IRC log started Sat Sep 11 00:00:01 1999 [msg(TUNES)] permlog 1999.0911 -:- MAxD [m.dentico@a118.teseo.it] has joined #tunes -:- MAxD [m.dentico@a118.teseo.it] has left #tunes [] -:- hcf [nef@me-portland-us1037.javanet.com] has joined #tunes -:- NetSplit: irc.linux.com split from forward.openprojects.net [07:44am] -:- BitchX+Deb1an: Press Ctrl-F to see who left Ctrl-E to change to [irc.linux.com] -:- Netjoined: irc.linux.com forward.openprojects.net -:- abi [nef@bespin.dhs.org] has joined #Tunes -:- hcf [nef@me-portland-us1037.javanet.com] has joined #Tunes -:- RAVI [user4659@ts1-4.colosseum.com] has joined #Tunes -:- RAVI [user4659@ts1-4.colosseum.com] has left #Tunes [] -:- SignOff hcf: #TUNES (Ping timeout for hcf[me-portland-us1037.javanet.com]) -:- eihrul [lee@usr5-ppp28.lvdi.net] has joined #tunes -:- pyro [tcn@cci-209150250058.clarityconnect.net] has joined #tunes -:- ElGato` [no@209.68.229.220] has joined #tunes lo hi bbl -:- SignOff ElGato`: #TUNES (Leaving) 09:50am hey re 10:10am -:- Icore [ahvezda@kalypso.cybercom.net] has joined #tunes -:- SignOff Icore: #TUNES (Leaving) -:- SignOff pyro: #TUNES (pyro has no reason) -:- lar1 [LarMan@dialup-209.245.143.220.SanJose1.Level3.net] has joined #tunes Hello lo eihrul: Hows the Bochs thing going? that's liar's project I thought you were doing it with him i just wrote some code for him :) :) 12:10pm u? Huh? Oh I am still reading up on Linux programming Slowly, its been real crazy over here, I transfered schools and I am quite tired * lar1/#tunes thinks he needs a sub-woofer Ohhh, that mp3 just wasn't bassy... Much better!!! My laptop is shaking from it you sure it's just not a broken drive? Well not shaking, but you can feel the bass in it :) Hmmm, Born to be wild lacks a good bass line 12:20pm ah, no wonder i got no occurences... i was grepping for ecPageShit when i should of been grepping for ecPageShift... stupid typos Hehe Thats a cool typo! yessir Have you gotten Bochs to compile? yessir With the debugger? i used an older version, as per brand's advice But don't the older versions have pretty imortant things missing? dunno 12:30pm ok Wow! I just found a spelling test from 3rd grade... 33%... ouch * lar1/#tunes wonders if it is a good idea to run his external drive sideways prolly not 12:40pm i know my external laptop drives emit junk data if you even nudge them while doing reads :) which is minorly annoying Ewww, yes I can't get it to stand in a way I like anyways... oh well I need to move a network cable... but if I do my mp3 will stip Ahh! the pressure! Just do it. Wow... it didn't really skip much! Cool! Grrrr, I moved all this stuff just so I could put my light in a better palce... but it was better before!! 12:50pm -:- lar2 [LarMan@dialup-209.245.142.187.SanJose1.Level3.net] has joined #tunes -:- SignOff lar1: #TUNES (Ping timeout for lar1[dialup-209.245.143.220.SanJose1.Level3.net]) Stupid modem What did I ever do to it? -:- lar2 is now known as lar1 01:00pm Putting a finishing nail into a stud with the back of a screwdriver doesn't work damn it... i forgot all about the EBP register i had a whole 7th register to use What are you coding? my kernel :) Which OS is yours? EMK Whats it stand for? eihrul's microkernel Ohhh I was close... I was thinking Efficent MicroKernel That the name of the OS, or just the kernel? microkernel i could care less about the OS... just making emk Isn't a uk useless with out the rest of the os? 01:10pm i'm not trying to make it practically useful i'm just trying to get a working uk, after that, i'd rather focus my efforts on tunes more besides, Linux is more than good enough as an OS Linux is good for the buisness world, but for the game hungry home user, the speed just isn't there what the fuck are you talking about? What aren't I talking about? "speed" i couldn't give a shit about speed in my OS The avergage joe is going to want speed A) linux is stable, B) it's a very good development environment, C) it does everything i need for now the average joe doesn't even know how to turn his computer on Thats not what I mean... the average joe just wants to play quake Yes, and Quake in X would be a joke i'm not concerned with him... since i don't play quake, my needs are different i do real work on my computer That the difference... I would like to play quake. I also like to do real work. I want an OS that is stable and fast and Linux is just that if you don't like X, make something better linux itself is not slow I don't like uKs It takes forever for linux to start up... on some of my boxes, windows loads faster so? I don't like waiting I want click, there it is A) linux doesn't blue screen, B) the applications are generally better, C) you rarely ever need to reboot linux that's your problem, not mine I never said that Linux is bad... I _love_ linux But I would like the power of linux in a little faster envrioment true so try BeOS No apps for it no apps for your OS either so what's your point? 