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 2022-12-31

00:06:14***Saving seen data "./dancer.seen"
01:19:59 Quit massiveH (Quit: Leaving)
01:31:08 Quit l0x0l (Read error: Connection reset by peer)
02:01:37 Quit Bobathan_ (Quit: ZNC 1.8.2+deb2+b1 -
02:06:15***Saving seen data "./dancer.seen"
02:08:37 Join Bobathan [0] (
03:45:55 Quit CH23_M (Read error: Connection reset by peer)
03:46:40 Join CH23_M [0] (~CH23@revspace/participant/ch23)
04:06:16***Saving seen data "./dancer.seen"
05:08:03 Quit hexadecagram (Quit: ZNC 1.8.2 -
05:20:09 Join hexadecagram [0] (~acc@user/hexadecagram)
05:22:35 Join lebellium [0] (
05:26:34 Quit CH23_M (Ping timeout: 252 seconds)
05:28:04 Join CH23_M [0] (~CH23@revspace/participant/ch23)
06:06:21***Saving seen data "./dancer.seen"
06:10:18 Quit CH23_M (Read error: Connection reset by peer)
06:10:37 Join CH23_M [0] (~CH23@revspace/participant/ch23)
06:13:52 Quit jacobk (Ping timeout: 252 seconds)
07:03:06 Join amachronic [0] (~amachroni@user/amachronic)
07:04:19rb-bluebotBuild Server message: New build round started. Revision 855540e1f5, 303 builds, 6 clients.
07:26:00_bilgus<l0x0l> (logs?) its left over cruft fro svn I believe basically a tag that would be autofilled
07:35:29rb-bluebotBuild Server message: Build round completed after 1870 seconds.
07:35:31rb-bluebotBuild Server message: Revision 855540e1f5 result: All green
07:37:42rb-bluebotBuild Server message: New build round started. Revision ea7b80dab0, 303 builds, 6 clients.
07:39:23amachronic_bilgus won't this commit prevent you from remapping things differently in locked context?
07:43:29amachronicI think the guy on the form wants rockbox to behave like the original m3k firmware
07:43:59amachronicit changes volume when unlocked but skips tracks when locked
07:51:40_bilguswell you can still by specifying them in the regular keymaps
07:52:31_bilgusproblem is if I allow the LOCKED_CONTEXT flag you then need to specify it it in the remmap as well and that kinda sucks
07:53:35_bilgusnd I read tht as once locked it reverts to volume change rather than the remap they set
07:55:12_bilgusI'm kinda leaning towards if you want that kind of control then compile from source
07:56:07_bilguskeyremap is already on the cusp of exploding most end users brains as it is
07:57:24amachronicyeah the revert happens because the remap isn't covering LOCKED when it should be
07:57:37amachronicie. the remap should really be in the LOCKED context only
07:59:25_bilgusbut then you force everyone to have to specify it
08:00:07amachronicyep the whole point is to make the buttons do different things when locked
08:00:11_bilgusnot that its hard to add another field but I question the add in complexity
08:00:34_bilgusi'd rather push those people to editing source
08:06:20rb-bluebotBuild Server message: Build round completed after 1719 seconds.
08:06:22rb-bluebotBuild Server message: Revision ea7b80dab0 result: All green
08:06:23_bilgusre previous commit I didn't see any more missing browse_context roots is there a way to warn at compile time if there is a mandatory field missing
08:06:25***Saving seen data "./dancer.seen"
08:06:46_bilgusI suppose the crash is a runtime warning
08:08:36amachronicI don't think it's possible to warn about stuff like that but we could add a NULL check in rockbox_browse()
08:08:42_bilgusoddly I din't notice it outright on device just hung the player
08:09:41_bilgusit still browsed just fine though weird
08:10:42rb-bluebotBuild Server message: New build round started. Revision 484a79fcc0, 303 builds, 6 clients.
