IRC log started Sun Aug 15 00:00:00 1999 [msg(TUNES)] permlog 1999.0815 !forward.openprojects.net!! Received :irc.linux.com SERVER hogan.openprojects.net from irc.linux.com !?! okay, so what's the prob now? the code dont work hahah the iolist holds blocks of unused ports it finds the right block that would contain the port but then it screws up the size and doesnt create another block to hold unused ports above it 12:10am ok it doesnt screw up the size it doesnt create a new node for the ports after 3b4 ah found it 12:20am * eihrul/#tunes sighs. it now works it's so hard to write good APIs in C heh whats wrong trying to write up the API for my kernel and it looks ugly as hell and is not conceptually sound, that's all i dont see how C has any effect on yer api isnt a bad api the designers fault? :) not quite >>> liar [brand@p0wer.qzx.com] requested PING 934702790 684277 from #tunes shoving a pretty package over a box of shit is not necessarily easy -:- SignOff Mr_Wrong: #TUNES (Ping timeout for Mr_Wrong[adsl-63-192-213-146.dsl.snfc21.pacbell.net]) some of the stench leeks through leaks :) i find that hard to believe 12:40am the language has no effect on how crappy yer api is not really, no... or are u talking about yer syscall convention? but it's as good as anything to blame :) no... i'm talking about ym syscalls themselves the conventions are but a minor annoyance so yer blaming yer choice of syscalls on C, thats good :) im gonna blame any problems i have with brix on visual basic no... i'm blaming convention on c i'm blaming choice on myself... C is very sloppy when it comes to such well it took me A VERY LONG time to get an api that i liked for brix years i dont like my current api but i like it ALOT more than any other api i have had right now i'm struggling how to cleanly and uniformly represent address spaces and memory objects mapped into them hehe thats fun actually... it's not so much how to do that as how to present that right because the API starts to reek of featuritis at that point i know what ya mean but they're necessary features... :( thats the part i most hate about my current api i would really like to only have a couple syscalls but... how many syscalls are you at right now? let me look 12:50am if only computing languages only utilized a single region of memory in the same manner eliminating the notion of code, data, and stack :) then i'd be fine 26 user and 10 system and efficiently at that eh? i'm just ranting... crap your api up on your site? something is setting my iobitmap pointer for osp0 to 1 so my iget thinks that one exists www.qzx.com/brix/api.cgi but its not current atleast i dont think it is well it's big :) there might be 1 or 2 functins missing hmm i think that is current hmm that more current then current kreload shouldnt be on that list its part of kset now 01:00am hrmm, yep... it looks like it's suffering the same way my api is :) heh just think of it like this, it has lots of features so its flexible but not so many that the programmer is overwhelmed 36 functions are easy to memorize but then again... as i tend to hear so often in here, representing complexity with simplicity is near a sin :) and if yer not writing system components than its only 26 i guess i'll just settle for a slightly, and blatantly hacky api ugh, i hate tracking down bugs that overwrite data mmmm... memory overrun... no it only modified the iobitmap entry for the osp sides i dont make memory overrun errors :) ugh, i hate tracking down bugs that overwrite data * eihrul/#tunes grins. we've all put a jg where we meant a jge or a > where we meant a >= :) uhh no it directly modifed a single memory location 01:10am well, it's your bug... :) if i give u the source will u find it :) given enough time ugh E is so slow. i need to get off my ass and put my dual board in this box so i can switch desktops faster i use blackbox myself... i was using afterstep and it was fast, but now i have this bloated E/gnome pig why? :) there's always twm i like some of its features so im forced to use the entire thing what features? well its panel and some panel apps featuritis strikes again :) hack the panel out of gnome if it doesnt run smooth after i install my dual then i will have to change some shit with E/gnome it'd be easier to just degnomitize the panel :) ofcourse its probably a swap problem cuz afterstep and my 40 windows only used about 150meg ram and no swap and E/gnome with 10 windows uses all 192meg + about 75meg swap ouch.... i never even touch the swap 01:20am i never touched swap with afterstep could i see a dump os your ps ax? :) i need a good laugh heh this is whats funny, i was using ax4.1 with afterstep and ax used 90meg of ram and now im using xf and it only used 50meg what are both programs, respectively? :) never heard of either accelX and xfree ah hungarian notation at it's finest ugh im glad u asked for a ps dump hrmm... i'm wondering if i should have a tag field in my message passing :) i just found a coupld things running that shouldnt be portmap et al? :) sendmail and gpm ah... 3meg right there whatcha think? message tags or no tags? tag? a dword sent along with the message for arbitrary purposes of course what kinda purposes? if a process wants to use it as a short 4 byte long message