#gemrb@irc.freenode.net logs for 21 Mar 2013 (GMT)

Archive Today Yesterday Tomorrow
GemRB homepage


[00:00:12] <-- |Blaze| has left IRC (Ping timeout: 276 seconds)
[00:03:16] --> Coriander has joined #gemrb
[01:20:48] --> skyphen has joined #gemrb
[04:26:18] <-- skyphen has left IRC (Ping timeout: 245 seconds)
[04:31:38] <-- brada has left IRC (Quit: brada)
[04:52:06] --> brada has joined #gemrb
[06:30:43] <-- brada has left IRC (Quit: brada)
[07:18:46] <-- |Cable| has left IRC (Ping timeout: 245 seconds)
[07:21:59] --> lynxlynxlynx has joined #gemrb
[07:22:00] --- ChanServ gives channel operator status to lynxlynxlynx
[07:32:07] --> |Cable| has joined #gemrb
[09:03:29] <edheldil> Cache should default to ${UserDir}/.gemrb/cache (or whatever is the exact name of that). If that does not exist, we could 1) create the dir 2) fallback to ./Cache. (1) is preferable, imo
[09:04:24] <edheldil> ( on windows, it should be ${USerDir}/GemRB/Cache, probably )
[09:08:17] <edheldil> so we should have some var that points at ~/.gemrb on *nixes and ~/GemRB on windows. Dunno about Mac
[09:09:30] <edheldil> ... although it might be more complicated if there should be deeper separation of config and cache on windows or Mac
[09:10:45] <edheldil> --> st. like UserConfig, UserRunState, UserData
[09:11:21] <edheldil> brada: btw: man 3 strtok ;-)
[09:11:33] <fuzzie> It's not a useful manpage, at least here.
[09:11:40] <edheldil> no?
[09:12:18] <fuzzie> My version doesn't even imply it's modifying in-place until the "BUGS" section, unless I'm missing something..
[09:13:04] <edheldil> well, it's in the BUGS section, so what? :)
[09:13:30] <fuzzie> so the BUGS section still doesn't explain what on earth it's doing :p
[09:14:06] <fuzzie> it's really confusing to me, it talks about "uses a static buffer while parsing" but presumably it's talking about storing a pointer somewhere
[09:16:39] <fuzzie> but to be fair I haven't had coffee yet
[09:17:03] <edheldil> you are right, might be a doc bug
[09:17:05] <fuzzie> and, yes, +1 on the making a cache somewhere sensible
[09:17:26] <fuzzie> but really you should have a unique cache for every gemrb session
[09:17:44] <fuzzie> so for me it would be 'nicer' to put it in a temp dir somewhere
[09:18:12] <fuzzie> the old code seemed to be concaternating '/Cache' onto UserDir though, which I assume would have the effect you describe
[09:18:55] <edheldil> freebsd 's manpage is different, though I am not sure it's better
[09:19:45] <edheldil> fuzzie: since we delete cache on startup, we effectively have unique cache for each session
[09:20:16] <edheldil> and if by "temp dir" you mean *shared * temp dir, that's not a good idea :)
[09:20:29] <wjp> fuzzie: the glibc one keeps a static pointer like you'd expect
[09:20:59] <fuzzie> edheldil: why?
[09:21:20] <fuzzie> and, well, testcase: run gemrb on bg1 in one terminal, run gemrb on bg2 in another terminal
[09:21:31] <fuzzie> if it breaks, fuzzie is frowning at you :P
[09:21:49] <edheldil> wjp: yeah, I guess it's a doc bug. Fuzzie has a right to immortalize herself by submitting a patch to glibc :-D
[09:22:03] <fuzzie> as if I don't have enough ongoing patches :/
[09:22:15] <fuzzie> ::despair::
[09:22:32] <edheldil> fuzzie: in that case bother with setting different cache directories :)
[09:22:56] <edheldil> ../. in the config files
[09:23:23] <fuzzie> well, or just put them in temp?
[09:23:48] <fuzzie> it seems like it's kind of a solved problem as long as you don't give them predictable names
[09:24:00] <edheldil> ah, you mean unique temp directory ala mktemp
[09:24:02] <fuzzie> and we don't preserve the content anyway, as you say
[09:24:09] <fuzzie> but I mean I'm just rambling here
[09:25:02] <edheldil> btw, should not we try to preserve the content? That would massively spped up the startup on embedded platforms, no?
