--- Log for 16.01.123 Server: calcium.libera.chat Channel: #rockbox --- Nick: rb-logbot Version: Dancer V4.16 Started: 21 days and 10 hours ago 00.14.30 *** Saving seen data "./dancer.seen" 01.55.29 Quit pixelma (Quit: .) 01.55.29 Quit amiconn (Quit: http://quassel-irc.org - Chat comfortably. Anywhere.) 01.56.27 Join pixelma [0] (marianne@p200300ea870c2100305e95fffec66ff3.dip0.t-ipconnect.de) 01.56.28 Join amiconn [0] (jens@p200300ea870c2100305e95fffec66ff3.dip0.t-ipconnect.de) 02.14.31 *** Saving seen data "./dancer.seen" 03.22.13 # <_bilgus> amachronic (logs) something between 2d9cb67 (add chunk_alloc to playlist.c #2) and head is hanging here->https://gerrit.rockbox.org/r/c/rockbox/+/5022/21/firmware/chunk_alloc.c#222 03.23.29 # <_bilgus> i'm not sure if its mine or yours but it works at 2d9cb67 and it hangs at head 03.24.01 # <_bilgus> but only when replacing a dynamic playlist 03.30.53 Join S|h|a|w|n [0] (~shawn156@user/shawn/x-4432647) 03.58.32 # <_bilgus> g#4956 is the first commit that breaks 03.58.35 # 3Gerrit review #4956 at https://gerrit.rockbox.org/r/c/rockbox/+/4956 : 3buflib: Remove CRC checks by Aidan MacDonald 04.09.20 # <_bilgus> hmm disabling CRC paranoia checks in the previous commit breaks it too 04.12.53 # <_bilgus> but setting BUFLIB_NUM_FIELDS 5 it works without the crc checks 04.14.34 *** No seen item changed, no save performed. 04.19.17 # <_bilgus> setting BUFLIB_NUM_FIELDS = 5 at head works as well 04.19.38 # <_bilgus> i'll look over my code again to be sure i'm not overflowing the buffer 06.14.35 Quit ats (Ping timeout: 260 seconds) 06.14.37 *** Saving seen data "./dancer.seen" 06.32.17 Join ats [0] (~ats@cartman.offog.org) 08.14.40 *** No seen item changed, no save performed. 10.14.41 *** No seen item changed, no save performed. 10.43.51 Join jacobk [0] (~quassel@47-186-81-17.dlls.tx.frontiernet.net) 11.54.11 Join amachronic [0] (~amachroni@user/amachronic) 11.56.37 # _bilgus, unfortunately I can't reproduce the hang at HEAD on device or sim. and everything is fine on ASAN, so probably not a buffer overflow. 11.57.03 # <_bilgus> yeah I'm not getting anything in ASAN either 11.57.53 # <_bilgus> so far Ive got making BUFLIB_NUM_FIELDS = 5 works and thats about it 11.58.43 # <_bilgus> I figured maybe I was overwriting the bookkeeping but I figured switching to your malloc version would catch that 11.59.43 # you can try with g#5070 11.59.46 # 3Gerrit review #5070 at https://gerrit.rockbox.org/r/c/rockbox/+/5070 : 3playlist: enable queue send by Aidan MacDonald 12.00.29 # apparently queue_send() has a fallback to nonblocking queue_post() without any warning :( 12.02.23 # <_bilgus> ouch 12.05.08 # <_bilgus> I had an idea to do that a different way i've already forgotten, hopefully looking at it will jar that loose 12.06.23 # i noticed while I was there that dcfrefs is in a buflib alloc that can move without warning 12.07.18 # if the thread is calling dircache_search() when it moves that will be a problem 12.07.47 # Build Server message: 3New build round started. Revision aae34b2e7f, 303 builds, 8 clients. 12.07.47 # 3playlist: enable queue send by Aidan MacDonald 12.11.27 # <_bilgus> oddly the sim talks slower and slower as time goes on 12.11.49 # <_bilgus> and tracs are playing ~10% slower too I don't remember that being a thing 12.14.44 *** No seen item changed, no save performed. 12.20.15 # I'm not getting any timing or pitch problems with playback 12.20.27 # I don't have speech enabled though 12.24.45 # <_bilgus> oh dircache_search yields in open_stream_internal? 12.25.01 # <_bilgus> *or could? 12.25.22 # it does I/O so yeah it can yield 12.25.34 # <_bilgus> what a mess 12.27.49 # <_bilgus> re your ltest patch hth did it work without sending enabled sheer luck? 12.31.22 # Build Server message: 3Build round completed after 1415 seconds. 12.31.25 # Build Server message: 3Revision aae34b2e7f result: All green 12.32.24 # <_bilgus> that commit doesn't fix it 12.33.45 # <_bilgus> i'll try blocking the move of dcrefs 12.36.12 # yep that patch doesn't do anything about dcfrefs moving around 12.36.30 # only the playlist changing things behind the dircache thread's back 12.36.46 # <_bilgus> oh that appears to have worked 12.41.01 # looks simple enough to convert dcfrefs to a buflib handle and core_get_data() 12.41.11 # except playlist_create_ex() 12.41.59 # it seems dcfrefs are added to a non-current playlist? won't they be unused? 