08:13:30_bilgusI was already in the middle of adding in context flags I think it might be better to just add them outright in the list though ie. CONTEXT_STD_LOCKED CONTEXT_STD_REMOTE CONTEXT_STD_PLUGIN rather than having a flag
08:15:58_bilgusI'm still unsure though it seems to be bordering on being kafkaesque
08:35:24amachronicCONTEXT_LOCKED is or'd with a normal context so it kind of has to be a flag
08:36:55_bilgusnot to the user though
08:37:19rb-bluebotBuild Server message: Build round completed after 1597 seconds.
08:37:21rb-bluebotBuild Server message: Revision 484a79fcc0 result: All green
08:37:34amachronicyeah you could put locked versions of the contexts in the list i just thought it'd be a bit unsightly :)
08:43:21 Quit amachronic (Quit: amachronic)
08:44:34_bilgusI think itll be less confusing in the end its easier to add flags but then you have to select it and its not at all intuitive
08:50:25 Quit kugel_ (Quit: Lost terminal)
08:59:21 Join speachy [0] (
08:59:22 Quit speachy (Changing host)
08:59:22 Join speachy [0] (~speachy@rockbox/developer/speachy)
08:59:22Mode"#rockbox +v speachy" by ChanServ (
09:00:42speachychris_s: Nice catch on those plugin hangs. We can change the optimization levels for those two games on the ipods but that's just papering over a bug. I wonder if it breaks on other targets too?
09:01:08 Join kugel [0] (
09:03:12speachyI wonder if the TLSF stuff itself is the problem
09:09:01speachyhuh, there's upstream tlsf updates.
09:13:02speachyapparently we've had problems with this on arm before.
09:14:26_bilguswas going to say that the last time moving to 0x8 alignment fixed it
09:14:34speachyupstream definitely has some relevant changes vs what we have.
09:14:40speachythre are other implementations though fwiw.
09:15:31_bilgusi'm going to say there wern't upstream updates the last time but it might have been that it didn't fix the issue
09:16:27speachywe have 2.4.4 plus our own changes, upstream I think is 2.4.6 −− 2.4.5 is some fixes for BSDs, and making printf optional (ie much of our local changes)
09:16:51speachy2.4.6 consists of "fixed a bug in the realloc func (discovered by the rockbox project)" :D
09:17:33speachybut here's another implementation that's a bit more recent:
09:20:16_bilguseh the 2.4.6 version was committed in 2012
09:20:29_bilgusso probably a previous bug we uncovered
09:21:36speachyupstream uses -O2
09:22:21speachyjust changing our tlsf CFLAGS might suffice to fix those crashes chris_s found.
09:22:37speachy(and possibly others, heh)
09:24:20_bilgusmemset isn't there some undefined behavior when using memset in a particular way on casted objects?
09:28:10_bilgusI bet if we align down to a multiple of 0x4 or 0x8 it'd work
09:29:22_bilgusor just do a memset the naive way
09:30:11_bilgusnot that we shouldn't go to the latest TLSF but I bet it still crashes
09:32:26speachyif nothing else I want to reduce the delta between us and them
09:36:30_bilguslaudable goal :p
09:39:01speachyit's tiny, how much work can it be? :D
10:02:47 Quit CH23_M (Ping timeout: 246 seconds)
10:04:47 Join CH23_M [0] (~CH23@revspace/participant/ch23)
10:06:29***Saving seen data "./dancer.seen"
10:24:32 Quit CH23_M (Ping timeout: 272 seconds)
10:27:08 Join CH23_M [0] (~CH23@revspace/participant/ch23)
10:40:12 Quit CH23_M (Read error: Connection reset by peer)
10:40:57 Join CH23_M [0] (~CH23@revspace/participant/ch23)
11:01:48rb-bluebotBuild Server message: New build round started. Revision 85410589d1, 303 builds, 6 clients.
11:26:45rb-bluebotBuild Server message: Build round completed after 1497 seconds.
11:26:46rb-bluebotBuild Server message: Revision 85410589d1 result: All green
11:40:30speachyhmm, we can't call panicf() from within a codec.
11:41:41speachyor any plugin.