[09:25:39] <fuzzie> well, that adds a lot of complications
[09:25:57] <fuzzie> since I think all the embedded platforms we care about are far more disk-space-limited than cpu-limited
[09:26:38] <edheldil> ... flush the cache only if config file changed or we load different save than the last one
[09:27:11] <edheldil> since we delete them on startup and not on shutdown, iirc, that's a moot objection
[09:27:20] <fuzzie> well, that's a really silly bug on our part
[09:27:27] <fuzzie> it's one reason I was thinking that temp would be preferable
[09:27:34] <fuzzie> i hope nowadays that we should delete them on shutdown too
[09:28:02] <fuzzie> but on the other hand I think we don't actually take *that* much space in our cache
[09:28:06] <edheldil> if you do not delete the files one by one, temp would not help you anyway
[09:28:15] <fuzzie> since we don't use it as a CD cache like IE does
[09:28:30] <edheldil> so far :)
[09:28:35] <fuzzie> well, temp is deleted on reboot, which I hear some people do, occasionally :)
[09:29:45] <edheldil> yeah, but imo it's better to have one unflushed cache more or less constantly than 20 of them until your next reboot :)
[09:31:26] <edheldil> what do you think about storing config file crc and save name in the cache and flush it if they do not match?
[09:33:12] <edheldil> or would it allow hard-to-detect bug, so rather steer clear of it? Or enable with cfg option?
[09:35:18] <fuzzie> i think preserving cache is really difficult
[09:35:46] <edheldil> does ie do it?
[09:35:59] <fuzzie> yes, but IE uses cache for something else
[09:36:14] <edheldil> where do you see the problem?
[09:36:17] <fuzzie> and maintains mutable files elsewhere
[09:36:29] <edheldil> ah
[09:36:46] <fuzzie> if you just think about performance then I would hope you can just preserve tilesets?
[09:36:55] <fuzzie> honestly it's been a long time since I thought about how this worked
[09:37:45] <edheldil> I just remember the time it took to start bg2 on n900 :)
[09:38:03] <edheldil> (though it was unusable for other reasons)
[09:38:31] <edheldil> never mind :)
[09:39:52] <fuzzie> well, we had a really nasty bug for some time which caused file loading to be limited by the graphics drawing rate
[09:40:13] <wjp> ah yes, the progress bar. That was fun
[09:45:01] <fuzzie> edheldil: well, it's sitll worth thinking about
[09:45:14] <fuzzie> admittedly it does not have my full attention this morning either
[10:27:40] --> fizzle has joined #gemrb
[11:13:47] <-- WingedHussar has left IRC (Ping timeout: 240 seconds)
[11:14:46] --> WingedHussar has joined #gemrb
[11:49:23] --> traveler has joined #gemrb
[13:07:38] --> brada has joined #gemrb
[13:22:48] --> rocket_hamster has joined #gemrb
[13:54:20] --> Coriander2 has joined #gemrb
[13:55:29] <-- Coriander has left IRC (Ping timeout: 246 seconds)
[15:06:18] <brada> fuzzie: do you think i should go ahead and commit that patch i posted last night?
[15:06:20] --> Fuzzlix has joined #gemrb
[15:09:02] <brada> this is the latest iteration: http://paste.debian.net/243409/
[15:09:05] <Seniorita> debian Pastezone
[15:27:11] <-- Coriander2 has left IRC (Ping timeout: 246 seconds)
[15:30:09] --> Coriander has joined #gemrb
[16:05:06] <-- WingedHussar has left IRC (Quit: WingedHussar)
[16:16:54] <-- Fuzzlix has left IRC (Quit: ChatZilla 0.9.90 [Firefox 19.0/20130215130331])
[16:53:13] <-- rocket_hamster has left IRC (Quit: bye!)
[17:17:34] <-- fizzle has left #gemrb
[17:28:29] <-- brada has left IRC (Quit: brada)
[17:57:35] --> brada has joined #gemrb
[18:01:21] --> Yoshimo has joined #gemrb
[18:12:59] <-- |Cable| has left IRC (Ping timeout: 256 seconds)
[18:13:58] --> |Cable| has joined #gemrb
[19:29:29] --> kingron has joined #gemrb
[20:22:21] <-- brada has left IRC (Quit: brada)
[20:23:56] --> brada has joined #gemrb
[21:05:28] <-- Yoshimo has left IRC (Quit: Yoshimo)
[21:23:48] --> nutron has joined #gemrb
[22:11:37] <-- kingron has left IRC (Quit: Leaving)
[23:04:47] <-- brada has left IRC (Quit: brada)
[23:27:49] <-- lynxlynxlynx has left IRC (Remote host closed the connection)
[23:30:29] --> brada has joined #gemrb