IRC log started Mon Sep 20 00:00:01 1999
[msg(TUNES)] permlog 1999.0920
-:- fare [fare@quatramaran.ens.fr] has joined #Tunes
-:- Zhivago [brian@th.merddin.com.au] has joined #tunes
<Zhivago> hmm, bfe looks interesting
-:- SignOff Zhivago: #TUNES (Leaving)
04:30am
-:- Zhivago [brian@th.merddin.com.au] has joined #tunes
<Zhivago> hmm, that bfe link was broken when I checked a little while ago
05:10am
-:- SignOff Zhivago: #TUNES (Leaving)
-:- fare is now known as Fare
-:- AlonzoTG [Alonzo@216-164-139-200.s454.tnt5.lnhva.md.dialup.rcn.com] has joined #tunes
-:- Zhivago [brian@th.merddin.com.au] has joined #tunes
* AlonzoTG/#tunes ignites a flamethrower and eyes Zhivago
* AlonzoTG/#tunes pitties *.au
<Zhivago> pities is the correct spelling
10:00am
<AlonzoTG> u can't surf for p0rn
<AlonzoTG> =((((((((((((((((((((((((((((((((((((((((((((((((((((((((((((
-:- SignOff AlonzoTG: #TUNES (Have Nice Day :))
10:10am
<Zhivago> what a strange person
-:- hcf [nef@me-portland-us717.javanet.com] has joined #tunes
10:20am
-:- krz [krz@gnat6.owo.com] has joined #tunes
-:- krz [krz@gnat6.owo.com] has left #tunes []
-:- Kaufmann [Kaufmann@dial605.infolink.com.br] has joined #tunes
<Kaufmann> Hey gang, anyone awake and willing to discuss math?
11:10am
<Kaufmann> Well, I'm going now. BRB perhaps.
-:- SignOff Kaufmann: #TUNES (Read error to Kaufmann[dial605.infolink.com.br]: Connection reset by peer)
11:20am
-:- s720 [s720@kneler.ii.uib.no] has joined #tunes
-:- s720 is now known as tmf
<tmf> anyone written CGI scripts before?
11:30am
-:- AlonzoTG [Alonzo@216-164-138-24.s24.tnt5.lnhva.md.dialup.rcn.com] has joined #tunes
-:- SignOff tmf: #TUNES (Leaving)
-:- core [core@core.suntech.fr] has joined #tunes
<core> people hi
* core/#tunes clears his throat
<core> heeeyy-lllloooo
01:40pm
* AlonzoTG/#tunes left clicks on core
* hcf/#tunes right chicks on atg
-:- SignOff hcf: #TUNES (Leaving)
<core> hi al :P
<AlonzoTG> om
<core> om :)
02:00pm
<AlonzoTG> topiq
02:10pm
-:- liar [brand@p0wer.qzx.com] has joined #tunes
<AlonzoTG> A handgranade falls from an umbrella in the middle of a dead channel... Does anyone hear it?
02:40pm
-:- SignOff liar: #TUNES (BRiX [http://www.qzx.com/brix] :: reboot)
-:- eihrul [lee@usr5-ppp119.lvdi.net] has joined #tunes
<core> hey eihrul :)
<eihrul> hey
<core> how's EMK going :)
<eihrul> good, i put out a small pre-release of code (http://www.lvdi.net/~lee.salzman/emk/)
<core> ah yeah, i saw that in the logs and downloaded it, but didn't look at it yet :)
<core> does it do anything yet? :)
<eihrul> nope
<eihrul> it just sits there and looks like assembly code
<core> hehe
<eihrul> tis not bad for 2-3 months though
<core> true.. depends how much it does, but yeah :)
<eihrul> i've been doing it backwards
<eihrul> i've been building all the syscalls first, then i'm going to make it bootable
<eihrul> so by the time it does work, it should be nearly feature-complete
<core> hmm.. well, backwards would be working from the application up into the kernel.. that's ummm.. middleward? ;)
<core> yeah, provided you defined all the syscalls you will ever need :)
<eihrul> considering that was as far as i planned to go on my OS
<eihrul> it's backwards :)
<core> hehe.. cool
<core> will you get anything to run on top of it after that? like emulating linux syscalls? or will you just scrap it :)
<eihrul> i might make a small little X-like server
<eihrul> but i'd have to get a small c-runtime environment going
<core> right.. a libc of sorts
<core> so you aren't going to write servers for your uK? :)
<eihrul> just one or two
<eihrul> i haven't really planned much of what i want to do with it
<eihrul> i just want to climb the mountain for the sake of climbing the mountain, ya know? :)
<core> mmm.. keyboard, mouse, ide disk, filesystem, event dispatcher, that's already more than 2 :)
<core> yeah i know :) it might be interesting to prove your design by getting some actual code to run on it tho :)
<eihrul> aside from the interrupt system
<eihrul> i think all the other things provided are pretty solid
<core> still it's fun to run applications on it ;) trust me :)
<eihrul> you can create memory objects, map them into another task, create threads with specified EIP/ESP, etc
<eihrul> i know, that's why i might get a small X-like server, as i said :)
<eihrul> perhaps i could rip some code from NanoGUI or the like
<core> well, technically the applications will not be running on your uK :)
<core> perhaps yeah
<eihrul> they could if they wanted to :)
<eihrul> the primitives are not exactly exokernel level, but....
03:00pm
<core> well, not if you only get an X server without storage of some sort to load from, or enough C library calls to get applications to compile
<eihrul> the server can be a boot module :)
<eihrul> and like-wise the applications
<eihrul> and then you could make a ram-disk boot module
<core> yeah.. that's how clementine currently copes with the lack of storage, for a couple more weeks or so
<core> but it's only good for a temporary solution or for a diskless box that will remain diskless
<core> well.. still you'll need a libc and most probably also a posix layer
<eihrul> no way am i making a posix layer :)
<eihrul> maybe a libc, but even that is a lot 
<core> well, it's sort of useful to get any X application to run
<eihrul> that's why i said X-like :)
<core> well, don't make your native applications _require_ either, it's only useful for legacy ones
<eihrul> not X itself
<core> ahhhh.. i see
<core> i thought you just wanted to make a diskless X terminal :)
<eihrul> in that protocol would go through messaging
<AlonzoTG> om
<core> yeah.. clementine dependencies abstract the way messages are actually routed, but i see what you mean
<eihrul> of course there will be a client side library so nobody has to send raw messages :)
-:- liar [brand@p0wer.qzx.com] has joined #tunes
<liar> anyone know much about kernel modules?
