IRC logs for #openrisc Friday, 2014-05-02

--- Log opened Fri May 02 00:00:30 2014
stekernah, a lot of build-deps for scummvm just 'automatically' has dissapeared while I was fixing single-stepping on mor1kx ;)04:37
stekernlibsdl-dev and libcaca-dev are missing now04:37
stekernlet's see if libcaca-dev is going to put up a fight04:38
stekernoh, it's putting up a fight05:34
stekernhah! in your face libcaca-dev!05:44
stekernlibsdl1.2-dev can be built now05:45
stekernand it's done06:15
stekernargh... scummvm builds fine, but that dash error prevent it from being installed06:26
stekernaha, it wasn't even done, it was just preparing by copying fonts onto the build system06:55
mohessaidI am facing some problems with or1ksim now!07:52
mohessaidI am using the glibc toolchain and when compiled linux and then tried to simulate him, the simulation stack in DHCP request in the xterm terminal and ERROR: failed to read packet from file07:53
mohessaidin the bash terminal from where I lanched the simulator07:54
stekernso... what .cfg are you using?07:56
stekernand did you setup tun/tap devices on your host?07:56
mohessaidthe default one07:56
stekerndefault from where?07:56
mohessaidusing openvpn and bctl07:56
mohessaidbut if I want to work without using the tap and just simulate it using rtx_type= 0 , it doesn't work either and he did the same thing?08:09
mohessaidwhat do you thing about the problem?08:20
-!- Netsplit *.net <-> *.split quits: bentley`, heroux, jeremybennett_, hansfbaier, ysionneau08:34
-!- heroux_ is now known as heroux08:34
-!- Netsplit *.net <-> *.split quits: balto, hno, javax, jeremybennett08:34
stekernmohessaid: I'm not sure where that "ERROR: failed to read packet from file" come from, but I would assume it has to do with that something is not working properly with your tun/tap setup08:42
stekernok, I located that error message to peripheral/eth.c in eth_read_file_packet ()08:48
stekern...which is only called when you are using the file interface, not the tun/tap interface08:49
blueCmdstekern: yeah, I've seen that before09:03
blueCmdtry without -O209:04
blueCmdnote that -O2 is multiple times in the command line09:04
stekernyup, it compiles without -O209:09
stekernI took a look at the code where it fails:
blueCmdoh, cool09:19
stekern...but can't say it unveiled any obvious reasons =)09:21
blueCmdwe could also narrow it down by enabling / disabling optimization flags one by one09:21
blueCmdmore than it's probably related to imaginary numbers, which I guess isn't used that often09:22
stekernyup, that's about the only thing that could be deducted from the place where it fails09:23
blueCmdstekern: no no09:23
blueCmdyou can also see that it's failing inside gcc09:24
stekernperhaps it turn it into some fancy construct that utilizes those modes and then we don't tell properly that we don't have any instructions to handle that and the bitsize checkup just fails09:24
blueCmdand not, for example, in .. ncurses!09:24
blueCmdvery important to know09:24
stekernah, sorry... you're of course right09:24
blueCmd(i'm being stupid now, but mpfr wasn't _inside_ gcc)09:24
stekernI wish that ICE was a bit more informative...09:26
blueCmdyes please09:27
stekernblueCmd: hmm... or...09:34
stekernwhich branch did you build gcc from?09:34
stekernwhich would make _alot_ more sense, since there actually is an assert there ;)09:35
blueCmdstekern: heh09:38
mohessaidI was tried to simulate linux with asterisk (79 MO in size) using or1ksim and it seem like the size is too big for the RAM of or1ksim so changed the size, the errors go but the simulator hang in10:06
mohessaid****************** counters reset ******************10:06
mohessaidcycles 103739, insn #8319210:06
mohessaid****************** counters reset ******************10:06
stekernyou need to use something else than RAM for your rootfs10:07
stekerne.g. nfs share10:07
mohessaidsomething else, I can't ping to the host from linux on or1ksim, but I can ping to other computer in the network? of course using the TUN/TAP10:10
stekernso... fix that?10:11
blueCmdwell, at least he seemed to be able to compile asterisk11:20
stekernoh no: /usr/lib/or1k-linux-gnu/ undefined reference to `__sync_fetch_and_sub_4'11:57
stekernblueCmd: if you can make fake atomics, then I'll do some real atomics as a counter measure and link with them =)12:21