it could if it wants to use it as a request number... it could what the hell, i'll do it :) i think those types of messages suck i like brix messages much more how do brix messages operate? fork/call no need to check message queues 01:30am i don't have any message queues a stinkin thread is created to handle the message then how do threads receive messages? it's synchronous they just wait :) xf doesnt use gpm does it? xfree uses it's own mouse drivers GPM is simply for console okay... i think i've succeeded in making my API somewhat less hacky but is it usable? for most purposes it's large enough wish it was smaller, but such is life a previous brix api didnt have those open/mopen/close/lookup/mlookup functions but i added them to make the system MUCH faster now all i have to do is implement all this and i found a big problem with brix today, it can have upto a 10ms delay before servicing an irq dont know how im gonna fix that that's the problem with microkernels :) they lose that uber-integrated speed no thats the problem with a feature in brix u can chain many threads to a single irq which is reminiscent of a problem common to many microkernels :) yeah... that could slow things down :) do you dispatch all the notifications before context switching? and when the irq triggers the kernel will resume all those threads but they dont get a timeslice until the current thread finishes ya ack, afk a minute... the coke i just drank is itching to get out usually u will only have 1 thread attached to an irq and it will take upto 10ms before it runs 01:40am if u have more than one then it will take 20ms before the second thread is run, 30ms for the third... interrupt chaining is evil :) hrmm, should i try implementing signals? i thought about skipping the current thread so the irq thread could resume immediately but if the current thread was a higher priority interrupt... maybe i could add a flag to the threads and set it if they are put on the irq events signals :) nah... i don't need signals :) me wonders about semaphores... ya make semaphores and events events are kewl messages do enough for that :) messages suck i can just fork off a thread to wait for messages from the kernel :) i need some kinda variables associated with objects how should i do them objects don't have variables? :) no 01:50am i thought the whole point of objects were that they were grouping of data named variables that only exist at runtime non-persistent stuff ala stack variables? :) no shared variables between different threads? ya make a heap of some sort i guess there are no processes and the osp does not have a static layout osp? object space, cr3 the memory space that threads run in u dont always have to change osp's when opening objects if u dont trust the object or just dont want it in yer osp then u open it in another. if it was already opened in another then u have no choice cuz u can have an object in two osp's simultaneously er cant ok... API at 26 functions so far :) heh down from 27 so i need some kinda named variables just enough to be on par with brix oh yer api is 26? 26 in count i have 36 but 26 to the application programmer :) maybe a compressed 64bit name and a 32bit value oh ya and half of my 10 syscom functions are only available to the startup and ofs actually... i probably need to provide reboot and shutdown syscalls :) dmabuffers is only used by some object that will allocate dma space and i don't have semaphores either no signals either :) well if u have a faulty kernel the user can make his own reboot call :) 02:00am kinda like in windows well i could just allow any privileged task to do it but actually... i couldn't it :) well the startup code uses kset to tell the kernel that it is the only object that can call reboot/shutdown i could maybe just give the boot task and only the boot task halt privileges and the only way to r/s is for the user to tell it to that way if any other process needs to shut down the system the boot task can decide policy allow the startup process to designate a process that can call it but that's a whole other syscall :) kset in brix does many nice things ah what the hell i'll make a hook for a supervisor process it tells the kernel what object will be handling the ofs who gets to use kmalloc, who to call for a keyword search... et_bool emk_supervisor_set (const et_thread); there we go :) damn those are some huge syscall names i like verbosity :) and and the nice thing about brix is that no one really has to know about the syscalls maybe it's that C++ has warped me from ever liking hungarian notation again who knows cuz brix has its own language that u are forced to use but what if someone wants to code for brix in assembly? :) they can but its not recommened cuz it breaks the nice distribution method that brix has what u code in asm will remain on yer system 02:10am it cant go into the master repository that the world will update from okay i just removed the supervisor system call :) hmm where the crap is sendmail started from? and just am making many privilege levels grep :) YA, killed gmc and my swap dipped a little :) brand 969 0.0 2.1 16644 4208 ? S Jul05 6:00 gnomepager_applet 17 MB FOR A PAGER!?!?!? well that is no longer on my panel what the hell is that pager doing my