<core> good :)
<core> liar: what specifically?
<liar> when i boot up it takes several minutes to find my midi module
<liar> and then it fails
<eihrul> core: also, how do you keep track of when to free/allocate page tables?
<core> liar: minutes? is it polling the joystick port or something?
<core> eihrul: when to allocate that's pretty easy.. when they aren't mapped :) when to free them.. whenever pages are needed :)
<liar> i have an awe32 and i ran isapnp and then load the modules
<core> eihrul: first empty page tables are freed, then empty object caches, then fail
<core> liar: hmm.. do you actually need the midi module (otherwise just alias it to off)
<eihrul> but... how do you know when they're empty, is what i'm really trying to ask?
<core> liar: if you do need it, i'm afraid it's not working :)
<liar> i have "alias midi awe_wave" in /etc/conf.modules and the awe_wav.o is in the right place
<core> eihrul: when they aren't referenced anymore :P
<liar> alias midi off?
<eihrul> core: ugh, you keep reference counts for each page table?
<core> liar: right .. and if you do need it, check that it's wave or wav on BOTH :)
-:- hcf [nef@me-portland-us201.javanet.com] has joined #tunes
<liar> -rw-rw-r--   1 root     root        64924 Jul 13 13:37 awe_wave.o
<liar> alias midi awe_wave
<core> eihrul: every page has a small struct MemPage, that allows for very high locality and speed in allocation, it's just kept there; it's also used for shared memory, to know when to actually free it.
<eihrul> ack, isn't that expensive on memory?
<core> eihrul: i use 16 bytes for 4K of physical memory
03:10pm
<core> eihrul: linux uses 48 bytes for every 4K...
<core> eihrul: 16 bytes/page is approx 500 KB for a box with 128 megabytes of memory, and it makes the allocator very speedy/efficient with dma/solving the problems of shared memory and page tables reaping.
<eihrul> hrmm, that would make page allocation considerably easy
<core> eihrul: i think it's a good tradeoff; doesn't use that much memory, and is very speedy/code is very small and general-purpose.
<core> eihrul: then again, YMMV.
<liar> i need a good benchmark program for linux
<liar> what should i use
<core> liar: what do you want to benchmark?
<core> (or benchmarket)
<eihrul> core: hmmm... i wonder if it would work better than my current scheme
<liar> cpu
<core> liar: use rdtsc around the call you want to measure i guess
<liar> ugh
<core> liar: what, it will work :P
<liar> i want a stinking benchamrk program to run tests on it
<core> liar: i don't know of any off the top of my head.. benchmark programs suck, anyway.
<core> bochs is a good benchmark
<liar> i had one called nbench(?) that just did cpu performance
<core> ah.. never really looked for one :)
<core> i know my p3 at work mostly feels like a 386 when running bochs :-)
<liar> well brix in bochs says 7mhz on a p233mmx, celeron 450 and a p2-450
<core> same value on all? try increasing the ips ?
<liar> YA
<core> ya ?
<liar> i changed ips from 750k to 33m and it says 342mhz
<core> liar: there :)
<liar> 342mhz 386
<core> hehe
<liar> ofcourse it took like a minute
<liar> the 7mhz 386 ran faster :)
<core> hehe.. well, you can probably find a faster value (on the celeron) of ips that does work.. it's a question of tuning :)
03:20pm
<liar> did u try bfe?
<core> no.. not yet. stupid 24-hour only days.. is it usable yet?
<liar> yup
<core> does it need anything special besides xforms, on a debian potato-derived box? :)
<liar> after i finish this cpu benchmarking i will finish up the memory dump
<liar> i already have the code on my pilot just need to move it over
<core> does it fill the structures with info yet? without that it's kind of pointless :)
<liar> xforms and bochs
<core> on your pilot? heh
<core> well, i have bochs obviously :P i'll really need to get the debugger to compile in 990718a tho
<liar> ya it kinda needs the debugger :)
<liar> and yes the structure stuff works
<liar> i will also be adding instruction history and step/next locking tonight
<liar> wtf, i did a search for benchmark on byte.com and it gave me an article on there newest version of their benchmark program
<liar> but the stupid article dont say where to get the program
<core> probably a losedose program..
<liar> no
<liar> it runs on every cpu out there
<liar> and every os
<core> liar: cool..
03:30pm
<core> probably wintel have tweaked their os and cpu to work around that benchmarketing tho :P
<liar> heh
<liar> i know that amd and cyrix tweak their cpus for benchmarks
<liar> thats why the k6-2 350 and faster had to have patches for windows
<core> yeah, that's about all they are good for, except for the mediagx :P
<liar> wtf, i just downloaded the source and there is no makefile
<core> ./configure perhaps?
<liar> emfloat.c  misc.c  nbench0.c  nbench1.c  nmglobal.h  sysspec.c  wordcat.h
<liar> emfloat.h  misc.h  nbench0.h  nbench1.h  nnet.dat    sysspec.h
<core> liar: guess you got to compile all .c's together :P
<liar> hmm must be seaparte programs
<core> liar: or maybe it's independant programs
<core> liar: right
<liar> stupid byte idiot
<liar> s/idiot/idiots/
<liar>  /usr/lib/crt1.o(.text+0x18): undefined reference to `main'
<Tril> maynbe because it is a Benchmark Kit and yuo have to write the program yourself?
<liar> this damn code hasnt been ported
<liar> ya
<liar> i used this before tho
<liar> ah its on linux.org
03:40pm
<liar> ugh metalab is slow
<core> it's not slow to me :)
<liar> blow me
<core> mmmm.... all things considered, no thanks :-)
<liar> Description:    Do low-level read & write SCSI benchmark over the first
<liar>                 SCSI disk. Destructively.
<liar> probably dont want that benchmark :)
<core> coolness, new bochs was released 5 mins ago, and includes fpu support !
<core> (and default cpu level is now 486)
<liar> i got that email too
<core> well, just propagating news :)
<core> i've been compiling for 486 cpu emulation for months, but fpu support is great. no more testing on real boxes for that one.
