IRC log started Sat Jul 24 00:00:00 1999 [msg(TUNES)] permlog 1999.0724 -:- ruiner [nate@ppp040.wi.centuryinter.net] has joined #tunes hmmmmm.....anybody looking for help with their os? out of curiosity... no response.....k I was just wondering if anybody had tried to enlist help via slashdot 01:10am -:- SignOff ruiner: #TUNES (Leaving) -:- franx [user9106@195.240.83.1] has joined #Tunes -:- franx [user9106@195.240.83.1] has left #Tunes [] -:- smoke [smoke@14dyn65.delft.casema.net] has joined #tunes -:- core [core@core.suntech.fr] has joined #tunes people hi 03:20am core! -:- Fare3053 is now known as Fare fare! t'es pas sur mon serveur, enfoire de toi :-) komenvatu? serveur? bien bien! en train de delirer avec le driver mediagx sous clementien :-) et toi? fare: /whois core s/tien/tine/ -:- SignOff Fare: #TUNES (changing servers) -:- Fare [rideau@quatramaran.ens.fr] has joined #Tunes hum mon serveur donc ;) ex irc.ggi-project.org -dans-son-coin :) comment ca va? t'etais parti en exil :) ouais, St Malo, Grenoble... oui on a vu ton rapport de globe-trotter sur la ML ;) t'as essaye de comprendre mon article? * Fare/#Tunes compiles retro and clementine Hum. Where is my floppy drive? j'avais pas trop le temps de convertir du .ps :) mais je le lirai ce weekend convertir? man gv apt-get install gv je sais comment ca marche, c'est pas le premier .ps que tu nous sors ;) bon, ben alors? ghostview est inmettable :) je prefere l'imprimer, j'ai une imprimante postscript ici gv >> ghostview less implement1.tex mais je lis tous les documents que tu ecris, meme s'il me faut un dictionnaire a cote, meme pas mal :) 03:30am :( pour un dico anglais-anglais, apt-get install dict hehe non, il *me* faut un dico parce que tu emploies des mots qui sont courants pour toi mais moins pour toi, mais une fois ca regle, c'est interessant et bien ecrit :) ton article sur la metaprogrammation et la disponibilite des sources a fait le tour de mes amis (bon celui la etait relativement facile en vocabulaire :) heureux qu'il y en ait qui apprecient... (a vrai dire, je me suis fais un ami grace a cet article: ca lui a tellement plu qu'il m'a contacte!) c'est quoi, la difference entre nasm 0.98pl3.6 et 0.98? bien sur que y'en a qui apprecient.. si y'avait plus du tout de recherche et de documents ecrits, on dirait tous sig heil a micromou 0.98 c'est la version finale.. qq bugs en moins, la doc terminee (achtung! separee de l'archive principale, il faut la prendre a part desormais !), etc.. autant upgrader a 0.98 j'suis paresseux :) hehe.. ben normalement clementine marchera, a vrai dire il marchait avec 0.97 tres bien alors.. si c'est plus complique que apt-get install nasm, j'ai peur hehe qq'un fera bientot des packages debian pour nasm, binutils et gdb patches pour clementine qq'un? soit lionel au boulot (qui a fait l'outil de construction d'image pour clementine qui remplace les makefiles pourris :), soit "quelqu'un" des que ca sort en open :) clementine is twice bigger as retro is this good or bad? :) 03:40am clementine has GGI :P neither hum. i think ggi takes half the source-space :) (even if it takes little runtime space) (well, it has some of ggi anyway; the drivers i have tested to work - vga, permedia 2, mediagx) retro on Kadath: error on early components fatal error :( retro or clementine? oops. I mean, clementine core: hi! retro works fine on Kadath hey crimson :) * Fare/#Tunes goes test on ZhengHe -- c u ! fare: oh, did you build it with bison or byacc? crimson: your cvs repository is messed up :) * Crimson/#tunes has done a scheduler now... core: really? core: it is plundis work... core:~/sources/kernels/chaos$ cvs -d :pserver:guest@chaosdev.org:/usr/local/cvs checkout chaos Sorry, you don't have read/write access to the history file /usr/local/cvs/CVSROOT/history (that's a server side issue btw :) core: hmm... i will send an email to plundis... okay :-) -:- DimGuy [dimguy@as002305.singnet.com.sg] has joined #Tunes fare: make sure you did make mrproper, and that you have byacc, not bison fare: the yacc prompt should say: yacc [-dlrtv] [-b file_prefix] [-p symbol_prefix] filename fare: if you did both, then make should build the entire OS (about 1 minute 15 seconds here :) セcoreセ: What does *.fr stands for? dimguy: franconia core: what os? os is never complete okay crimson: oh. eh. clementine :) OS is operating system * Fare/#Tunes is back dimguy: *shrug* france obviously. hum. Fatal: Error in early components Oh.. Sorry.. I wanted to find out core: the pagelevel protection works fine now... maybe I did sumthin wrong at compile time? 03:50am fare: read above whether there are any servers which are none other