blueCmdhah :D12:31
blueCmdI tried that and things exploded12:31
blueCmdbut feel free12:31
stekernnuclear blast exploded?12:34
blueCmdI compiled libc with it though12:34
blueCmdit was horrible12:34
blueCmdsegfaults.. segfaults everywhere12:35
stekernwhat is 'it'?12:35
blueCmdthe event12:35
blueCmdso many programs died that time12:35
stekernno, I meant, what is 'it' in "I compiled libc with it though"12:36
blueCmdah, my real atomics12:37
blueCmdinstead of the libc atomics I let libc use the atomics in gcc (__sync / __atomic) - it's the new fancy way12:37
stekernah, I'm not using any fancy ways, I just created an .o with __sync_fetch_and_sub_4 and __sync_fetch_and_add_412:38
blueCmddoes it work? :)12:41
blueCmdI added it to the repo anyway12:43
blueCmdmaybe something build-depends on it, you never know!12:43
stekerndon't know, it fails at my first attempt to run it in qemu-user because it doesn't find a directfb screen12:43
stekernwill start a vnc session and see if that makes it happier12:43
stekernlibsdl is in that dir too12:44
blueCmdapt-get install postgresql-9.312:47
blueCmdI dare you12:47
stonersanta_hello, how is the nodejs port going?12:50
blueCmdyou want to run nodejs on or1k or run or1k in nodejs?12:55
stonersanta_but on nodejs first12:56
blueCmdso emulate or1k on nodejs? I guess that's poke53282's area12:58
stonersanta_yup.. he is on it..12:58
blueCmdfor nodejs on or1k you would need to port v8 to or1k I think13:01
stonersanta_or wait till the debian port is ready13:04
blueCmdhah well yes, that would be a dead lock in my case13:05
blueCmdI will probably not port it for a while13:05
blueCmdand since it doesn't help stekern play old video games I doubt he will either13:05
blueCmdstekern: my point is that v8 is machine specific so I cannot package it to or1k Debian easily13:06
blueCmdstonersanta_: *13:06
stekernbah... it segfaults...13:16
stekern...or rather some lib it uses does13:18
stekernsomething for the evening to play with ;)13:19
blueCmdstekern: have fun! :) file a lot of bugs!13:19
stekernit's in libc13:20
stekernwhich libc is it that I get if I install from now?13:22
blueCmdthe New Improved Destructor version13:24
stekernok, good. I think. =)13:24
blueCmdstekern: tried postgresql yet?15:15
stekernum, no. I don't know what to try it with. ;)15:16
blueCmdsee if it explodes if you install it15:16
stekernthat I can do15:16
blueCmdI bet it does15:17
stekernman-db did, but that's not something unusual15:24
stekernFATAL:  semctl(950300, 16, SETVAL, 536) failed: Numerical result out of range15:26
stekernbut this was in chrooted env... maybe I can try on the real target15:26
stekernhmm, this scummvm failure is interesting17:15
stekernit originates in pa_log_levelv_meta() in libpulsecommon17:15
stekernthat function allocates a *huge* stack (17068 bytes)17:16
stekernthen it calls __errno_location17:16
stekernwhen __errno_location tries to save stuff on the stack it segfaults17:17 increasing the user stack to 64k fixes the segfault17:23
stekernbut then it goes: "Could not initialize SDL: Not enough resources to create thread!"17:26
stekernoops, I accidently had set the stack size to 65553618:04
stekernI get those too: SJK DEBUG: unix_mkname: len = 112, sizeof(*sunaddr) = 110, sizeof(short) = 218:05
stekernwhich means that there are struct pad mismatches somewhere too18:06
stekernmaybe I should dump where those are called from too, that'll make it a lot easier to pinpoint the offenders18:08
poke53282I remember, that you compiled scummvm already some time ago without any problems.21:02
poke53282@stonersanta: Today I will put the stuff I have in a branch in the official github repository.21:03
stekernpoke53282: yes, but this is the "compile with all you can find and some more enabled" debian version21:05
poke53282Ahh, Ok21:27
stekernsomething is trying to allocate 128MB of mem...21:29
autuus<poke53282> Cool! like me the repo then21:39
stekernthe key was to disable the audio (like we did last) with SDL_AUDIODRIVER=dummy22:00
stekernit's insanely slow when running it through vnc though...22:27
poke53282you mean running it on an FPGA?22:54
poke53282at least it works.22:54
poke53282Is it possible that you run on your FPGA. I am curious about the results.22:56
--- Log closed Sat May 03 00:00:31 2014

Generated by 2.15.2 by Marius Gedminas - find it at!