IRC logs for #openrisc Thursday, 2014-04-17

--- Log opened Thu Apr 17 00:00:08 2014
stekernblueCmd: hmm, but how does one get a broken libmpfr? from a borken gcc? =)03:29
stekernblueCmd: and just for the record, the handholding invite wasn't a romantic one ;) I'm about ready with my atlys board to try out debian on real hw, but don't know how to start05:02
stekernI guess the README here is a good entry point? https://github.com/bluecmd/or1k-debian05:22
stekerndo I need to do the debootstrap with qemu, or can I do that on the board?05:37
stekernoh, right, I'll need your atomic patch too...05:38
stekernthat and the TLS patch05:39
stekernis there anything else?05:39
olofk_stekern: I've only run the ethmac regression tests in orpsoc-cores. Haven't gotten around to actually use it on a board05:46
olofk_Got my half-finished lx9 microboard and ordb2a ports that should use them05:47
stekernolofk_: didn't you see that I amended that question? =)05:49
stekernit's working like a charm on the atlys board05:50
olofk_Yeah, I saw that. Just wanted to give you the glory of being the first one to run it :)05:51
olofk_I still haven't gotten around to back port some of the hacks from orpsocv205:51
stekernah, fusesoc pioneering!05:52
olofk_IIRC, there's one fix that adds an extra receive fifo for double buffering. I think something like that would be good to have05:53
olofk_There might have been some weird CDC issues in that implementation though, so I didn't want to upstream it directly05:54
stekernok, this has worked solid so far05:54
stekernhaven't done so many rebuilds though, in orpsocv2, it was a 50/60 chance of failing05:55
olofk_I'll be on parental leave for eight months starting in May. Just need to get wifi to the sandbox and I can start working on my TODO list :)06:13
olofk_Think of the children! Unhandled CDC are dangerous!06:14
stekernlol06:17
stekernblueCmd: I had to do this in addtion to the other binfmt instructions: sudo update-binfmts --enable qemu-or1k06:25
blueCmdstekern: really? hm ok07:50
stekernmight be ubuntu specific(?)07:52
blueCmdstekern: aha, I didn't understand the handhold request! hah. btw, I lied before, weston wasn't compiled, but I will try to compile it.07:52
stekern=)07:52
olofk_blueCmd: Did your colleague just make the boards for fun, or are you planning on using them for something internally?07:53
olofk_At work I mean07:53
stekernI'm having problems with the debootstrap, mandb takes forever07:53
olofk_Not, like in your body internally07:53
blueCmdolofk_: for fun07:53
olofk_oh07:54
blueCmdstekern: I think that's a race, it freezes and never completes like 10% of the time for me - I haven't debugged that yet07:54
stekernso I killed mandb, and then the bootstrap continues, but in the end it goes: I: Configuring libc-bin...07:54
blueCmdI would just ctrl+c and redo it07:54
stekernW: Failure while configuring base packages.  This will be re-attempted up to five times.07:54
blueCmdyes, it will do that anyway07:54
blueCmdit's not very polished07:54
blueCmdIt will probably work just fine, when you do apt-get install X it will do stuff it didn't complete07:55
stekernok...07:55
stekernI've tested to boot it (on hw), but I get permission denied problems07:56
stekernwait for pastie...07:56
stekernstill wait...07:57
stekernhere it comes...07:57
stekernhttp://pastie.org/908694407:57
blueCmdweird, I haven't gotten that07:58
blueCmdsure your nfs isn't ro?07:58
stekernshouldn't be, this is my export line:07:59
stekern/home/stefan/openrisc/debian/initramfs 192.168.255.0/24(rw,insecure,no_subtree_check,sync,no_root_squash)07:59
blueCmdlooks fine07:59
blueCmdi wonder if the kernel starts with it ro and since your fstab is empty (I'm guessing) it will not remount it rw08:01
stekernhmm, interesting theory08:25
stekerndid you ever run or1ksim at the point '# Now you can either:', or only later?08:25
stekernnm, it works now after I rebooted the nfs server08:32
olofk_stekern: Is sublime compatible with cursynth? :)08:33
olofk_One of the user comments for cursynth said the exact same thing I was thinking "Why would anyone want that?"08:34
olofk_And why on earth is GNU involved in making a curses-based synthesizer?08:35
stekernno, it's not ;)08:35
blueCmdstekern: building mpfr natively causes the 'log' function to checkfail when running the self-tests. it tries to set precision 1 instead of 65 (if I compile it natively 65 is set, might be different for or1k, but 1 sounds a bit.. low)08:35
blueCmdI'm intrigue, time for work08:35
stekernhttp://pastie.org/908704308:44
olofk_stekern: Cool09:03
blueCmdstekern: nice!09:08
blueCmdstekern: how long time did it take to boot?09:09
stekernblueCmd: didn't measure, but 1-2 minutes perhaps09:21
olofk_stekern: Is that using sysvinit or is systemd implemented already?09:23