than usa and singapore to link to a network.. hee dimguy: no, my server is in france France... > yacc 13:03:54 usage: yacc [-dlrtv] [-b file_prefix] [-p symbol_prefix] filename dimguy: openprojects has servers al over the place Eh.. core fare: okay. did you do make mrproper as the first thing? It is not linking here... s/al/all/ dimguy: well, open a server in .sg :) SG servers are slow core: I didn't. retrying. damn laggy hee SG? fare: please do, i changed too much singapore :_ core: did you update your Review-page? fare: the system builds on ~ 15 different boxes out of cvs checkout, there must be some compile time issue yes :) crimson: no :(( i must do that today. .fr means german conquerors enslaved the people in this country. Anybody can help me to link to a network? crimson: did you update your THANKS file? :) core: think so... =) crimson: hehe, just curious fare: want to work for an innovative french start-up ? *grin* DG: apparently you're already linked to the internet! Eh... dimguy: then use a nearby USA server, what can i say :) I wanted a dedicated one... US? well, US is full of em core: maybe when I'm fired from my PhD studentship I am not in US... fare: that'd be nice :) (not that you get fired of PhD, eh, that you came here.) dimguy: singapore links mostly to the US to go to the rest of the world Hmm.. No no I want to run ircd. :) dimguy: then run it..? * Fare/#Tunes fucks the french fisc fare: you too? we can start a club. But this irc network wants servers from other places dimguy: they will accept .sg Nope. SG servers are damn slow dimguy: why? ssh? :) zeeya! dimguy: they don't acre (actually, "the french fisc fucks fare" would be more accurate) s/acre/care/ seeya crimson care? They really do reject sg servers i have tried us and sg servers but got rejected fare: again, we could start a club :) (i'm single [taxes-wise anyway], young, cadre, nicely paid, and have shares.) core: that makefile replacement, you mean, mkimage? fare: yep *sigh* fare: it does an integrity check, albeit a little inflexible at the moment (ie. checks all deps of all components) retrying on Kadath... fare: does it run? "error in early components" can only denote a build problem still same error what the hell shall I rm -rf clementine, and check it out once again? you updated all patches, right? patches? rebuilt binutils and nasm aaaarrrrrrggggggghhhhhhh! * core/#tunes finds a nice spot where fare can't hack at him with an axe * core/#tunes decides to hide there for a while 04:00am what about nasm-j5 ? whatever that is? :) john fine's patch #5 ? yup hum, and where is the binutils source to patch? not in linux/GCC anymore, nor in /pub/gnu/binutils !!! well, i'd think p5 is < 0.98, but try it if you like :) -:- DimGuy [dimguy@as002305.singnet.com.sg] has left #Tunes [] core: where did you find the binutils sources? fare: ftp://ftp.varesearch.com/pub/support/hjl/binutils/ fare: well, it was on metalab at the time i got it, but it's there as well where on metalab? this URL seems to be the home of H.J. Lu tho. fare: it's not on metalab anymore ouch. HJLu. Not someone I'd trust on code reliability. This guy is a specialist in kluges you're biased by NFS :) (he's maintaining linux nfs code :P) so he also wrote linux NFS code? that explains a lot! yeh twisted asian mind. i'd think he also wrote IRIX's if it wasn't a closed system, but anyway :) cannot do anything straight! fare: damn asians! :) [hello, dear mr. irc log. for the record, and for people who won't get the joke later, Fare is. thank you, end of sidenote] core: hey, you're late! There binutils 2.9.4 out already! is it? well, i've been in apnee for 2 months, i guess i should update :) whee, i have a new way to annoy people who work on clementine :-) and 2.9.5.0.3, actually * Fare/#Tunes wonders which version to try... oi, still 260 release critical bugs on potato :/ fare: ah, ok. well, compile 2.9.1.0.25 for now, and i'll update :) 04:10am ok (since you have cable now, i don't feel too bad making you download 10 MB tarballs :) fare: btw, in mid-august or so, we will setup a dedicated box at suntech, for opensource projects. if tunes needed any extra services, you'll be able to host them there. great! a reliable french mirror for tunes wouldn't hurt. well, you'll have it :) don't let the rosbif be the only european to host tunes! * Fare/#Tunes hopes alaric doesn't read the irc logs :) OSS is part of suntech's engagement to investors, so we'll respect it fully. OSS server, oss system, etc. :) lol.. he probably will. hi alaric! :) why not announce it NOW? why not announce what now? your OSS support? oh well, i don't want to make empty announces the box will be up in another probably 3 weeks or so.; why not publish open source before the announce? at this time, we can make an announce and have people start using it immediately :) cause we need more time than that :) clem will be released later than mid-august * Fare/#Tunes hits core's head with a clementine tree or worse even, with a binutils tree! oh, 2.9.5.0.3 is dev 2.9.1.0.25 is the latest stable * core/#tunes ponders "stable"! what a joke! well, think about it - 2.9.1.0.25 is stable for H.J. lu.. what will 2.9.5.0.3 be :) HJLu makes plug and pray a reality! 