11:46:24speachy g#4978
11:46:28rb-bluebotGerrit review #4978 at : tlsf: Update to upstream 2.4.6 release by Solomon Peachy
11:46:55speachyI'm waiting for my mini2g to charge up sufficiently to test this out
11:49:34 Join Schimon [0] (
11:58:27speachyboomshine still seems b0rked on the mini2g with those changes. problem might not be tlsf though −− maybe it's unique to ipods due to them having more than one core?
12:01:18speachyok, it's not crashing; music keeps playing but no input.
12:02:04speachywhich means the updated tlsf would appear to be good.
12:02:13speachy(also works fine on the x3)
12:06:30***Saving seen data "./dancer.seen"
14:06:33***No seen item changed, no save performed.
14:06:36_bilgusboomshine is lua so proib something I missed
14:24:38 Join fourHZ [0] (
14:36:33speachyI'll hold off on merging that until chris_s tests things out on his device.
14:37:17speachybut otherwise it's apparently a win-win. One thing this newer version gets us is panics if it detects internal corruption. so that'll eat a little mroe code space.
14:49:59 Quit hexadecagram (Ping timeout: 260 seconds)
14:59:25 Join chris_s [0] (
14:59:30chris_sI was able to launch both Puzzles and Boomshine on the iPod video.
14:59:45chris_sLike you said, boomshine not reacting to input must have another cause,
14:59:46chris_sso maybe it's fixed on device, although both still crash in the Simulator for me.
15:03:27 Quit chris_s (Client Quit)
15:04:05 Quit Nezumi-sama (Ping timeout: 268 seconds)
15:07:26speachywhoops, didn't check sim build; had to fix a few things in the new tlsf.
15:07:56speachystill building...
15:08:02speachy(I miss my beefy workstation...)
15:08:43speachyboomshine runs in the sim for me, at least with the newer tlsf
15:09:56speachythat said, there's this: ERROR: path is not absolute: "$/rb_defines.lua"
15:10:41 Join chris_s [0] (
15:10:50chris_sprobably because I'm on aarch64
15:11:49chris_sor doing something wrong
15:14:46 Quit chris_s (Client Quit)
15:15:26speachyhmm, that's a definite maybe. :D
15:16:03 Join Nezumi-sama [0] (
15:16:14speachythere's probably plenty of places where ARM is hardcoded as 32-bit
15:18:37speachyjust for giggles, let's see how well this RPi3 does with compiling (and running) the simulator. :D
15:21:13speachyI'm gonna merge it
15:22:05rb-bluebotBuild Server message: New build round started. Revision 485e96d6be, 303 builds, 6 clients.
15:27:33speachy....let's see how long this rpi3 takes to build an ipodvideo simulator.
15:49:09rb-bluebotBuild Server message: Build round completed after 1624 seconds.
15:49:11rb-bluebotBuild Server message: Revision 485e96d6be result: All green
16:03:43speachyA bit over 25 minutes to do a from-scratch sim build on an aarch64 rpi3.
16:06:35***Saving seen data "./dancer.seen"
16:12:19rb-bluebotBuild Server message: New build round started. Revision 6ebec601f9, 303 builds, 6 clients.
16:12:51speachyand it just segfaults on me. :D
16:13:05speachy(granted I'm trying to use it over an SSH connection...)
16:16:23_bilguschris_s I always get that absolute path message in the sims
16:22:43_bilgusI still get a seg fault in the 4g sim too
16:23:01_bilguswith boomshine
16:26:15speachyI wonder if it's a threading thing.
16:31:10_bilgusIts all of lua
16:31:37_bilgusI guess thats worse
16:35:51speachyUgh, really wish we used pthread natively.
16:36:36speachythe ipods are our only SMP targets...
16:46:51_bilguswell at least the sim reproduces it
16:46:59_bilgusthe clipZipSim crashes as well
16:47:08_bilgusi'll try on device
16:47:56rb-bluebotBuild Server message: Build round completed after 2138 seconds.