<liar> bbiaf
-:- SignOff liar: #TUNES (BRiX [http://www.qzx.com/brix] :: reboot)
03:50pm
-:- SignOff hcf: #TUNES (Leaving)
-:- SignOff core: #TUNES (Ping timeout for core[core.suntech.fr])
-:- SignOff AlonzoTG: #TUNES (Have Nice Day :))
-:- ElGato [dave@bespin.dhs.org] has joined #tunes
<ElGato> :D
<eihrul> 'lo
<ElGato> i've been gone for a little bit :(
<ElGato> you guys all that talk on the mailing list about high and low level stuff i thought of before
<ElGato> ain't i l33t :D
<eihrul> 31337 you mean
<ElGato> 133t is shorter
<ElGato> for some reason i ditched the high and low shit though
<ElGato> mabye it wasn't such a bad idea
<ElGato> i still don't see where natural language has a place in programming thoguh
<eihrul> it doesn't
<ElGato> yes
<ElGato> you can't specify something when it's fuzzy
<eihrul> natural language can be precise
<ElGato> but not as precise
<eihrul> just as easily as it can be not precise
<ElGato> what's the point though
<eihrul> i have to disagree, just because something allows for impreciseness, does not mean it can't be precise
<ElGato> natural language is only optimised for human to human communication
<eihrul> the point, so far as i can tell, is just that
<eihrul> that most humans use it, hence we're stuck with it :)
<ElGato> bah
<ElGato> the point of computer science is to advance ourselves
<eihrul> there is nothing stating we cannot do such
<ElGato> we shouldn't dwaddle in our own stupidity
<ElGato> :D
<eihrul> we're not... we're helping others dwaddle in their stupidity
<ElGato> if tehy wanna do that they can program in cobol :P
<eihrul> intercal!
<ElGato> it's been short but i gotta go
<ElGato> gotta do homework :\
<ElGato> bubye
<eihrul> short is relative impreciseness of natural language :)
<ElGato> :P
<ElGato> is water at sea again?
<eihrul> it has been a quantity of seconds, much less than an hour
<eihrul> where he is, he is not here often anymore
<eihrul> er whereever
05:00pm
<ElGato> im gonna send him physical mail
<ElGato> :D
<ElGato> later
-:- SignOff ElGato: #TUNES (ElGato has no reason)
05:10pm
-:- Downix [down@d-gnaps-29.ici.net] has joined #tunes
-:- SignOff eihrul: #TUNES (Ping timeout for eihrul[usr5-ppp119.lvdi.net])
-:- eihrul [lee@usr5-ppp119.lvdi.net] has joined #tunes
-:- liar [brand@p0wer.qzx.com] has joined #tunes
<Downix> hey liar
<liar> ahhh
<liar> i just stuck my dual in this machine and updraded to 256meg rm
06:00pm
-:- KorGrey [kor@pm-ppp48.triton.net] has joined #tunes
-:- KorGrey [kor@pm-ppp48.triton.net] has left #tunes []
-:- hcf [nef@me-portland-us922.javanet.com] has joined #tunes
<Zhivago> re
!irq:*! i am considering glining *.av-el.co.il, any problems?
<Zhivago> natural language's imprecision may be essential for creative communication
07:20pm
!Teknix:*! gee,
!irq:*! yeah?
!Teknix:*! goforit
!irq:*! done allready :)
!irq:*! how about ns*.jp ?
!spice:*! do it
-:- SignOff liar: #TUNES (BRiX [http://www.qzx.com/brix] :: reboot)
-:- SignOff Downix: #TUNES (User abortion with 5 coathooks)
-:- liar [brand@p0wer.qzx.com] has joined #tunes
<liar> does anyone here have an smp linux box?
08:10pm
-:- beholder [beholder@ppp-023.m2-1.sub.ican.net] has joined #tunes
<beholder> ping
<beholder> Oops :)
<beholder> Hey all
<liar> hey
<liar> got any smp linux machines?
<beholder> liar: Have you ever used EPIC4?
<liar> no
<beholder> liar: If I had SMP boxes, I'd use em ;)
<beholder> liar: What IRC client do you use?
<liar> bx
<beholder> Ok, thanks :)
<liar> 3why?
<beholder> What do you need SMP boxes for?
<liar> i have one
<liar> im on a dual 550
<beholder> I've got a dedicated internet sharing server, and I'm trying out Epic4 right now
<beholder> Liar: Dual 550?!
<beholder> Liar: Wow, that must be nice :)
<liar> it was my game-tv machine for the last several months
<liar> but i figured that since win98 wasnt using the second processor...
<beholder> liar: Ahh, how does it perform in Linux?
-:- AlonzoTG [Alonzo@216-164-129-5.s259.tnt1.lnhva.md.dialup.rcn.com] has joined #tunes
<liar> well its faster than my 450 but i dont know if both cpus are doing stuff
<beholder> liar: Hmm... did you enable it in the Kernel config?
<liar> when i randomly change desktops with them all loaded it doesnt seem much faster
<liar> yes
<liar> they are both receiving interrupts
<beholder> liar: Hmm... you'd probably be better off asking someone in a news group, I'm sure lots of people have dual boxes.  I used to administer one a while back, but I never saw it without the second processor (and the load average never went high)
<liar> oh my load average isnt going above 50%
<beholder> liar: Maybe the bottleneck is your IDE drives?
<liar> even when i was doing the random high speed desktop change
<liar> i have 256meg ram so its all in ram
<beholder> liar: Hmm... have you pulled one of the processors out and tried it?
<liar> why would i do that?
<liar> i want one of those load meters like beos has :)
<beholder> liar: To determine what the speed is when it's running in single mode :)
<liar> i want brix THATS WHAT I WANT :)
<beholder> liar: I thought you used BeOS?
08:30pm
<liar> no, i have beos on my game-tv machine
<beholder> liar: How goes Brix dev?
<liar> well i spent the last 2 weeks writing bfe
<beholder> liar: Good :)  Oh, what termcap should I use when I telnet in? 
<beholder> liar: VT220, seems to be limited :(
<liar> have u seen bfe?
<beholder> liar: BFE? Nope not yet
<liar> www.qzx.com/bfe
<beholder> liar: I've only seen your concept shots on your page
<liar> its 99% functional
<beholder> liar: Gettin there, slow connection :(
<beholder> liar: What is Bochs?