01:20pm and seeing as BeOS is in a state where you can actually develop apps for it :) But I can makes apps for my OS, while I don't really feel like learning how to code in yet another way (in the same language) umm... beos uses gcc My point being, You code for Linux console differently then you do for dos and you code different for Windows then you do X. I assume that BeOS will be different as well -:- pli [per@spinner.rsn.hk-r.se] has joined #Tunes -:- pli [per@spinner.rsn.hk-r.se] has left #Tunes [] so? hows your OS going to be any different? learning is a part of life... and from what i've heard, the BeOS is a good deal nicer and saner than either Windows or X or DOS Hrmmm, perhaps I am just being lazy perhaps so I can justify not using BeOS just as well as I can justify using asm Oh, that reminds me, I had a thoght about asm As you said, C(++) code will 'evolve' as the compiler gets better But shouldn't the asm compiler optimize and 'evolve' the code as it gets better as well 01:30pm -:- SignOff abi: #TUNES (Ping timeout for abi[bespin.dhs.org]) no assembly language is not a high level language, if i specifically tell the assembler to use an instruction i EXPECT it to use it the best an assembler can do in light of those constraints is produce the best encoding of the instruction it can, which in itself is not hard -:- abi [nef@bespin.dhs.org] has joined #tunes But it _could_ be made to optimise? even if you had a weirdo optimizing assembler, anything it can do is not comparable with the optimizations a C(++) compiler can do the lower level your specification, the harder it becomes Hm, I see what you are saying i.e. a C/C++ compiler can use whatever registers it wishes without messing things up because no registers are specified in the code but an optimizer for assembly language would have to be careful to make sure all desired results are in specified registers, etc * lar1/#tunes adds optimizing compiler to list of things to do... there is nothing wrong with writing in assembly language, just make sure you have the right reasons for doing it, dude! :) What are the 'right' reasons? i can't tell you general reasons because there aren't any i'm doing so for personal opinions which may not apply to you at all Hmmm i just don't like C... it's a horrible language, and i don't like interfacing assembly and C code Why don't you like C? * eihrul/#tunes cackles. take a guess 01:40pm Is c++ okay by you book? it is better, but inherits all of C's problems so it has C's problems PLUS C++'s problems I don't see any problems with C besides lack of Object orentation... And the only thing I see with C++ is a few quirks of flexability BRB - food -:- lar1 is now known as lar_eating you're not used them enough then :) -:- NetSplit: varley.openprojects.net split from irc.linux.com [01:46pm] -:- BitchX+Deb1an: Press Ctrl-F to see who left Ctrl-E to change to [varley.openprojects.net] -:- Netjoined: varley.openprojects.net irc.linux.com -:- abi [nef@bespin.dhs.org] has joined #Tunes 01:50pm -:- lar_eating is now known as lar1 Heh Tu hablas espanol? si, yo hablo el espanol Bueno! Estudio espanol en la clase "Spanish 3-4"! -:- NetSplit: varley.openprojects.net split from irc.linux.com [01:59pm] -:- BitchX+Deb1an: Press Ctrl-F to see who left Ctrl-E to change to [varley.openprojects.net] 02:00pm -:- Netjoined: varley.openprojects.net irc.linux.com -:- abi [nef@bespin.dhs.org] has joined #tunes Whats up with abi? Shes like on the fritz abi: lar? wish i knew, lar1 abi: x to french baugette avec fromage lar1: fromage d'avec de baugette abi: x from french baugette avec fromage lar1: Nevermind Should my UPS power my external modem too? 02:10pm I have to go... I'll be back later tonight mabye -:- SignOff lar1: #TUNES (RavenOS -- Its my OS, I'm gonna plug it :)) 02:20pm -:- liar [brand@p0wer.qzx.com] has joined #tunes liar: done yet? 03:50pm done? with bfe? no im still trying to get yer code to work i checked the return value and it had failed 04:10pm you have to remember to skip the prompt and actually, you have to skip all the initial junk that bochs outputs to even get to the prompt, initially :) how? want me to write a routine to do it? ya okie dokie actually, i don't have to :) just do a prompt_read (readpipe, NULL) initially and after every command you need to do that as well ok btw in string_read() u have a while(){}{} loop. wtf is the second brackets for? it's only one loop -:- kryogenik [groove@ppp05b.nobel.com.br] has joined #Tunes meaning what? you can declare a block anywhere in the C language so that you can declare variables as needed i declared that block so i could have that tmp var i didn't want to declare it locally so the