand starting from the template results in a non-booting system.
16:49:51Bilgus g#899
16:49:51speachythe more expensive a software tool is, the worse its user experience tends to be.
16:49:53fs-bluebotGerrit review #899 at : iPod Classic: Prevent low battery shutdown if no battery is present by Michael Sparmann
16:50:18Bilgusnah that isn't the one
16:53:51vupyeah the vivado ip stuff is great ;) i hope the open source tools will soon support xilinx fpga's
16:55:26speachywe're using bleeding edge stuff so I'm not going to hold my breath. meanwhile I'm tripping over bugs first publicy reported five years ago. :/
16:59:05Bilgus#define NO_LOW_BATTERY_SHUTDOWN
17:00:05speachyso I updated g#2071, incorporating Bilgus's comments. Anyone else with comments? I've been using builds (Rocker, X3) with it (and the follow-on) changes since Feburary, and others have been using Poretsky's various builds for much longer.
17:00:06fs-bluebotGerrit review #2071 at : FS #7704 - Talk support for plugins by Solomon Peachy
17:01:55Bilgusspeachy I have something to go into plugins.h as well let me know before you push it and I'll put mine in with it
17:02:52 Quit dys (Remote host closed the connection)
17:12:29speachyI could go ahead and push it now.
17:12:45speachyyour changes are new additions, or is it something I could fold into this patch?
17:14:07Bilgusno not related just can go with the api version bump
17:14:37BilgusI'll put mine on top as a child and when I push mine it'll push yours
17:18:04speachyokeydokey. I have five follow-up patches staged that talk-enable more plugins, and then I can resume cherry-picking more stuff out of Poretsky's tree.
17:19:00Bilgusare you wanting to push it today I have about an hour I can rb wuith
17:19:44speachyit's been more or less ready since feburary. It can wait until you're ready. :)
17:26:40 Quit noobineer (Ping timeout: 250 seconds)
17:42:54 Join noobineer [0] (~noobineer@2601:401:8200:26d8:e57a:837b:402c:7a7a)
17:55:08Bilgusspeachy I found an issue in the plugins you didn't move some of the entries in both plugins.c and plugins.h
17:55:11Bilgus rbversion,
17:55:11Bilgus root_menu_get_options,
17:55:11Bilgus root_menu_set_default,
17:55:11Bilgus root_menu_write_to_cfg,
17:55:11Bilgus root_menu_load_from_cfg,
17:55:12***Alert Mode level 1
17:55:12Bilgus settings_save,
17:55:23Bilgusat least these but there might be more
17:56:46speachycrap, okay. gimme a moment to fix
18:05:13***Alert Mode OFF
18:05:35 Quit pamaury (Quit: Konversation terminated!)
18:05:37speachydoing a build with -Werror this time
18:09:14speachyokay, it's no longer complaining. can't believe I missed that.
18:09:32speachynew patch set pushed to gerrit.
18:10:08BilgusI only noticed it as I was moving some stuff around to make it more logical
18:10:46BilgusI'm out of time for today so I'll pull the new set down this evening
18:11:38Bilgusthank you! :p
18:11:41 Join Rower [0] (
18:36:13***Saving seen data "./dancer.seen"
18:42:32 Part robertd1
19:00:00 Quit Moarc (Ping timeout: 245 seconds)
19:01:34 Quit vmx (Remote host closed the connection)
19:18:59 Join lebellium [0] (
20:19:07 Join Moarc [0] (
20:36:17***Saving seen data "./dancer.seen"
22:36:19***No seen item changed, no save performed.
22:51:25 Quit Moarc (Ping timeout: 272 seconds)
22:54:02 Quit speachy (Ping timeout: 260 seconds)
22:55:37 Join Moarc [0] (

Previous day | Next day