<abi> Bochs is a PC emulator at http://www.bochs.com that runs on win32, un*x, beos, mac, os/2 and amiga
<beholder> liar: You wrote that frontend? Wow, very nice.
<beholder> lair: What's it written in?
<liar> C
<liar> it uses xforms
<beholder> liar: Xforms?
<abi> Xforms is, like, a widget toolkit at http://bragg.phys.uwm.edu/xforms
<liar> and if u think that image looks kewl u should try the real thing
<beholder> liar: I'm really not up on my Linux programming I guess ;)
<beholder> liar: You use Enligtenment?
<liar> ya
<liar> thats why i need dual 550 :)
<liar> i also use gnome :(
<beholder> liar: Hmm... I use a box, no keyboard or monitor, I assume I'd have to use a X-Emu on my desktop machine
<beholder> liar: Hmm... Bochs is part of the ports collection, that makes it easy to get :)
08:40pm
<beholder> liar: How well does it emulate Windows 95?
<liar> it dont emulate win95
<liar> it emulates a PC
<beholder> liar: That's what I mean, have you tried to emulate entire OS's on it yet?
<liar> no but others have
<liar> i just run brix in it
<beholder> liar: Sweet little dev tool. How fast is it?
<liar> up until today it emalated a 386
<liar> but now it does 486
<liar> and soon it will do pentium
<liar> instructions that is
<beholder> liar: Wow, they're really working on it
<beholder> liar: How fast?
<liar> not sure
<liar> if IPS=750k brix says 386-7mhz
<liar> when ips=5M brix = 52mhz
<liar> but it runs alot slower than 7mhz
<liar> and when ips=33M brix = 243mhz
<liar> er 342mhz
<beholder> IPS?
<liar> instructions per second
<liar> its a config thing in bochs
<beholder> ahh
<liar> damn my rxvt windows slide really fast
<beholder> Sounds very interesting. Will your front end work in FreeBSD?
<liar> almost threw the window off my monitor :)
<beholder> liar: :)
<liar> yes it will
<liar> if eihrul has his way it wouldnt
<liar> er had
<beholder> eihrul doesn't like FreeBSD?
<liar> he had me using some signal function that only works in linux
<liar> but now it uses io callbacks from xforms
<beholder> liar: Ahh.  that would not be good.  Same as the /proc FS... they have different info in both OS's
<liar> so it will work on anything that xforms works on
<liar> xforms dont work on windows
<beholder> liar: Who cares about windows ;)
<liar> hmm xosview stinks
<beholder> liar: How do you get the uptime of a system?
<liar> i will have to do what a few in linpeople suggested and find out how xosview gets it info and make a beos like meter
<liar> uptime
<beholder> liar: I need it in a constant format, uptime changes constantly
<liar> huh?
<beholder> liar: I wrote a script that awk's out the stuff I need, but it doesn't work when the system hasn't been up for a day, or when the minutes are 0x
08:50pm
<beholder> liar: Any ideas?
<liar> ask eihrul
<liar> he is good at that shit
<liar> he did that for bfe
<beholder> I'll ask him when he's on sometime
<liar> he is on now
<beholder> Eihrul: You here?
<liar> beep him till he gets here :)
<beholder> liar: ;) thanks :)
<liar> crap xosview is written in C++
<beholder> liar: yeah, what's wrong with that?
-:- SignOff AlonzoTG: #TUNES (Have Nice Day :))
<liar> other than it sucks and is hard to read?
<beholder> liar: Ahh, yep, C++ sucks for that :)
09:00pm
<liar> wow
<liar> i have my own bytemarks for a celeron 450, 550 and p2 450
<liar> the celerons are 1, 1.2 and 6.2% faster than the p2 in memory, interger and fpu respectively
<liar> funny how the celeron is that much faster in fpu
09:10pm
<beholder> liar: It's a good chip
<liar> u know where i could find out info about /proc/stat?
<liar> cpu  18251 0 19556 938899
<liar> cpu0 9212 0 9804 469337
<liar> cpu1 9039 0 9752 469562
<beholder> liar: Proc Stat isn't even on FreeBSD, so I'm clueless :(
<beholder> liar: Do you know what this permission is: "-r-xr-sr-x"  
<liar> ya
<liar> everyone can run the r/x the program but those in the group will run it as the user 
<beholder> liar: Ok, wasn't sure what the "s" was for
<liar> like if the program needs to be run as root u do chown root.group and anyone in that group can run the program
09:20pm
<beholder> liar: Ahh, never thoght to do that :) Very useful 
<beholder> liar: I'm looking for the code for uptime right now so I can make my own version ;)
<liar> and a nice security hole too :)
<beholder> liar: How do I interperet load average numbers?
<beholder> liar: They're always x.xx, whats the format?
<liar> process/minute
<liar> processes
<liar> the 3 numbers are timeframes
<beholder> liar: Ahh, same format for top then also?
<liar> ya
<beholder> liar: Yeah, 1min, 5min, 15 min or something
<liar> ya something like that
<beholder> liar: Mine are always 0.0x very small load I guess
<liar> ya so are mine
<liar> u need to be "using" the machine to have a load :)
<beholder> liar: It's used all the time, it's my NATD/Samba/FTP box.  
09:30pm
<liar> crap i gotta reboot. my damn cdroms arent working
<liar> looks like i plugged the cable in backwards :)
<liar> wonderful
<beholder> lair: That sucks :)
<liar> now that my win98 box has the single cpu mobo it cant find any drivers
<liar> so i had to do the "add new hardware" thing so it could reinstall all drivers
<liar> so it asks for the win98 cd
<liar> but it has no cdrom drivers
<liar> i love windows
-:- SignOff liar: #TUNES (BRiX [http://www.qzx.com/brix] :: reboot)
09:40pm
-:- liar [brand@p0wer.qzx.com] has joined #tunes
<liar> everytime i have booted i get a crapload of hda crc errors when linux gets to the login prompt
<liar> on this smp box
<beholder> liar: Hmm... not good :(
<beholder> liar: Is the HDD in good shape?
<liar> perfect
<liar> i got no errors on the other mobo
<beholder> liar: Hmm.. I think you're gonna have to post to a newsgroup to find out whats wrong.  Sounds pretty serious.