compiler wouldn't bungle up the first loop or atleast globally to the procedure ah i want it in a register, but not at the cost of the first loop :) i see though i dunno if it even makes a difference to gcc ultima said it was a c++ thing what was a C++ thing? the brackets no... the deal with C++ is that you don't need to declare a block to declare a variable that's not at the beginning of a procedure in C, you can declare a variable so long as it's at the beginning of the block, and you can stick blocks anywhere :) you can do it in C++ too, but you don't need to 04:20pm damn, cli/sti is expensive why? this manual says 7 cycles, on original pentium's probably slower on most intel's cpus now take a look at prompt_read and tell me if its correct that for() dont look right :) for(inputChar = fgetc (file); what's not right with it? -:- core [core@core.suntech.fr] has joined #tunes people hi hey core hey qz^Wliar :) liar: you never sent me your brix image :P damn... took you only 1+ weeks to get back uhh then yer gonna have to find someone that has it cuz i dont have it anymore :) i've forgotten my question already! liar: you can't compile a working image of your own os? :) (or assemble anyway) not right now 04:30pm i work on it, fix it up and make a binary, work on it some more, fix it up... right now im in a work on it stage and you delete the binaries? how efficient :) actually im in a make better tools to help me work on it stage * eihrul/#tunes tries to remember his question for core. yeah i sort of saw that in the logs.. even if i'm not sure how your front-end will help :) no i dont delete the binaries they get deleted when i rebuild a non-working version hi eihrul, i don't know what it was either :) it was something to do with scheduling methinks... liar: well, i mean, like, keeping the working binary and sources somewhere, so you have some point of reference of something that works, in case it doesn't anymore eihrul: why, did i say brb and leave for a week? :) core: it will display the contents (with names) of every variable/structure in the os you disappeared in the middle of me asking questions eihrul: oh; sorry.. it's probably that annoying "real life" thing that popped up again ;) (that, or females) liar: oh, gdb linked over the serial port does that for me, actually it even follows my sourcecode :) sheesh... it takes me 40 instructions just to set up the mappings for IPC -:- SignOff kryogenik: #TUNES (changing servers) the actualy copy takes about, oh 5 instructions liar: ddd does that graphically on top of it, i can watch the internal structures of the components being modified as the system runs. (and yes, in the memory of the _remote_ box :) ddd? liar: data display debugger; a graphical front-end to gdb core: here's a new question: how do i handle two messages in the same quantum without a TLB flush :) liar: it's very cool, even though i usually end up typing commands in the gdb window anyway. or is that why you're allocating address space on demand for messages? gdb doesnt do asm eihrul: uh, two sync messages? you don't what do you mean you don't? :) liar: it sure does, if you can cope with at&t syntax i cant :) eihrul: if your thread sends two sync messages, there is no "in the same quantum" thing right now i flush the TLB on messages > 1 no... two messages can be received liar: well, that's your problem then :) i can (with some mental effort :P), and it works fine.. :) and does it handle lists and shit like that? liar: most of the system is C anyway, and i test code extensively before i go to the upper layer, so i don't have much debugging problems, anyway. -:- lar1 [lar1@1Cust244.tnt4.sfo3.da.uu.net] has joined #tunes liar: yeah, it does. graphically :) hi hey lar1 :) liar: You run E, right? Hola core oh but u have to run it over serial to another machine liar: ddd is a quite neat front-end to gdb; i have patched gdb's bfd to handle XCOM, and clementine has a remote gdb representative, so i just need to run clem on my guinea pig box, and gdb on my linux host, and it debugs at source level, and displays all structures :) this will run over bochs on the development machine liar: yeah, of course; you need two boxes. but it's very rare that i have to actually use a debugger; i have assert's and such built into all code i write. liar: yeah, so does clementine when compiled with assertions etc. on; it will even print to the console bochs was started on, in case something bad happens to the page mappings and you can't write to the emulated screen anymore ;) -:- SignOff eihrul: #TUNES (Ping timeout for eihrul[usr5-ppp28.lvdi.net]) -:- supahman [shayne@pm162-108.dialup.netidea.com] has joined #tunes (that's what my --enable-port-e9-hack that made it into the standard bochs code, does :) core: you run bochs? lar1: yep. i even participated to