| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Comments:
<0> kthx^: depends on the state
<0> in your state, can be illegal
<1> but poll wont return coz it working on lots of sockets
<1> lol
<2> not true.
<1> eh
<2> poll returns as soon as there is something useful to communicate.
<1> no
<2> yes.
<1> ...
<1> i meant not that
<1> i know how to do poll for multiple sockets
<1> but how do you do both in a single thread
<1> is what i miss
<2> just like i showed.
<1> because poll() is supposed to keep going through all the sockets ans never return tho
<2> wrong.
<2> poll returns as soon as there is something useful to communicate.
<0> kthx^ you are right
<0> in your state, it works so
<2> *boggle*
<1> ignore communicate coz that would just go to parsing not
<1> going to accept();
<1> hence if the timeout was infiinite and no data was sent
<1> how could you call select it would be on a endless loop?
<2> loop { call poll. poll returns. a listen fd is marked POLLIN, you accept then add the fd to your list. handle any other fd's marked. } end of loop
<2> if you are using an event dispatcher then it usually has a way to modify the fd list it is watching.
<1> i think i get it now
<1> you saying use poll for accept and recv not just recv?
<2> of course. all sockets, and any interactive fd's as well.
<1> ah
<2> disks are always ready to read or write, so don't bother with file fd's.
<2> pipes too.
<2> err, use poll for pipes as well.
<3> yes poll works on pipes
<1> eh so how do you accept a connect in poll?
<2> you put the listen fd in the poll list.
<1> ah so a connection event is POLLIN?
<2> when poll says it has activity (POLLIN is customary) you accept.
<1> ah k
<1> whats a good backlog for listen()? coz ive tested a backlog of 1 - 100 with multiple connections and havnt seen a differnece>
<0> hmmm why yum can't find mplayer
<2> yum ****s?
<2> oh, you have fedora.
<2> you need to add the atrpms.net repository.
<3> I got the source and compiled it
<0> nice now it found it at atrpms.net
<0> jeffloc: I'm probably missing dependency libraries, too
<4> http://rafb.net/paste/results/jRLmN218.html <--- how can i fix this so it will compile?
<1> typedef on second one?
<4> er whoops yeah
<1> lewl
<4> http://rafb.net/paste/results/GrzAIK23.html :)
<5> typedef struct a a; typedef struct b b; struct a { b *b; }; struct b { a *a; };
<4> why "a a" ?
<5> no.
<0> to confuse u
<4> Huh?
<2> avoid typedef?
<2> struct a { struct b *b; ... }; struct b { struct a *a; ... };
<5> the units are ((typedef (struct a)) a)
<5> there is no "a a"
<4> hmm
<0> where do you see a a ?
<4> typedef struct a a;
<2> AndroMish: in mauke's first typedef.
<2> JoelS: define a as a synonym for struct a.
<4> Right
<4> Sigh whyyyy
<2> i tend to avoid typedef's.
<4> Still having 3 errors
<4> whew!!!! thanks
<3> AndroMish, would you like a list of the libs it needs?
<3> I'm using debian so it was easy. ldd then apt-file search to find out what packages I needed
<3> mine might not be compiled the same as yours.
<1> iPoll = (struct pollfd *)realloc(iPoll,sizeof(struct pollfd)*iPollSize+1); <<-- would that work?
<3> realloc returns a void* so you don't need to cast it if you #include <stdlib.h>
<6> hey
<6> mornin'
<6> hey Dach, you awake?
<6> hmm wee this webchat is annoyingly inconsistent
<7> Gambit-- is awake, zid_ is awake
<6> maybe if I keep talking i'll actually get delagged
<6> stranger things have happened
<6> hey AndroMish, did you respond to what I said abou tthere still being a deadlock condition with your approach?
<0> ??
<0> why
<6> still haven't seen all th elines I've typed
<6> hello zid
<6> i see you're awake.
<6> i noticed that this webchat doesn't deal well with no idle chatter.
<0> jeffloc: thanks, I installed it using yum
<6> Ah, he is awake
<3> shouldn't be missing anything vital then
<6> Basically you were saying that everyone starts with the lowest ID and locks upwards from there
<3> did it install all the codecs too . they might be a seperate package
<0> i'm not sure
<0> what's name of codecs pkg
<0> Gambit: right
<3> AndroMish, get yum to list them and | grep codecs
<6> AndroMish, actually, not a deadlock, but a starvation condition. If everyone locks objects On and On+1 and Tnmax locks Onmax and then gets busy with it for a while, no one can do anything until the entire thing rolls back.
<0> you wanted locking, you get locking
<6> it works fine as long as nothing gets in lockstep, but if you have a small number of objects that's near the number of threads, then it's actually pretty possible.
<0> don't complain that you get locking if you wanted locking
<0> and i said this alg doesn't deal with starvation
<0> all i said was it resolves the deadlock issue
<6> sooo laaggggeedd
<0> but if your purpose of
<0> locking is to avoid deletion of the object in the middle of the op,
<0> then refounting is better
<6> Yeah, sorry, my brain spun off elsewhere when I said it deadlocked, my apologies :0
<6> er :)
<6> heh might as well be using an abacus then :P It's still a good approach, I'm just not sure it's what I'm looking for.
<0> what's your purpose of locking: avoid crash if obj is deleted in the middle of the op, or something else ?
<6> hrm it's 07 now
<6> i wonder when i'll see this
<3> if that was all he would just use refcounts
<6> well not to avoid deletion -- i have shared_ptr's to take care of that -- it's just the general dining philosophers issue I'm analyzing atm.
<3> 13:15gmt
<3> see how lagged you are
<0> hmm mplayer desc says it includes codecs
<0> but it doesn't play metacity clips
<6> man over two second lag
<6> harsh
<0> Gambit_: bo appetite
<6> no the locking is to avoid race conditions
<3> a whole stack of them.. they should be in /usr/lib/codecs or /lib/codecs
<0> hmm nothing there
<3> mine are in /usr/local/lib/codecs because I compiled it there
<6> crazy lagged.
<6> and it's getting worse, too :(
<6> er, two minute lags
<6> now it's around three minutes
<0> Gambit-: you can't say that you add some locking into the program to resolve philosophical issue
<0> Leave philosophical issue to philosophers
<6> ah ****it, i'll be back later when i'm not lightyears behind everyone else.
<3> Gambit-, its 13:17:40 and you don't have the record for being lagged. I saw a poor fellow who was lagged by more than 1 hour. no idea how he stayed connected
<0> You don't normally add some overhead to the sw unless there s some practical problem
<0> jeffloc: he was on alpha centauri, 1 light away
<0> *light year
<8> jeffloc: You're lagged two seconds from me here, and you said it was 17:40 when it was 18:35. Time to install ntpd? :)
<0> or light hour maybe
<3> I know. I was going to change it when I finished the line but forgot :)
<0> if you guys don't behave, i'll trigger this thing
<8> 2006-07-05T13:20:52+0000
<8> /exec -o date -u --iso-8601=seconds
<0> I only have /usr/lib/helix/codecs, no other codecs
<0> i'm going to build it from srcs
<0> european clock signals (MW) are not receivable in IL
<0> it ****s
Return to
#c
or
Go to some related
logs:
INDIAN POITICS AND RELIGIONS
#beginner
#goal
#microsoft
sriptise
winregistrycleaner
#nhl
#beginner
The rise and fall of Rashneesh
#debian
|
|