<liar> <SoundBlaster EMU8000 (RAM0k)>
<beholder> liar: That's and AWE-32 right?
<liar> i think that is an even worse problem
<liar> i have 8meg on that thing not 0k
<liar> :)
<beholder> liar: Did it detect before?
<liar> never had
<liar> has
<beholder> liar: Ahh, well not to worry about then :)
<beholder> mmm... terminal games...
09:50pm
<liar> i need a good linux game
<liar> that is multithreaded
<beholder> liar: Quake?
<beholder> liar: Q3 Test might be better
<beholder> liar: Not many games I know of take advantage of threading at a rendering level though
<liar> q3 is multithreaded
<beholder> q3 would be a perfect test then
<liar> at least it is on windows NT
<liar> i would assume it is on linux too
<beholder> liar: Makes sense. I hear it's programmed in ANSI C... 
<beholder> liar: I upgreaded my FreeBSD machine from 3.2 to 3.3, while it was running... 
<liar> so
<liar> i can that with linux
<beholder> liar: Yeah, just thought it was neat :)
<beholder> liar: I've never done it before.
<liar> i can upgrade win95 to win98 while its running too, just a couple reboots but technically its still running :)
<beholder> :)
<beholder> liar: I don't have to reboot though, which is nice :)
<eihrul> liar: only a couple...
<beholder> eihrul: I need to get the uptime of my system in a C program, any clue on how to do this?
<eihrul> hrmm, lemme poke around the syscalls
<beholder> eih: Ok, thanks :)
10:00pm
<eihrul> man 2 times
<eihrul> looks like it might do what ya need
<liar> beholder: told ya he would know :)
<beholder> eih: I'll check it out, thanks :)
<beholder> liar: :) 
<liar> eihrul: any idea what the cpu values in /proc/stat mean?
<liar> i am trying to make a beos like meter to show the load of my dual cpus
<liar> and xosview reads /proc/stats but i cant read the xosview code cuz its in c++
-:- lar1 [lar1@dialup-209.245.130.238.SanJose1.Level3.net] has joined #tunes
<liar> hey lar1
<lar1> Hey
<lar1> 64 bytes from 206.63.100.13: icmp_seq=13 ttl=242 time=33320.1 ms Loook at that!  God damn... I hate earthlinke!
<beholder> Hey lar1
<lar1> Hey, Liar, beholder
<beholder> lar1.. that's some serious lag  ;)
<liar> Sep 20 18:52:47 p0wer swapon: swapon: warning: /dev/hda7 has insecure permissions 0660, 0600 suggested 
<lar1> beholder:  Yup
<liar> how do i fix that?
<eihrul> liar: nope
<lar1> liar: chmod 600 /dev/hda7?  I dont' really know
<liar> nope what?
<eihrul> liar: negative on /proc/stats
<liar> oh
10:10pm
<liar> how do i remove a kernel module
<lar1> liar:  There isn't an instruction history in bfe?
<liar> lar1: not yet
* beholder/#tunes is playing the "quiz" terminal game and finds it sad that he doesn't know Canadian capitols...
<lar1> liar: Hmmm, that would invole just storing all previous instruction in a buffer (array of strings)? right?
<liar> lar1: no
<lar1> liar:  How would you do it?
<liar> there will be a list box, u just add the current instruction to the list, when the list reaches a given limit u remove the top line in the list
<liar> so if the user wants a 100 instruction history then u keep adding until u reach 100 lines then u just remove the top line before adding new lines
<liar> i could code that up in 5 minutes
<liar> u wanna do it?
<lar1> I am thinking so
<lar1> I'd have to lean about lists in Xforms
<liar> ok u need to define a new form in forms.inc
<liar> fl_add_browser(list, char *)
<lar1> Should instruction history be another one of those 'tabbed' windows?
<liar> fl_remove_browser_line(list, 1)
<liar> yup
<liar> then find the code in main where i add tabs and add a history tab
<lar1> Ok, you want to send me the latest bfe?
<liar> its on the page
<liar> i can dcc it
<lar1> Thanks mucho
10:20pm
<lar1> Ouch... that was slow
<eihrul> liar: anyothing new in BFE today?
<lar1> Where did you move my ConfirmQuit() function too?
<liar> buttons
<lar1> ok
<liar> eihrul: no i upgraded my machine
<lar1> Where do you want the code for history?
<liar> the update code should go in updates and the form code should go in forms.inc
<lar1> Uhhh what form code?
<liar> that code that defines what the tabbed window looks like
<liar> u can pretty much just copy the structures form
<liar> u will need a new object, i_list
<lar1> ok
<liar> oh and put the addresses in there too
<lar1> Ok
10:30pm
<lar1> Should I mark my changes like last time or no?
<liar> no
<lar1> ok
<liar> i will just extract yer stuff over mine
<lar1> When I am done should I just send you a tarball of all the source as I have it?
<liar> oh crap wait
<liar> i made changes this morning
-:- core [core@core.suntech.fr] has joined #tunes
<lar1> Hey core
<liar> let me send u a new copy
<lar1> Ahh!  Ok ok
<liar> give me a minute to add my memory code
<lar1> Ok
<liar> im just copying it over from my pilot
<lar1> You write code on yer pilot?
<liar> ya
<lar1> Where did you wirte this code?  At the cafeteria at work? ;)
<liar> at my desk at work
<lar1> Okay
* lar1/#tunes thinks the palm pilot is just a social status object, much like SUVs...
<liar> hey they are useful
<liar> i use mine everyday
<lar1> I didn't say they wernt useful
<lar1> But $500 for them?  They aren't worth that much in parts and code methinks... 1000% profit margin for 3com
<liar> they are worth it
10:40pm
<lar1> Hmmm, what does the pilot do that the $150 one doesn't?
<liar> what $150 one?
<lar1> ls
<lar1> Oops
<lar1> liar:  All those other 'pilot-like' orginizers
<liar> the pilot has the palmOS
<liar> the others have wince
* beholder/#tunes thought Palm's were toy's till he saw sysadmins programming routers with em
<lar1> Hmm,ok,whatever... perhaps I don't understand because I am not in a position to need one
<lar1> although, if mom got me one, I'd take it ;)
<lar1> WHo is coding Clemintime?