it some. 04:40pm (added --enable-port-e9-hack, improved the keyboard and hdd emulation) core: Hmm, I can't get it to compile the debugger in lar1: the latest versions are a little rushed. core: ya since u helped with bochs we shall all complain to u about how it wont compile :) lar1: there is also some trouble with hddrv.c on win32 and others; -:- supahman [shayne@pm162-108.dialup.netidea.com] has left #tunes [] core: What version do you reccomend liar: hehe, well, i've made quick and dirty fixes to compilation problems, but i'm not recommending them to anyone :) liar: I'm just asking him cause he runs it, not cause he devloped it why wont anything newer than 981102 compile? lar1: 990718a, since it can run 486 system code finally. liar: 990718a compiles for me, except for the debugger. core: Ok, thanks ya i got 990212 to compile and a couple others but they either didnt have debugger or color lar1: having to ifdef out instructions like invlpg just to run in 386 mode for Bochs was quite annoying (it's built into my work, but it's annoying to recompile all just to test on a real box at real speed) liar: color works for me, but the debugger requires major hackery to compile on any 99* version core: Yes core: do u know why that is? liar: also the latest versions' serial port code doesn't compile on linux anymore, unless fixed by hand. liar: yes - kevin is 100% on freemware and doesn't pay much attention to bochs quality anymore.. IMHO figures (serial port code, as in, what lets you simulate a serial port with the start console of bochs) -:- eihrul [lee@usr5-ppp28.lvdi.net] has joined #tunes -:- pyro [tcn@cci-209150250124.clarityconnect.net] has joined #tunes So if I want the debugger, I need to run a 98 version? lar1: no, you can fix the code, it's just a pain to do. lar1: 981102 the debugger isn't very useful anyway. hey, all the cool people are here yes it is unless you're debugging really low level code, but i don't do that often :) hey tcn^Wpyro :) long time no see^Wtalk^Wwrite :) i do liar: i figured as much :) core: I am doing real low level code ;) lar1: what are you working on? core: an OS liar: settle on low level designs and move up to usability stage already :) larOS lar1: i figured as much, what is it called :) core: crapola, did you reply anything i said "right now i TLB flush" core: RavenOS... yes I know... the name isn't that great, I really should change it er after i said eihrul: i haven't seen that? lar1: oh; i've listed you on the tunes OS review. oh, i'll paste > right now i flush the TLB on messages > 1 > no... two messages can be received > if there are already two threads blocking on the receiver > or if you do a receive, send > perhaps i could have different PDE's for sending and receiving, but 2+ receives still require a flushing lar1: i go see your site on a regular basis, but i haven't seen any updates since end of july :) core: _You_ read the Raven site? eihrul: ok, reformat into something that makes sense :) lar1: sure, i wrote a review for Tunes. core: basically, a fixed portion of the address space is reserved for mapping messages eihrul: no dynamic mapping? core: now if there are N threads blocked on sending to a given thread eihrul: oh; you only get one message at a time, is what you meant? core: Well, I don't maintain the page somone else does, I am writing a new one. The current one is way to flashy and frilly... the new page will be at http://ravenos.dhs.org core: the thread can receive N messages, quantum permitting eihrul: did u tell core about yer software tlb thing? core just got here :) lar1: http://www.angelfire.com/az/ravenos/ is your current page? that's the project I listed on the OS review lar1: i think i even sent you email about it :) eihrul: he has email capabilities u know core: also, another problem i have is that if an IPC operation takes place hehe liar core: Yup thats where the current page is... but its gonna be rewritten and moved and since my ipc operation is preemptable software TLB? sounds recursive software TLB sounds slow lar1: ok :) well, ok, i was aware of your project then. it's on the review :) if the receiver or sender is destroyed and the IPC continues, there are problems eihrul: IPC must have cancel or abort exceptions (cancel from client to server; abort from server to client) 04:50pm either exceptions must propagate up or down the IPC chain so that everyone can terminate cleanly core: I was thinking about doing some work for the review. Going through it all and putting it in a db. I assume that job is still open i have to then signal a thread as executing a transaction so i can do that kind of notification though the job is open lar1: yeah, very open. i have no time; i had started, so did tril, but nothing came out of it yet.. i don't want to lose the richness of the review, so the