swap just made a bigger dip well it is a 3d pager featuritis :) heh hmm whats the diff between X and Xwrapper? is it just me... or do you think it's pointless to return a failure boolean in a halt () syscall? :) hehe what is halt for? it halts the system :) -:- ElGato [dave@bespin.dhs.org] has joined #tunes why? are u too lazy to have yer scheduler halt when idle? why is it that so many american tunesters are on and few international ones liar: er i mean, it shuts the system down COMPLETELY :) ElGato: u scared em off :D so halt = shutdown think i should rename? :) ya 02:20am done halt is something that should be done when u have no threads to run i went to take a nap at 2:00pm and told my brother "wake me up when you're done with the internet" to put the cpu to sleep it is now 2:20 am HAHAHA i just woke up 12 hour nap eh? :) yep when i start i don't stop obsessive compulsive napping now i've heard everything ugh, gnome restarted gmc slay it all! leave not a single bloody gnome alive! that isn't obsessive compulsive it's close enough who tought you about psycology? it's not like "if i don't nap the world is gonna end" i just over sleep who taught you how to spell? cheap shot but, of course i write at the 8th grade level heh and yer in what 12th grade? hah college grad? im going into the 9th grade suckah :D and yer STILL writing at the 8th grade level???? YES u should be writing at the going_into_9th_grade level well let this be a lesson to you when one is idle and halts, start up again when not idle liar: ohhh thanks for clarifing that eihrul: here ya go brand 32138 0.0 1.5 5144 3068 tty1 S Aug14 0:02 gnome-terminal -- brand 32139 0.0 0.2 1288 552 tty1 S Aug14 0:00 gnome-pty-helper each terminal needs a pty-helper laugh at that use xterm :) rxvt it's half as expensive i remember my 800k rxvt i had with afterstep, the rxvt with rh6 is 1.9meg debugging symbols perhaps? dunno and why x11amp needs 4 8.5meg processes is beyong me hmm... i wonder where i can get the source code to ps 02:30am why? i think it's part of bash isn't it or any other shell -rwxr-xr-x 1 root root 60656 Jul 6 16:14 /bin/ps just wanna find out how it queries the states of processes :) hmm 9meg for my slashdot panel app !sterling.openprojects.net!! Received :irc.linux.com SERVER hogan.openprojects.net from clarke.openprojects.net !?! ack brix has an info() syscall for that actually... who needs to know information about processes anyway? :) well the user might wanna see all the objects/threads running and maybe kill some 02:40am but that should all be managed by some task :) as this is a microkernel... ya * ElGato/#tunes is away: (Auto-Away after 10 mins) [BX-MsgLog On] you build system API's over the MK api :) info() doesnt dump text, some app is needed to convert the results from info() to text userland can read directly from kernel memory but then it can break if a kernel structure changes info() will always return the same values even if the kernel structures change there, i just put some hacks in for system info :) :) 29 syscalls who needs small api's anyway? :) haha would having 29 different interrupts for syscalls be kosher or overkill?:) i would use call gates and using callgates will make yer uk faster i gots to save eflags anyway so there's no benefit for just that reason why do u need to save eflags? many system calls cause context switches and i don't want to have to wait for the things to get preempted in user land 02:50am er kernel land :) what syscalls cause context switching message passing primarily does the sender wait fora reply? nope it's just a context switch + data transfer when i send a call() message it blocks the current thread and when the new thread is finished it resumes the blocked thread i might have a message_call () syscall in addition to send and receive but it would simply be a composite send/receive yer message passing is just a resume + data copy and the data copy is optional :) what is the point in passing a message if the data is optional? resume/suspend :) some times the contents of the message is irrelvent oh and what happens if a thread that is suppose to be waiting for a message is processing a message it just received? do u block the sender until the receiver goes back to waiting? yessir the sender or receiver can both configure how long they're willing to wait for transaction to complete +s 03:00am hrmm... i should probably go to sleep now well im gonna goto sleep haha cya :) -:- SignOff eihrul: #TUNES (Leaving) -:- SignOff liar: #TUNES (BRiX [http://www.qzx.com/brix] :: sleep) 03:10am Plundis? Crimson? Tril? Anyone awake? 04:30am -:- overfien [overfiend@m229.telcomplus.net] has joined #tunes y0h hey what brings you here? a few peeps from my college chill here ;p ahhh where you go to college? western wwu.edu transfering to berkley you're leaving berkeley? wtf would you want to do that? i'm going to berkley. oh ok i'm getting out of washington. to goto cali ;p :) my grandma went to berkeley i think it's a cool school i have a few friends down their.. that can hook me up wid cisco security testing jobs ahh cool heh comin down to california hehe what state u from? california :) oh hehe what part? souther (san diego) i got a summer house in sacromento.. actually its a condo "summer room" hehe