<beholder> Not sure :(
<liar> core
<liar> is
<lar1> Ohhh, ok, I was confusing emk with it
<liar> eihrul is coding emk
<lar1> Yes
<lar1> eihrul:  how does emk pass messages?
<core> hi :) sorry, i didn't know i was on :O
<core> hey lar1, liar, can you take more confusing nicks? ;)
-:- lar1 is now known as lair
<lair> core: no
<liar> heh
<lair> core: :)
<eihrul> lar1: message.s
<beholder> gotta go later :)
<lair> later beholder
-:- lair is now known as lar1
-:- SignOff beholder: #TUNES (ircII EPIC4pre2.003 -- Accept no limitations)
<lar1> eihurl:  Yes, I see the function to _recive_ the message...but howdoes the thread know it has a message waiting?
<eihrul> if the thread doesn't have messages waiting, or the receiver isn't waiting for a message
<eihrul> it's blocked until one of those conditions becomes true
-:- SignOff hcf: #TUNES (Leaving)
<lar1> eihurl:  So threads must be contantly be fed with messages?
10:50pm
<eihrul> no
<lar1> But you just said if there is no message the thread is blocked
<core> lar1: the new ravenOS page isn't visible yet? :)
<eihrul> lar1: it is unblocked when someone tries to sed a message to it
* lar1/#tunes grins wide
<lar1> core:  Nope :) sorry
<lar1> core:  I need someone to nag me about it on a pretty daily basis
<lar1> eihrul:  So you are saying if no ones is sending a message then the thread it blocked?
<eihrul> only if it tries to receive a message
<eihrul> and there are no threads blocked waiting for that thread to receive
<lar1> So a thread has to call the Getmessage functino or whatever inorder to get its next message?
<eihrul> yes
<liar> YES, bfe dumps memory now
<eihrul> if there are no messages waiting, it's blocked... and when some thread does a MessageSend, it's unblocked
<lar1> liar:  Very, very cool!
<core> lar1: i can put a "nag lar1" job in crontab if you like :)
<core> lar1: so, did you make any progress on it since the last 2 minutes? :)
<eihrul> cron is for weenies
<lar1> core:  Not really :) I do like the crontab job idea though
<core> liar: you never sent me a brix binary or sources like half of the channel seems to have :P
<lar1> core:  set it for every hour :)
<core> lar1: be careful what you wish for :)
<lar1> core: Uh oh....
* lar1/#tunes shuts mouth
<core> hehehe
<core> don't tempt me, i do have boxes on a permanent link :)
<core> hmm.. muscles hurting, throat sore, sneezing.. i guess i show most symptoms of flu :P
<lar1> eihrul:  What if thread a wants to tell thread b somthing but thread B hasn't called getmessage because it doesn't know anyone wants to talk?
<eihrul> thread A is blocked
11:00pm
<eihrul> until thread B tries to receive a message
<eihrul> i keep a list of all threads with messages waiting to be received
<lar1> So all thread are required to call getmessage everyonce and a while?
<eihrul> they don't have to...
<lar1> core:  Hmm, I have a headache
<eihrul> just if they don't, they won't receive messages
<core> lar1: i've had a latent one since friday :/
<eihrul> most often, threads will simply block trying to receive any message
<lar1> core:  That sucks
<eihrul> use messages to do some transaction
<core> eihrul: and all threads that send messages to them will stay frozen forever :)
<eihrul> then the server goes dormant again waiting for another request
<eihrul> core: the breaks of life :)
<eihrul> core: i may implement block timers, i may not... but as i've said, i'm not trying to make this perfect
<core> eihrul: yeah i know.. or you would have implemented thread migration :)
<lar1> eihrul:  But what if the system want thread z to termanate... and thread z ain't getting its messages... then what?
<eihrul> then the threads are unblocked
<lar1> Thread migration?
<eihrul> and returned an appropriate failure 
<eihrul> value
<lar1> eihrul:  Which threads?
<eihrul> any threads blocking on another thread that gets destroyed
<core> lar1: the client thread detaching from its context in the client and attaching to a context in the server, servicing itself, then returning into its home base into the client. this way, you needn't have a big pool of threads listening for requests in the server, and work is accounted to the client, as it should be :)
<eihrul> core: but umm, that's evil!
<eihrul> is that not insecure?
<core> eihrul: why?
<eihrul> giving a client complete access to server data?
<lar1> Thats mean!  Just cause a thread tryed to send a message youre gonna kill it!  Hmmph... I'll have a word with the ASPCT!! heh
<core> eihrul: no.. the client has no way to touch the server. it connects to an entirely new context (stack and everything)
<eihrul> lar1: i said unblocked, not destroyed
<eihrul> core: but how is it a "thread" in the server's context? how does it access server data?
<core> eihrul: no.. you don't. it doesn't access anything. it executes the request in a new context in the server, one it has no access to.
<core> eihrul: the thread is split in two: the thread proper (priority, time slice, kernel stack, etc.), and an "activation" (address space, registers, user stack, etc.)
<eihrul> but it's the "activation" i'm concerned with here :)
* lar1/#tunes is pretty sure message passing, migration and all that other stuff is surely the tool of the devil
<eihrul> how does it interact with server data having no access to server data?
<core> eihrul: when doing sync IPC, the thread connects to a new activation in the server, executes the server code in the context of the server (ie. _no_ client code is executed), then pops the server activation off the activation stack, and comes back into the client
<core> eihrul: you don't get it. the activation then executes server code, in the server context, so the server code does have the same access to server data as any other of its threads :)
<eihrul> so we're talking something like sun "doors" here?
<lar1> I need to study for el examen a la manana!
<core> eihrul: we're talking something like Mach's thread migration, implemented by bryan ford and jay lepreau (the os kit :)
<core> eihrul: the original point is that sync IPC becomes something like 3 or 4 times faster, not having to block and unblock any thread :)
<eihrul> but are we talking "doors" here :)
<lar1> core:  You are speaking in abstract terms... Ambigous operators require parenthesis!
<eihrul> core: which sound vaguely similar to what you're describing
<core> eihrul: NFI what those are :)
<core> lar1: er? :)
<core> eihrul: anyway it's implemented in clementine and works fine ;)
<eihrul> core: a server sets a thread entry point, then other processes can then when given rights to the door, "step through the door", into the entry point defined for that door, do the work, then return home :)
<lar1> core:  Server, activation, client, activation stack?  What?