olofk_Or is busybox still handling everything?09:24
blueCmdstekern: that's nice. if you do:09:24
blueCmdecho "NOSWAP=yes" | sudo tee -a initramfs/etc/default/rcS09:24
blueCmdolofk_: that's systemd09:24
blueCmdstekern: if you do that you will skip the swap init steps09:24
blueCmdolofk_: and no busybox09:25
olofk_ah yes. Saw some systemd printouts now09:26
olofk_Haven't really learned what the systemd boot looks like. It's so blazing fast on my laptops so I never have time too look at the output :)09:26
olofk_Got confused by the run levels. Thought that was a sysvinit thing09:27
stekernblueCmd: ok, will test that, just running an 'apt-get update' now09:29
blueCmdstekern: weston is now built, you should be able to apt-get update; apt-get install weston10:25
stekernblueCmd: let's see if I can apt-get something simpler first ;)10:40
blueCmdhah10:40
stekernheh, my 1-2 minutes estimate was pretty accurate: 1m 38.46s10:44
stekernwhich of 10s was kernel10:45
stekernapt-get update: ~15 min11:06
blueCmdstekern: hah, wut!12:31
blueCmdhm, I guess snapshot.debian.org is quite big though12:32
mafm_snapshot.d.o unstable should have about 20-21k source packages, and 40k+ binary packages (I don't know what's the percentage of "arch all", if that's what you are using)12:44
blueCmdmafm_: hi there!12:45
blueCmdmafm_: I'm starting to narrowing gcc crashes down I think. mpfr seems to be broken, which would cause gcc to act strange. I'm trying to figure out why.12:46
mafm_very nice :-)12:47
mafm_I have to push another batch of packages to your mirror, cannot estimate how many this time12:47
mafm_but nothing very important12:47
mafm_the issue with krb5 is affecting a lot of packages, also12:49
blueCmdmafm_: I built weston and libcolord today, with some trucks loaded with luck it might just work :)12:49
mafm_:-)12:50
mafm_that reminds me my work with xorg packages, I have to show you my packages for review before I send them upstream12:50
blueCmdso building with -O0 and --enable-logging --enable-warnings --enable-asserts fixes mpfr12:54
blueCmdso this might be optimiziation related12:55
olofkblueCmd: A Heisenbug? :)12:56
stekernblueCmd: yeah, it sits at the [100% yada yada zip src yada yada] for most of thetime13:04
stekernapt-get install gcc went smoothly13:05
mafm_blueCmd: in all of my packages, I've been setting DEB_BUILD_OPTIONS=noopt, to try to build with -O0 always -- not all packages respect that, but it helps13:07
blueCmdmafm_: that's a good idea14:00
blueCmdolofk: yes14:07
stekernI think I'll run an experiment where I lock all 0xc0000000 adresses to 0x014:28
blueCmdmafm_: I have a mpfr deb that will remove at least one of the crashes now15:21
blueCmdmafm_: if you do an update + dist-upgrade you should get it15:24
blueCmdit's compiled with -O0 --enable-logging and --enable-warnings, so it might output stuff, but it never has for moe15:24
blueCmdme"15:24
blueCmdexcept in the tests15:24
mafm_blueCmd: do you know which kind of crashes fixes?15:27
mafm_the ones with -O2?15:27
blueCmdno, just floating point crashes15:27
blueCmdhttps://github.com/bluecmd/or1k-debian/issues/10 that one15:27
mafm_ok, upgraded15:33
mafm_blueCmd: I guess that there's a good reason to not include gfortran?15:44
blueCmdmafm_: it doesn't build :)15:44
blueCmdmafm_: well, maybe it builds but the configure check that enables it tests with gfortran by compiling a program that uses sin, which causes gcc to crash - possibly because of mpfr15:45
mafm_blueCmd: ok.  there are some important libraries depending on that one.16:10
mafm_anyway, about to go home... have a nice week-end/holidays/whatever :-)16:10
blueCmdso it appears that MPFR uses TLS to become threadsafe. I bet qemu doesn't like TLS16:22
blueCmdenabling logging disables threadsafe MPFR16:22
stekernbut didn't you say it happens in or1ksim too?16:24
blueCmdstekern: yes, it does16:24
blueCmdand I just saw that I did implement TLS for QEMU16:25
blueCmdMaybe, just maybe, my perfect TLS code is too perfect16:25
stekernthat of  ourse doesn exclude that it might be tls related16:26
blueCmdstekern: should be possible to boil this down nicely though16:27
stekernI bet that's it, I usually write slightöy imperfect code, and it most often works :)16:27
blueCmdstekern: so I need to bring my perfect code down to your imperfect codes ABI, is that what you're saying?16:28
stekernyes, it doesn't want to socialize with to posh code16:30
stekerntoo16:30
blueCmdETOOPOSH16:33
stonersantaHello, I'm whondering if17:04
stonersantaor how hard would it be to port open risc on nodeJS?17:04
stonersantaas in, if there is someone that could help me get started? running virtual machines on node would definetly be worth it. And if you could run node on a virtual machine that is running on node would really be interesting.17:06