database must be quite extensive i could either make the IPC code do it, or the preemption interrupt check if EIP points to within the IPC handlers core, lar1: it might be just as fast as normal paging but it requires only 2-3 page tables to map everything into tril: i'm sorry i didn't give you feedback on the db, really busy and all :/ I think the .html file is a good enough db Tril, Core: Yah, well, I have been _very_ buisy but I am planning on learning SQL and when I do, I'll take a shot at doing the review project dbifying pyro: all i personally need would be an automatic way to check and correct links.. sites are moving targets, especially if the projects grow and move away from angelfire or geoshitties :) the data isn't very structured.. it's mostly English lar1: you're welcome to propose something anyway :) pyro: right.. that's why putting it into a db isn't evident :) i don't want to lose the semantics. and fare did a big job with that, i'd feel bad if we did. maybe something with 'sed'.. :) eihrul: manipulating any sort of pointer during IPC sounds Bad :) pyro: yeah, maybe i can just look for href's i guess :) eihrul: ta for the file, i'll read it you could test them automatically, fix them manually core: So we are looking at a DB or an HTML or ...? or maybe have the checker insert [MIA] pyro: right.. i guess the wiki server could help for that some way or other.. lar1: some sort of db would be nice, but something flexible enough to include all info currently on the page in some structured way pyro: right :) core: what is with your use of ta? it evades me :) I know i've seen scripts for this already eihrul: oh, 'thanks' :) core: i'll just set a flag on the thread that it's involved in a transaction lar1: so when will the site will be up? are you making progress on raven? and if the thread is getting destroy, the transaction will be canceled accordingly eihrul: or the thread destructor can send cancel requests to all current transactions? core: I don't know when the site will be up, hopefully very soon... but most of my work on everything has been halted the past week or so. I am getting more free time now, and since the bug in the BB was fixed Raven should be going quicker lar1: what does it do at the moment? (just out of curiosity) core: basically... but i think i'll need to disable interrupts until the flag is set core: because if the thread gets destroyed before i have a chance to or the flag, icky :) -:- hcf [nef@me-portland-us701.javanet.com] has joined #tunes eihrul: _something_ has to destroy the thread after it isn't scheduled anymore. that _something_ can flush pending transactions. core: Not alot, just going in to protected mode and puts an A on the screen. The kernel is alleged to have IDE and all video services written and the GUI allged to make windows and move them around core: and it will indeed... but right now i'm just trying to make sure i don't have race conditions lar1: any specific idea behind the system, or just a fun and educational project? eihrul: there are finer-grained ways than disabling interrupts (btw, on SMP systems this is a terribly expensive IPI) core: spinlocks? core: Mostly fun and education (i know you're doing that for fun, but you may want to scale up sometime :) mutexes, semaphores, etc? eihrul: that's one, yes; spinlocks don't involve rescheduling, so it's quite inexpensive eihrul: right mutexes just being a special case of semaphore. this isn't an SMP kernel though, so yielding/context switching is more expensive 05:00pm Whats a mutex? mutually exclusive lock lar1: a semaphore whose initial counter is only 1 lar1: so there can only be one thread that can acquire the semaphore at a time core: Whats the point of that? lar1: serialising lar1: as in, making sure a particular section of code is only executed by one thread at any point in time. lar1: like, say, when you're modifying a memory allocation chain, etc. core: Ohh core: How does playing with the counter of a Semapore let you do that? lar1: the easiest form of mutex is disabling interrupts lar1: if you initialise the counter to 1, only one thread can down() before the others have to wait. okay, race condition eliminated... core: do cli/sti stall the whole CPU for N cycles or can other instructions execute? eihrul: anything visible for EMK yet? :) yeah core: So when a semaphore's counter gets to 0 then all other threads wanting that device are blocked until up() is called? eihrul: um, read the intel docs, i don't think they're pairable core: about 6+ assembly files and almost as many headers it's not pairable... but in optimization manuals it says certain long instructions like division lar1: right :) can be 'shadowed' eihrul: ah, yeah.. then yes, unless it is mentioned as a serialising instruction eihrul: ah, does it boot? :) core: ok, thanks