04:20am hehe.. maybe it's an evil m$ employee that's going to subvert linux with crappy code :) core: too late :) eihrul: what is? it's already subverted with ugly code... hum. Looks like 16-bit code is supported by gas, at last? fare: yeah, to some extent eihrul: ah well, for a "regular" system, it works, i guess. ugly code can work doesn't make it easy to maintain/modify however at last, I've found a new maintainer for the Asm-HOWTO! fare: who ? asm-howto? asm-howto is really mismaintained eihrul: fare maintains the assembly-howto just odd, a howto for a language? eihrul: duh! eihrul: no, it's a howto about how to write assembly under linux. then shouldn't it be the assembly-under-linux howto? eihrul: all HOWTOs are part of the linux documentation project, what do you think? howto's need not be limited to linux :P eihrul: well, say, there's a Bootdisk-HOWTO. Should it be renamed How-To-Make-A-Linux-Bootdisk-HOWTO ? eihrul: no, but it's a word that linux documentation people have coined, and everyone knows it means "HOWTO..do it under linux". "Konstantin Boldyshev" okay, i can live with that eihrul: would you think the 3DFX-howto is how to install your 3dfx in your computer too, or how to make it run under linux? :P fare: any good? fare: did you build binutils? not built it yet ahh built & installed nasm, unpacking binutils 04:30am core: btw, i've been tossing the bitmap scheme around in my head for a night or two... i cannot for the life of me find a way to avoid probing for a new string of raw pages once the current pointed to string is taken okay eihrul: then read the buddy scheme again :) bitmap? fare: free pages bitmap it's not that i don't get the buddy scheme it's just i don't get how you're doing it so easily with a bitmap is a bitmap efficient? i use lists, but samuel falvo does it with a bitmap just fine :) when memory is heavily loaded, bitmaps might suck but you still have to probe in some cases i can't find a way around that without leaving holes of unused pages fare: well, not scanning through them - more like using the buddy scheme i use, on them how is that a bitmap, then? fare: well, you have bits set when pages are allocated :-) who ever checks those bits? the question is: is the bitmap the sole mechanism for doling out pages fare: the buddy allocator :) or do you have helper structures such as lists? eihrul: again, _i_ use lists :) (not a bitmap) well, if i were to use a bitmap :) eihrul: download dolphin code i guess, or ask sam dolphin code? eihrul: http://www.dolphin.openprojects.net been looking at L4 though, it has given me some ideas the idea of user level page allocation/management seemed nice because then however you allocate pages is really insignifigant because it can always be replaced easily the way i allocate _memory_ can be easily replaced. even by something not using paging if you want insecurity and speed for some embedded application. i.e. you can easily remove the paging mechanism? yeah, if you write a memory manager that doesn't use it, and still complies to the clementine interface for memory management. so paging is completely implemented at the user level or is it just that it provides a linear map of pages? other components don't care about paging they just ask for memory. 04:40am would anyone here know a fast way to call a userspace routine from kernelspace in linux? (or are linux questions very offtopic here?) no paging means no more TLB misses! smoke: you don't it means more predictable real-time behavior, too smoke: there is no PL0->PL3 call scheme. but it means no more hardware help for virtual memory smoke: you have to send a signal to the userspace process, so it can be woken up by the scheduler fare: that's a bad thing :) smoke: the kernel definitely does not want to call user-space routines! fare: well, you can develop the system using a paging mm, and when everything is bootstrapped with your safe compiler, you can automagically get rid of H/W protection :) core: hm. the thing is that signals are handled asynchronously afaik, and i need about ~1ms accuracy smoke: you can't