16:47:58rb-bluebotBuild Server message: Revision 6ebec601f9 result: 5 errors 1 warnings
16:49:31_bilgusasan on the clip zip sim gives a few warnings /home/bilgus/Desktop/RockboxDev/rockbox/firmware/pcm_sw_volume.c:369:32: runtime error: left shift of negative value -340
16:49:54_bilgustlsf.c:345:24: runtime error: left shift of 1 by 31 places cannot be represented in type 'int'
16:50:06_bilgustlsf.c:328:26: runtime error: negation of -2147483648 cannot be represented in type 'int'; cast to an unsigned type to negate this value to itself
16:50:22_bilgusand tlsf.c:350:26: runtime error: left shift of 1 by 31 places cannot be represented in type 'int'
16:52:21_bilgushmm the clipv2 doesn't have a hold button wonder whats up with that
16:58:07_bilgusre. TLSF it works on the device still but thats also an older compiler
17:02:29_bilguson the clipV2 thats a weird error maybe a linker problem??
17:02:45_bilgusaction.o: In function `get_action_worker.constprop.7
17:08:13 Join amachronic [0] (~amachroni@user/amachronic)
17:11:55_bilgushuh maybe the clipV2 does have a hold button
17:14:16amachronicspeachy: you're forgetting the multi-core ipods don't have cache coherency between cores. major footgun there!
17:15:24rb-bluebotBuild Server message: New build round started. Revision 3def8fee8c, 303 builds, 6 clients.
17:16:46_bilgusamachronic I found a few naggles adding HAVE_LOCKED_ACTIONS but I think I have them worked out now −− we could probably add it to all device with softlock at this point and just make it fallthrough to the std actions as I did with the clips
17:17:18_bilguswell sand clipv1/v2
17:17:23amachronicsounds good to me!
17:35:11amachronicdoes anyone have an opinion about killing off named buflib allocs entirely?
17:35:46amachronic_bilgus i know you were concerned about making debugging harder
17:36:22amachronici'm of the opinion it's rarely useful, memory errors are better caught in the sim, and it reduces buflib complexity greatly.
17:36:33amachronicalso it saves a good chunk of binsize
17:38:07amachronicgating the name support behind a debug flag works but would uglify the code with IF_BUFLIB_NAME("foo",) stuff like we have IF_MV(x) for multivolume, which I'd rather not do.
17:43:08_bilgusamachronic I've had to track down buffers before it'd suck to have to write in a bunch of splashses or logfs to do it
17:44:24_bilgusbut I can't argue with reduced complexity since it is strictly a debugging concern
17:44:49rb-bluebotBuild Server message: Build round completed after 1765 seconds.
17:44:50rb-bluebotBuild Server message: Revision 3def8fee8c result: All green
17:45:38amachronici always thought the only place you can see the names is in the debug menu which isn't that useful
17:48:25_bilgusthat is the only place so it'll still allow you to see what chunk of ram an error is reffing but no clue what code unit to start in
17:49:14_bilgusI suppose they are in order so that might be enough context idk
17:50:51_bilgusluckily its been a long while since I've had any buflib issues so eh toss up
17:51:45amachronicI guess there'd be some use for bugs that only repro on device
17:52:19_bilgussomewhere around this time g#2006
17:52:23rb-bluebotGerrit review #2006 at : Buflib add range checks blocks and crc_slot raise panic if out of range by William Wilgus
17:53:18_bilgusah the commit just before that I tracked down
17:54:21_bilgusnot that it was strictly needed it allowed me to see what was before and after the buffer to go looking
17:54:28amachronicyeah I see how it'd be useful
17:56:03amachronicwouldn't malloc-backed buflib + ASAN achieve pretty much the same thing though?
17:56:48amachronicif I'm reading right g#2005 was a buffer overflow into some other buffer
17:56:50rb-bluebotGerrit review #2005 at : Fix tree.c->tree_get_entry_at() buffer overflow by William Wilgus
17:57:52_bilgusyeah it probably would
17:58:10_bilgusone way to find out
17:59:25amachronicguess I'll implement a malloc version of buflib then :)
18:01:37amachronicby the way g#4982 might fix lua woes on ipod
18:01:39rb-bluebotGerrit review #4982 at : lua: rocklib_events fix by Aidan MacDonald
18:03:11amachronici remembered coming across the odd COP thread there when you were doing the playlist mutex patches
18:03:23_bilgusyeah probably the issue of no actions but its crashing in the sims atm
18:04:28_bilgusI probably should get an ipod but i've been so anti apple for so long its hard to justify
18:06:39***No seen item changed, no save performed.