poke53281Hi stonersanta. I suppose you are the same who sent me an email. I am the developer of jor1k.17:15
stonersantaOh yes :)17:15
poke53281Most of the code is browser independent. Especially the part, which runs in an web worker.17:16
poke53281This part should run out of the box in nodeJS.17:16
poke53281The terminal emulation runs in the master thread. But I suppose, that is not required.17:17
poke53281As well as the terminal input. So both will not work in nodeJS.17:18
poke53281The network part requires web sockets which I suppose is not available in nodeJS. But I could be wrong.17:19
poke53281The rest is just basic organizing stuff like loading images from the web server.17:20
stonersantaNode has a lot of stuff. yes. Okay. I get that I need to code up the input and output, but will i at least be able to run linux kernel on it out of the box?17:20
poke53281So, yes. With a little bit justification, it will work in nodeJS.17:20
poke53281Yes, it will run under nodeJS.17:21
poke53281You need something to print ascii chars on screen something to load files from the hard drive. That is the minimum requirement.17:21
stonersantaway cool. websockets btw work in node. haven't done much with them, but node is pretty flexible17:21
poke53281To work with it you should be able to check asynchronous for keyboard input.17:22
poke53281non-blocking I mean.17:23
stonersantayeah, node is a master at asyncing stuff17:23
poke53281Good, then think it can be done in a few hours. At least for me who knows the code ;)17:25
stonersantaLOL! ^^17:25
stonersantaI could pay you to do it. I'm not rich, but i really do appreciate your know-how17:26
stonersantabut, it would be really nice to have it run on node, so that I could make x number of instances of linux, that can open oprts, or even run node themselves ^^17:27
poke53281jor1k is just a fun project for me. So, no money required. I will try to port it.17:29
stonersantathat's my goal, to make a node fractality. Anyways, i can find very many uses for mini-vm on node. the browzer based linux with X windows is really cool, but because it cant open ports, and act like a server its use is limited.17:29
stonersantaway cool! ^^17:29
poke53281Yes, that would be cool. But the biggest problem is compile the programs. But there is an debian port underway.17:30
poke53281jor1k still needs a proxy to connect to the internet.17:30
poke53281At least a tap device to which the ethernet frames are sent.17:31
poke53281tap+bridging. This is the current implementation.17:31
poke53281Can nodeJS open Linux devices and run ioctl?17:32
stonersantahmm.. node could be used as a proxy..17:32
stonersantayes17:32
stonersantai mean, I dunno. at least it can run commands on the console17:32
poke53281you get ethernet frames from jor1k. So let's say IP packets. They have to be handled someway.17:33
poke53281The easiest thing is a Linux tap device.17:33
poke53281The emulator QEMU e. g. has even a user-mode TCP/IP stack implemented, which I obviously don't have. That would require to expand the current code by at least a factor of two just to implement something like this.17:34
stonersantayeah, well, we can worry about that later ^^17:36
stonersantaeven running the emulator on node would be great, even if the network didn't work, cuz node can send data to the emulator itsels anyways17:37
stonersantaeven the possibility to be able to run x86 c++ programs on node would totally expand its horizons.17:38
poke53281well, why not compile them directly?17:38
stonersantabecause its fun to code with node17:39
stonersantaand exiting to have little mini virtual linux machine instances as a part of the node family, and npm packages.17:41
stonersantaI'm allso planning on making a mini browzer game, where you can boot up a computer in the game, and it would actually start a mini VM instance17:43
stonersantathat would not die if the browzer jams.17:44
poke53281Hehe, sounds good17:44
stonersantaand might even have internet access ^^17:44
poke53281Does nodeJS support asm.js?17:44
stonersantaI donät think so :/17:45
stonersantait has V8 engine17:46
poke53281Ok, that's probably fast enough.17:46
olofkstekern: Is there an FPU in mor1k?17:46
poke53281I think to run a tap device you need an addon for nodeJS. Something like this: https://github.com/bramp/node-ioctl17:48
poke53281So, everything is there. Now I need 3 hours of free time :)17:49
stonersantaway cool!! ^^17:50
olofkDo I need to sign the copyright assignment thingie if I just want to add a patch to binutils?20:01
amsolofk: if it is significant, yes (or a disclaimer), if not, no.22:23
--- Log closed Fri Apr 18 00:00:10 2014

Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!