| |
| |
| |
|
Page: 1 2 3
<opeth> in what folder does the .conf need to be
<AndroSyn> etc
<opeth> it is in there and still when i run ./ircd it doesnt start the ircd
<AndroSyn> try ./ircd -conftest
<opeth> its in both etc/ and /etc
<opeth> thanks
<AndroSyn> syntax errors i ***ume?
<opeth> bad cidr and something with ipv6
<wfl> andro: I think im gunna create a new 2.1 branch
<wfl> and do some memory cleanups in it
<AndroSyn> sure
<AndroSyn> i think my connection is finally fixed at least
<AndroSyn> i just thought about hostmask.c and how its likely not going to probably behave as nicely as it should on 64bit platforms
<AndroSyn> most all of that code ***umes unsigned long = 32bits
<wfl> im gunna remove refcounting on struct User
<wfl> because the only reason we have it is to attach it to server links
<AndroSyn> yeah
<wfl> im probably also gunna do the client_p->name cleanups
<wfl> but ill do them somewhat safer than in -devel
<wfl> as ill just rename name in struct
<wfl> then I can validate all references to it easily ;-)
<AndroSyn> yeah
<wfl> new branch is just svn cp?
<AndroSyn> yup
<wfl> reet, hometime
<AndroSyn> i'm cleaning up all sorts of type stuff since we've got pretty reliable u_intX_t detection now
<AndroSyn> laters
<anjuta> what was the exact encap syntax ? i cannot find it in doc/
<spoogle> one of these days I'll learn how to use this svn shti
<spoogle> s/ti/it/
<AndroSyn> its not hard
<anfl> andro: im making chptr->mode.key dynamic
<AndroSyn> k
<AndroSyn> what part of svn co url is hard?
<AndroSyn> :P
<AndroSyn> or svn commit, or svn update
<AndroSyn> :P
<spoogle> by learn how to use
<spoogle> I mean "find which port it is"
<spoogle> and type "make"
<AndroSyn> cd /usr/ports/devel/subversion && make && make install
<AndroSyn> :P
<spoogle> ta
<spoogle> can you type "su" too
<spoogle> :)
<anfl> if(strcmp(mode.key, oldmode->key) < 0)
<anfl> what you reckon
<anfl> add a SAFE_STRCMP() macro?
<AndroSyn> you mean to check the length?
<AndroSyn> why not just use strncmp()
<AndroSyn> :P
<AndroSyn> and that needs to be strcmp() != 0
<jilles> no
<jilles> I think that's the compare in SJOIN
<anfl> if chptr->mode.key is dynamic
<jilles> what's the point of making chptr->mode.key dynamic?
<anfl> we can now no longer guarantee chptr->mode.key != NULL
<AndroSyn> just do if(chptr->mode.key != NULL && strcmp() != 0)
<anfl> jilles: we save at least 500k of ram
<AndroSyn> though what difference does 500k of ram make..when you're using 150MB :P
<anfl> it wont be just 500k though
<anfl> theres nothing big we can save ram on, but a lot of little cleanups will add up :)
<AndroSyn> though, is it worth it to do that and save maybe 5MB?
<anfl> yes.
<AndroSyn> i'm not seeing the advantage..
<anfl> I think our mem usage is too high personally :)
<jilles> you'll have to allocate separate bits of memory for +k channels though
<anfl> which will be a very small minority
<AndroSyn> well the thing is, the more you hit the malloc heap, the slower things get
<anfl> im aware :)
<jilles> perhaps use a block heap with KEYLEN pieces?
<AndroSyn> if we wanted to save memory, you'd strdup() hostnames :P
<anfl> but personally I think we've gone too far
<AndroSyn> has anyone complained? :)
<anfl> I dont like it! :p
<AndroSyn> .za is currently using 133MB, which imho is fairly reasonable
<anfl> im doing it in a branch, stop complaining ;-)
<AndroSyn> for a server that really has no users
<spoogle> if nobody else has complained, I will
<AndroSyn> that's because you have a sandy vagina
<AndroSyn> well 2.0.x i have no complaints about in terms of memory usage
<AndroSyn> i don't think 2.1 is much worse
<AndroSyn> ratbox 30144 3.9 27.2 143100 141536 ?? Ss 6Oct05 1921:05.33 ./ircd
<AndroSyn> thats igs, with 5300 clients
<AndroSyn> i don't think 140MB is unreasonable
<anfl> you know, this refcounting on struct User is completely pointless
<anfl> because nothing uses serv->user
<AndroSyn> heh
<anfl> I think im gunna remove user->server too
<anfl> because we can just jump client_p->serv->name
<anfl> err, servptr
<anfl> if servptr doesnt exist, we're in a world of **** :p
<jilles> it's from the time before QS :)
<anfl> but youll still have servptr
<anfl> as you will always exit users before a server
<AndroSyn> you damn well better exit them before servers ;)
<anfl> why does whowas link to a struct Client?
<anfl> err, struct Client link to a whowas
<AndroSyn> where does it do that?
<anfl> struct Whowas *whowas; /* Pointers to whowas structs */
<anfl> in struct Client :p
<AndroSyn> now that is a damn good question
<anfl> also, I dont see why hopcount is in struct Client :)
<anfl> we might as well add a macro for it.. MyClient((x)) ? 0 : (x)->from->serv->hopcount
<jilles> serv->up can go away as well
<anfl> err wait
<anfl> thats not right
<anfl> something like that anyway :)
<AndroSyn> the whowas is for when you have a client that's online
<anfl> youll have to elaborate, as I dont see why you wouldnt just do a find_client() to see if its online
<anfl> :p
<jilles> it's for the part of whowas that finds clients who have recently used a nick
<anfl> ahh
<anfl> nick chasing
<anfl> now it makes sense :)
<AndroSyn> would be nice to make the nick chasing stuff die!
<AndroSyn> die die die!
<jilles> no
<AndroSyn> would be nice if TS5 could go die
<jilles> nick chasing is still necessary for commands from local clients
<Quai> 20:26 -!- E xml_update -24121 seconds
<AndroSyn> heh oops
<Quai> I just wrot a module with that eventIsh..
<Quai> and stats E now shows 24 events like that with the same name
<Quai> (and one running event)
<Quai> http://tg.quai.no/m_linkxml.c
<Quai> is there something I have missed? :P
<AndroSyn> jilles: it wouldn't be needed for clients getting killed remotely
<AndroSyn> well i could see the need for it..but i never did like nick chasing for oper issued kills
<anfl> I think other things follow nick chasing
<anfl> like mode
<anfl> not sure :)
<AndroSyn> its still a hack :P
<jilles> kill, kick, mode +ov
<jilles> it makes it less likely that someone avoids an action against him by changing nicks
<anfl> err, how do I trap raw s2s traffic in xchat?
<AndroSyn> windows->raw log
<AndroSyn> err window
<anfl> indeed, ta :)
<AndroSyn> well if you're using TS6 only, the only time that would happen is if the server got the nick change shortly before you issued the kick
<AndroSyn> if they are both in flight at the same time, the kick is issued against the UID
<AndroSyn> no amount of nick changing is gonna help him then ;)
<jilles> IRC clients don't understand that
<jilles> they rely upon the server to do chasing
<AndroSyn> well if the nick change got there before..
<AndroSyn> and then you send a kick against a nick that changed
Return to ratbox
or
Go to some related
logs:
eurochat
|
|