18:08:49_bilgusamachronic running the sim with that patch lua now works do simulators get COP even when the device doesn't?
18:09:04_bilgusor its just another alignment thing
18:09:07amachronici just built a 4g sim myself and it worked
18:09:26amachronicno idea why
18:09:39_bilgusi'll try adding some junk at the end of the plugin
18:12:47amachronicwell no COP emulation on the sim it's only single core
18:13:11amachronicis this an ARM only bug?
18:18:59 Quit CH23_M (Ping timeout: 255 seconds)
18:19:35 Join CH23_M [0] (~CH23@revspace/participant/ch23)
18:20:03_bilguswas the last time
18:23:59_bilgusah bringing your patch up to HEAD the crash returns
18:34:36amachronicyeah that was my mistake
18:35:57_bilguswell messing with alignment doesn't seem to fix it this time
18:36:41_bilgusits probably just conflated with speachys tlsf update but IDK for sure
18:36:56_bilgusi'll have to try the build just before that
18:38:20_bilgusprobably the issue was the COP on the ipod though
18:38:59_bilgusyeah going to review -d I056ca2c0ee27e6cc7946166cf52461efda0ee1fb
18:39:12_bilgusfixes the issue with the clipzip sim
18:40:09_bilgusodd that it doesn't carry over to the device so might just be (un) lucky but I'm asan is probably right
18:40:22_bilgusI'm sure*
18:42:08amachronicI stubbed out tlsf in the sim with system malloc and still got a crash so it's a lua bug, not a tlsf bug.
18:44:11amachronicer sorry i forgot stdlib.h and it did something stupid
18:44:30amachronicit works now that i stubbed out tlsf so it IS a tlsf bug.
18:45:25 Join massiveH [0] (~massiveH@2600:4040:a992:a300:acdf:c010:6d16:f5a8)
18:46:55_bilgusyeah I don't see anything that jumps right out and bytes me heh
18:47:13_bilgusbut I guess we will eventually track it down
18:47:44_bilguslua is a good test for tlsf it is very heavy in the memory churn
18:54:00speachy_bilgus: just realized/remembered that my local toolchains are all GCC 8.5
18:56:00speachyamachronic: and yeah, I know about the lack of coherency. tlsf has locking support but we've never had them turned on...
18:56:30speachyif lua is threaded it might trigger tlsf more.
18:59:36_bilgusits only pseudo threaded
18:59:44_bilguswell except on ipod
19:00:08_bilgusessentially it locks the state and jumps into c
19:01:38_bilgusbut it does churn the ram I did some tests a while ago and it is all over the place allocating and freeing like crazy
19:02:01_bilgusjust start up along boggled my mind when I was logging it
19:04:54 Quit CH23_M (Read error: Connection reset by peer)
19:05:12 Join CH23_M [0] (~CH23@revspace/participant/ch23)
19:19:23 Quit amachronic (Quit: amachronic)
20:06:43***Saving seen data "./dancer.seen"
20:12:34 Quit Nezumi-sama (Ping timeout: 260 seconds)
20:14:05 Join Nezumi-sama [0] (
20:22:15munkisgit bisect becomes less useful when debugging an intermittent fault which only appears on hardware
21:41:02__builtinspeachy: apologies if I'm out of the loop, but why does flyspray inject a bunch of "Warning: Undefined array key..." errors into the HTML output? for example
22:06:45***Saving seen data "./dancer.seen"
23:17:23 Quit Schimon (Ping timeout: 256 seconds)
23:37:10 Quit m01 (Quit: Konversation terminated.)
23:39:25 Join m01 [0] (
23:40:48 Quit Piece_Maker (Quit: ZNC 1.8.2 -
23:41:21 Join Piece_Maker [0] (

Previous day | Next day