<eihrul> core: that was the main idea of doors....
<core> eihrul: right, except there are many 'doors' then (as many as there are exported calls in the interface)
11:10pm
<liar> if(*(s--) != '\t'){ *s = '\t'; s++; }
<lar1> core:  How does sync. IPC work?  Implemntation wise?
<core> lar1: just discussing IPC :)
<eihrul> core: though... how do you support parameter passing?
<liar> oops
<core> lar1: that's what i just described :)
<eihrul> core: do you just instead of copying data to an arbitrary location, copy it into the stack?
<lar1> core:  Well, I didn't understand a word of it :)
<core> eihrul: right :) and identifiable data zones (strings or data with a known length) are copied or mapped as usual.
<lar1> thanks liar
<eihrul> core: how are result values propagated?
<core> lar1: instead of marshalling a request into a message, blocking the client thread, sending the message to the server that has to be listening, wake it up, give it the message, unmarshall it, call the right function .. the client thread simply migrates into the server.
<core> eihrul: returned into EAX ;)
<eihrul> core: okie...
<core> eihrul: the thread of execution proper, isn't broken by the scheduler, so you can simply do that :)
<liar> can anyone tell me why it chops off the last digit off the address in the memory dump? its in updates.c:update_memory()
<eihrul> that does sound nicer than synchronous IPC....
<eihrul> as you don't need a sending thread and a receiving thread
<eihrul> just a sending thread
<eihrul> i'd go so far as to say i almost like the concept better than synchronous IPC :)
<lar1> core:  Ok, makes sense, what are you defining as 'server'?
<core> eihrul: it avoids the big pool of listening threads in servers, too
<eihrul> core: > as you don't need a sending thread and a receiving thread
<liar> lar1: damn u accepted first and finished last :)
<core> lar1: oh, any component that provides functions to the outside world. most can be both client and servers
<core> eihrul: right :)
<lar1> liar:  My connection is pretty crappy tonight
<core> liar: now send brix :)
<eihrul> you go and tell me this after i have messaging already implemented damn it! :P
<eihrul> i must beat thou with a rolled-up-hardwood-floor
<core> eihrul: it can be implemented later.. ford and lepreau implemented it in Mach after version 3.0 :)
<eihrul> did i not say there might not be a later with emk? :)
<lar1> core:  How does thread A say "Hey, wanna go to a movie with me" to thread b?
<core> there are a lot of other nice properties to that too
<core> the fact you can propagate a 'cancel' request down the activation chain..
<liar> lar1: heheh
<eihrul> core: do you still have "normal" sync messaging?
<core> the fact that cpu time is accounted to the client.. :)
<core> eihrul: yeah, if explicitly called but nothing calls it right now (and actually the necessary code isn't checked in cvs, it reminds me :)
<lar1> liar:  I'm a teenager... I suppossed to be obsessed, right?
<eihrul> is all the information needed to return to the sending context stored on the user stack?
<core> lar1: there is no thread B.. it just goes to the movie alone :)
<core> lar1: it just goes with thread B's car :)
<lar1> core:  Nah, thats only in my life :(
<core> eihrul: user stack == bad. no, activations are simply in a linked list. when exiting from an activation, it returns into the previous one in the list, if none is left, the thread is dead.
<eihrul> core: i know they're bad... that's why i asked, because the server context could manipulate the return information and do evil things
<core> eihrul: actually it can't
<lar1> L-l-l-l-linked list?!?! Ahhhh!  Core speaks of the unholly data type!
<liar> damnit, someone fix my code :)
<core> eihrul: this method gives neither the client nor the server, more privileges than usual :)
<eihrul> i'm saying if the return information was saved on the user stack :)
<core> eihrul: right.. but that's not an option :)
<eihrul> indeed
<liar> core: how does the client thread "migrate" its ass into the server?
<lar1> liar:  When ever I read bfe I keep thinking big friggn enviroment!  Its weird
<core> lar1: linked lists are excellent if used properly, it avoids things like "maximum 26 drives" :P
<eihrul> core: i bets it fun when the client context destroys itself while the server context is activated, no? :)
<liar> lar1: i also think of something else but it aint that ;)
<core> liar: the thread proper is just a priority, kernel stack etc; it needs an activation to run; the activation is private to an address space. it just migrates by connecting to an activation in the server address space :)
<lar1> liar:  Heh!
<liar> core: eh?
<lar1> core: Hmm, well, I had a bad experence implementing linked lists.. so I don't like them
11:20pm
<eihrul> core: you have to allocate a temporary stack for the thread in the server context, right?
<core> eihrul: in this case, the request is silently cancelled and return values are ignored.. the thread proper is only deleted when the request returns from the server
<core> eihrul: right, except in the spirit of what i told you yesterday :), activations aren't deleted immediately, since ipc is usually a ping-pong.. only when needed. which makes it fast :)
<eihrul> ah
<eihrul> i'd have to restructure way too much of my microkernel to support that
<eihrul> since thread stacks are kept directly in the thread structures themselves :)
<core> lar1: i have standard macros for doubly linked lists, and usually also use single linked ones, no problem with that :)
<core> eihrul: as i said it can be done later. clementine did work without migration for a while :)
<eihrul> core: stop throwing that word at me! :P
<core> eihrul: what, the l word? :)
<eihrul> yes!
<core> hehe
<liar> heh
<eihrul> i'd rather have clemetine do that stuff than EMK
<core> eihrul: diversity is good :)
<lar1> heh
<lar1> core:  When will you relese clemintine's source?
<eihrul> but when diversity is maintained by a single man, it gets tiresome!
<core> lar1: soon.. this fall. we're getting there i guess (today is fall :). i just miss one functionality to implement, to make a complete prototype people can play with.
<core> eihrul: tell me about it :)
<eihrul> core: that being?
<core> eihrul: me for clementine (at least i was for most of the time)
<eihrul> > i just miss one functionality to implement
<lar1> I miss one functionality to implement as well
<lar1> I need to implemnet the os
<eihrul> core: that being? :)
<core> lar1: heheh
<core> eihrul: physical disk drivers :) i do have components for filesystem abstraction, i do have them exporting a dynamic filesystem of their properties, i can browse it and all, i just miss persistence :) that, and getting a legacy unix application to run, which will happen at the same time since someone _else_ is working on it :)
<core> (and is making good progress.)