12.42.29 # worse they are shared with current_playlist 12.45.35 # <_bilgus> and thats why it fails when a current playlist is up and I try to make a new one I suppose 12.47.03 # <_bilgus> wonder why adding an extra byte to the field made it work.. moved the alloc pattern just enough 12.47.38 # <_bilgus> multithreading woe is me 12.51.30 # <_bilgus> luckily the only caller of playlist_create_ex() seems to be the playlist_viewer and it supplies its own buffer 12.52.39 # <_bilgus> ah ifplayling == true 12.53.06 # I'm just going to get rid of the sharing because it seems only current_playlist can use the dircache anyhow 12.58.25 # <_bilgus> i'm not sure of the intricacies involved there I presume you would load the dircache so viewing a playlist would go smoother 13.02.30 # the dircache thread only fills in current_playlist so I *think* it would be unused 13.07.30 # <_bilgus> yeah i'd agree 13.12.25 # g#5072 13.12.29 # 3Gerrit review #5072 at https://gerrit.rockbox.org/r/c/rockbox/+/5072 : 3playlist: pin dircache fileref buffer during background scanning by Aidan MacDonald 13.19.52 # <_bilgus> nope still crashes 13.22.43 # <_bilgus> and now BUFLIB_NUM_FIELDS = 5 doesn't fix it so I guess we now have 1 more piece of the puzzle but no box lid 13.24.13 # is it the same error too?? 13.25.18 # it's still not crashing for me, with or without that patch 13.25.56 # <_bilgus> no errors just hangs at alloc 13.27.52 # how do you know it hangs at alloc? 13.31.09 # <_bilgus> added a splash(0 before and after 13.33.27 Join lebellium [0] (~lebellium@2a01cb040109a6000d9a7d6993fd2072.ipv6.abo.wanadoo.fr) 13.35.06 # <_bilgus> maybe i'm doing something stupid like freeing the buffer out from under it but I don't think I am 13.35.24 # <_bilgus> plus I its a straight alloc 13.39.53 # <_bilgus> I'll try enabling debugf this eve maybe I can catch it in buflib_alloc_ex 13.40.10 # <_bilgus> I haven't tried a serial debug session before 13.41.12 # oh yeah usb serial 13.41.19 # that does work actually 13.42.07 # <_bilgus> good since it hangs I wasn't able to get logf 13.43.41 # <_bilgus> I can't repro it on the sim but on device its any dynamic playlist (even one track out of the db) and 1 track loaded from disk 13.44.38 # <_bilgus> I thought it might have something to do with start_screen resume pb but it happens even from a fresh boot 13.47.49 # <_bilgus> huh damnit removing the config file its now gone :/ 13.49.04 # <_bilgus> luckily I saved it 13.54.57 # <_bilgus> amachronic, I'm sorry to waste your time IDK wth is going on 13.55.15 # <_bilgus> copying the very same install back the crash is gone 13.55.28 # <_bilgus> maybe its a bad drive 13.56.44 # at least a few bugs got fixed along the way :D 14.00.37 # <_bilgus> I guess but i'm incredlous I literally made a copy of the install deleted the config file and nvram bin and the crash stopped 14.00.56 # <_bilgus> copied back the two files and crash still gone 14.01.54 # <_bilgus> if the file was corrupted why would the copy not be 14.02.05 # yeah it doesn't make much sense 14.05.12 # <_bilgus> ok well bbl maybe it'll reappear hope its not a heisenbug 14.14.46 *** No seen item changed, no save performed. 14.16.26 Quit amachronic (Ping timeout: 255 seconds) 14.38.16 Quit jacobk (Ping timeout: 252 seconds) 15.12.20 Quit l (*.net *.split) 15.13.26 Join l [0] (launchd@bitbot/launchd) 15.13.29 Quit l (Max SendQ exceeded) 15.14.30 Join launchd [0] (launchd@bitbot/launchd) 15.43.29 Quit pixelma (Ping timeout: 246 seconds) 15.43.29 Quit amiconn (Ping timeout: 246 seconds) 15.43.49 Join pixelma [0] (marianne@p200300ea87178e00305e95fffec66ff3.dip0.t-ipconnect.de) 15.43.59 Join amiconn [0] (jens@p200300ea87178e00305e95fffec66ff3.dip0.t-ipconnect.de) 16.14.48 *** Saving seen data "./dancer.seen" 16.32.04 Quit Romster (Quit: Leaving) 17.14.35 Quit lebellium (Quit: Leaving) 18.14.52 *** Saving seen data "./dancer.seen" 18.40.07 Join JanC_ [0] (~janc@user/janc) 18.40.18 Nick JanC is now known as Guest8630 (~janc@user/janc) 18.40.18 Quit Guest8630 (Killed (lithium.libera.chat (Nickname regained by services))) 18.40.18 Nick JanC_ is now known as JanC (~janc@user/janc) 18.58.13 Quit JanC (Remote host closed the connection) 18.58.37 Join JanC [0] (~janc@user/janc) 19.33.16 Quit JanC (Ping timeout: 256 seconds) 19.33.30 Join JanC [0] (~janc@user/janc) 20.14.54 *** Saving seen data "./dancer.seen" 20.41.05 Join jacobk [0] (~quassel@47-186-81-17.dlls.tx.frontiernet.net) 22.14.57 *** No seen item changed, no save performed. 23.37.16 Quit m01 (Quit: Konversation terminated.) 23.39.26 Join m01 [0] (~quassel@vps-b172b88b.vps.ovh.net)