do that. boohoo smoke: async is the only way to trigger an action from kernelspace to userspace core: paging is not just protection; it's also a way to achieve various tricks. smoke: but CPS can help transform async into sync fare: i KNOW :) like copy on write, partial mmapping, etc. core: what i want is to have 70fps video (means polling the retrace) with a modplayer running in another thread; until now i haven't found a solution fare: what is cps? smoke: you can't do that :) core: i can poll the retrace, but when the musicthread is running at the time the retrace occurs i miss a frame.. so i need to cancel the musicthread out just before the retrace may occur smoke: if you signal the process, and it has higher priority than the mod thread, it'll wake up first. (i'm sorry if i am messing up a discussion here, please tell me to shut up about non-tunes related topics) core: wow, dolphin code is pretty clean and small... eihrul: and not very functional.. but it's clean eihrul: sure if i'm going to read it, i'll take clean over functional linux is functional, but it's an absolutely bitch to figure out how it works sometimes linux code is interdependent on everything. they have macros all over the place living in header files that you think are just normal functions that have various ugly side effects all my macros are uppercase. well, some are uppercase in linux, some aren't hence the chaos 04:50am eihrul: Clean *is* a functional language :) CPS is Continuation Passing Style fare: sometimes you have to trade cleanliness for functionality (or lack thereof :>), but only to a certain limit * Fare/#Tunes make install the binutils fare: is cps a mechanism in linux? 05:00am smoke: no, it's a general implementation technique although it's used in linux in the form of sigreturn 05:10am aha fare: did you build binutils? yup fare: does clementine work now? * Fare/#Tunes make mrproper; make (i assume you also make install'ed binutils and nasm :) * Fare/#Tunes is having disagreements with the new HOWTO maintainer about using direct linux syscalls fare: is the new asm-howto maintainer any good? fare: why, you think it's a good idea to use syscalls directly? :/ 05:30am fare: allo? 05:40am hmm * smoke/#tunes is puzzled by how to solve the modplaying problem :( 05:50am fare ?! core: he looks ok fare: it boots now? although he has something against the libc ld: unrecognised emulation mode: i386xcom32 fare: you have make install-ed ouch. Did I forget to apply the patch? fare: or didn't ./configure right ./configure --enable-shared --enable-targets=i386-clementine-xcom32 oops bummer! hope your box is fast :) 06:00am :( :( * Fare/#Tunes is rewriting the Asm-HOWTO I wish to make a common release with the new maintainer before to make him officially so and to retire fare: please re-recompile binutils and retry making clementine, i have to know if it works :-) fare: is the new maintainer any good? (#3) :) fare: i can't stay connected all day :( not from home anyway why use section .text when you can just say .text ? habit i guess did you build binutils? i'm on isdn at the moment :( 06:10am oops, forgot to type "make" after configure --... *sigh* :) /usr/local/bin/ld:archive.lo: file format not recognized; treating as linker script wtf is that? methinks you should take /usr/local/bin out of your path for the time to recompile and make install binutils again, then put it back yup core: do you know where PLundis is? i heard PLundis was the ruler of the known world. I control this world at the top of Plundis's fingertips. crimson: no? crimson: i thought you would know :) why ? * Fare/#Tunes puts "cdplay" on the input buffer of the "make" window, so that he'll be notified hehe core: you know, we use the same bot on irc.chaosdev.org... crimson: yeah, so i heard crimson: you should like to OPN s/like/link/ core: how to use .bss in nasm? yeah... i should... ;) fare: section .bss should set defaults? fare: or stick 'nocontents' or whatever ELF uses, next to it I mean, how to reserve space in .bss? what's the diff between resc, resd, resb, etc? fare: oh. resb reserves bytes, resw words, resd dwords, und so weiter resc? i don't know this one. chars? *shrug* must be that i can look it up if you like i don't see any resc retro uses it in forth.asm!!! 06:20am hmm, okay. not a macro of his? maybe doesn't seem to be what are alignment and ordering guarantees in .bss? depends on the fileformat can the assembler reorder entries? must it align them? nasm will never reorder ELF has tunable alignment for any section in its header (like xcom) ok, installing xcom why is the configure flag not implicit in presence of the xcom patch?