<lar1> core:  Perhaps when I write the Raven page I'll take you up on some of that space so I can set up a mirror :)
<eihrul> core: just don't take a year...
<core> eihrul: nope :)
<core> lar1: you can have it :)
<eihrul> core: or else i'll be forced to implement doors....
11:30pm
<lar1> core:  Ok, let me know when it becomes available
<core> lar1: sometime today. actually, everyone who needs webspace, emails, a cvs repository or ftp access for his project can have some.
<core> lar1: i got the server hardware yesterday, i get the rack this morning, i'll get one of my cow-orkers to put the former into the latter, set it up, and voila :)
<eihrul> core: curiously, in practice, what are the relative costs to transfer from client context to server in thread migration and sync IPC?
<liar> cow-orkers? :)
<core> liar: yep :)
<lar1> liar:  Its a new milk harvesting method
<core> haha
<lar1> lier: orking the cow is much better
<liar> core: hey are u aware of any idea bugs with smp in kernel 2.2.7?
<core> eihrul: thread migration is faster :) for actual benchmarketing, read bryan ford and jay lepreau's paper in .ps about thread migration :)
<liar> s/idea/ide/
<core> liar: 2.2.7? probably, upgrade to 2.2.12 :)
<core> liar: but none i'm aware of in particular, though
<core> liar: ide bugs.. sounds like overclocked SMP
<liar> i get a crapload of hda crc errors and then is says it disable dma
-:- _ruiner_ [nate@ppp367.wi.centurytel.net] has joined #tunes
<core> yep, overclocking allright :)
<liar> hey so what
<liar> 366 to 550
<core> OUCH :-)
<core> gee, i really wonder why it doesn't like it ;)
<liar> it shouldnt give me errors
<lar1> core:  Celerys can do that!
<liar> ya it will do 567 in single cpu mode but not smp
<liar> beos wont boot at 567 or 550 in smp
<liar> i have to use 460 in beos
<lar1> liar:  what can it do smped?
<core> liar: well, ultradma won't like it.. linux warns you it turns it off, probably windows does it silently
<core> ok, i really need to dash to work. see you :)
-:- SignOff core: #TUNES (ircII EPIC4pre2.003 -- Accept no limitations)
<lar1> later core
<liar> great so i dont get udma
<liar> that bites
* lar1/#tunes always finds it cool that while I'm about to sleep, somebody is going to work
<lar1> liar:  When you said fl_add_browser(list, char *) what is list?
<lar1> ermind
<lar1> nevermind
<eihrul> core: where can i find this paper?
<lar1> liar:  Why doesn't udma like overclocking... the bus speed too much?
<eihrul> core: also, if you have thread migration across a network, does the net-effect not just become synchronous messaging of sorts (you have to create a thread on the other side)
<liar> core left
11:40pm
<eihrul> sheet
<lar1> liar:  How do I know how long the history should go on for?
<eihrul> just when i have questions, he leaves!
<eihrul> he always does it, without fail!
<lar1> eihrul: Heh
<liar> ugh i gotta crash
<liar> lar1: have fun with the code
<liar> eihrul: hey my memory dump has a bug
-:- BlackPhoenix [black-phoe@adm.univd.kharkov.ua] has joined #tunes
<eihrul> liar: but of course! :)
<liar> in updates.c:update_memory() it is chopping off the last digit of the address
<lar1> Liar:  How do I know how long the history should be?!
<liar>                         if(*(--s) != '\t'){ *s = '\t'; s++; }
<liar> i think its that line
<liar> well i know it is
<lar1> Thats a very screwy line
<liar> add a history_length variable
<liar> and we will set it in the config file
<lar1> liar:  What should its deafault be?
<liar> 50?
<lar1> liar: ok
<eihrul> that code is umm... sloppy man :)
<liar> eihrul: duh
<lar1> eihrul:  Whats *(--s) mean?  the address of s -1?
<eihrul> *--s
<eihrul> S = S - 1; *S
<liar> does it actually dec it
<lar1> Ewwww, thats nasty code
<eihrul> yes
<liar> crap
<_ruiner_> what else would you expect from liar?
* eihrul/#tunes cackles.
<lar1> Heh ruiner
<eihrul> ruiner: i haven't seen your code yet, so :)
<liar> lar1: change that line to *(s-1)
<eihrul> s [-1]
<lar1> liar: Uhhhh, Where is that line?
<liar> *(s-1)
<_ruiner_> my code would ruin you for all other projects
<eihrul> liar: s [-1] damn it
<_ruiner_> its just that sweet
<liar> updates.c:update_memory() 
<eihrul> ruiner: i doubt it
<liar> line 31
<eihrul> s [foo] is preferrable to *(s + foo)
<eihrul> and indeed equivalent
<lar1> liar:  Are we going the QZ style or the eihrul style?
<eihrul> and is a savings of two chracters as well
<liar> qz style
<liar> :)
* eihrul/#tunes frowns.
11:50pm
<lar1> liar:  You sure?  Ok.... *(s-1) it is
<eihrul> lar1: don't litsen to him!
<liar> fine do it his way
<lar1> Now that I saved and recompiled...
<lar1> liar: fine :)
<lar1>  *s[-1] it is
<liar> no
<liar> s[-1]
<lar1> Oh :) ok
<lar1> hh it didn't like that
<liar> hmm once the intruction history is done and resize problem is fixed bfe will be ready for release
<liar> it does like it
<liar>                         if(s[-1] != '\t'){ *s = '\t'; s++; }
<liar> it works fine here
<lar1> OHHHHHHH, how stupid, I left out a parenthsis
<liar> hhaha
<lar1> if s[-1] != '\t'){ *s = '\t'; s++; } :)
<lar1> Ok, all better now
<lar1> liar:  I will finisht the history tommarow... I'm not sure how to fix the resize problem
<lar1> Whats that for?
<liar> the history form
<lar1> I already have it
<lar1> I made it
<lar1> I just need to add that code to updates.c
<lar1> !gaim &
<lar1> Oops
<eihrul> dern it... can't find the paper on threa dmigration
[msg(TUNES)] newlog 1999.0921
IRC log ended Tue Sep 21 00:00:00 1999