#pentagram@irc.freenode.net logs for 2 Nov 2003 (GMT)

Archive Today Yesterday Tomorrow
Pentagram homepage


[00:37:43] <-- Fingolfin has left IRC ("42")
[01:45:18] <-- watt has left IRC ("using sirc version 2.211+KSIRC/1.2.4")
[02:22:36] <-- Dark-Star has left IRC (""ZZZzzz..."")
[04:15:17] <-- Kirben has left IRC (Read error: 104 (Connection reset by peer))
[04:15:23] --> Kirben2 has joined #pentagram
[04:30:43] --- Kirben2 is now known as Kirben
[05:05:39] <-- Kirben has left IRC (Read error: 104 (Connection reset by peer))
[05:16:13] <-- SB-X has left IRC ("brb")
[07:09:18] --> Kirben has joined #pentagram
[07:09:18] --- ChanServ gives channel operator status to Kirben
[07:22:52] --> SB-X has joined #pentagram
[08:05:29] * SB-X is away: sleep
[10:16:17] --> Fingolfin has joined #pentagram
[10:16:17] --- ChanServ gives channel operator status to Fingolfin
[11:54:36] --> Dark-Star has joined #pentagram
[13:26:15] <-- Kirben has left IRC ("System Meltdown")
[14:18:12] <servus> Coren_ and I are having a 3D Duel!
[16:50:07] --> watt has joined #pentagram
[17:14:13] <watt> hey all..
[18:09:23] <watt> I've got a small cleanup for the combat patch, but I added a new struct to do so.. Is there a way to include new files in a diff?
[18:14:10] <Coren_> watt: diff against an empty file.
[18:14:24] <Coren_> patch knows to Do The Right Thing(tm) then.
[18:42:24] * SB-X is back (gone 10:36:42)
[18:48:31] <watt> using cvs diff? or just diff?
[18:48:48] <Coren_> cvs diff is different, you want to:
[18:48:51] <Coren_> cvs add new_file
[18:49:11] <Coren_> Before you do the commit.
[18:50:17] <watt> but.. adding takes commit access.. And I really doubt you guys want to give me commit access just yet.
[18:50:39] <watt> (Because I wouldn't give me commit access yet)
[18:50:50] <Coren_> "You guys" excludes me. :-) This isn't my project.
[18:50:56] <Coren_> Then just do the empty file trick.
[18:51:05] <Coren_> >old/new_file
[18:51:18] <watt> ah
[18:52:44] <Coren_> It /is/ confusing. There is this set of rewrite projects within which many members participate in the others. :-) Like minds and all, I guess.
[18:53:42] <watt> well.. google time..
[18:54:17] <Coren_> I lead LoW, participate occasionaly in uwadv, gave a (small) hand for nuvie and am a fan of exult and pentagram. Willem (wjp) leads pentagram, is a major contributor of exult, participates in uwadv and low, etc...
[19:06:39] <watt> ewww.. It's suggested to append the diff with the extra diffs for the new files.
[19:07:29] <Coren_> Nah; just do the empty file trick. It works fine.
[19:08:17] <Coren_> Add an ampty file named like the old one in the origina dir then diff yours against the original. The diff will include your new file diffed from 'nothing' and patch understands.
[19:08:38] <Coren_> s/named like the old one/named like the new one/
[19:09:41] <watt> right, but the diff from the other files is still done using cvs diff -u
[19:10:27] <Coren_> Do the entire diff the same way-- you can request a diff of one dir against another which is probably what you want.
[19:10:51] <Coren_> It'll make a diff of everything you changed (or added)-- this way garantees you didn't forget anything too.
[19:11:22] <Coren_> Oh, I get it
[19:11:29] <Coren_> Don't use cvs diff.
[19:11:56] <Coren_> checkout a new copy of the tree in /another/ directory and diff one directory against the other
[19:12:32] <Coren_> [make sure you check out the same versions you modified, or cvs update in your tree, otherwise your diff will include 'undo' of the revisions since]
[19:14:20] <Coren_> At least that's how I do it when I diff against a cvs tree I don't have commit access to. It garantees your patches are good.
[19:14:35] <Coren_> I did this all the time with Xfree86.
[19:14:46] <watt> oh.. well, that should be simple enough..
[19:28:11] <-- Fingolfin has left IRC ("42")
[19:29:03] <watt> so I'm guessing somthing like diff -Nru <old-dir> <new-dir>
[19:30:05] <Coren_> Yes, /if/ your new-dir is clean. Otherwise, drop the -N and add the file in old-dir.
[19:30:22] <Coren_> (If your new-dir has intermediate files, they will be added to the diff with -N)
[19:30:56] <watt> Yeah. I think you're right on that.. all my eclipse files are getting diffed too.. ugly.
[19:33:08] <watt> touches the CVS "entries" files too.. hmm. should I keep that in the diff?
[19:34:10] <Coren_> Normally not. There are a few arcane files where you'd want to (mostly having to do with repository maintenance) but otherwise they are just so much garbage.
[19:35:12] <Coren_> I.e.: given you don't have commit access, it's fair to presume you're not trying to mess with the repository itself; so no-- don't keep 'em.
[19:35:34] <Coren_> You may want to add --exclude='*CVS*'
[19:36:59] <Coren_> Or is that --exclude='.*CVS.*'? I can't remember offhand if it's a glob or a regexp. Try both, one will work. :-)
[19:37:34] <watt> well *CVS* returned a pretty clean diff
[19:38:26] <Coren_> Then that's the one to use.
[19:39:57] <Coren_> What's 'eclipse' anyways? :-)
[19:42:04] <watt> It's a damn fine IDE for java. Multi platform and limited C/C++ support as well
[19:42:51] <Coren_> Ah. Call me old-school, but I like my current ide fine. It's called 'bash'. :-)
[19:44:38] <Coren_> Typing 'make' is faster for me than removing my hands from the keyboard, digging the mouse from under the papers and books that accumulated over it since I last touched it, and move/clicking it. :-)
[19:44:38] <watt> right.. I'm a big vim fan... but it just wasn't adequate when I started programming java.
[19:45:51] <watt> and I like having tabs for multiple files
[19:46:22] <Coren_> ^W^W does fine for me. :-)
[19:47:33] <Coren_> I don't blame you, though, I guess I'm just an old dog. When you've been coding for >18 years one way, anything else just doesn't work. :-)
[19:48:09] <Coren_> Besides, IDEs just won't work well when I code from my balcony with my trusty vt420. :-)
[19:49:04] <watt> yeah.. realizing I needed 64+ free megs just to program kinda did bum me.
[19:50:33] <watt> hmm my code does use tabs.. but pentagram does.. best fit to spec.
[19:51:09] <watt> think a vim search and replace is a godd way to change the diff?
[19:51:53] <Coren_> That's iffy because it's easy to break. Better fix the source before the diff, usually.
[20:02:18] <watt> I gotta get used to not using expand tabs again
[20:03:34] <Coren_> :g/^\(\t*\) /s//\1\t/g
[20:03:36] <Coren_> :-)
[20:13:18] <Coren_> Hrmph. I'm trying to figure out a /fast/ test for collinearity of three points. I can't think of anything that doesn't involve two square roots and a dot product.
[20:15:07] <watt> it's been a while since my linear algebra stuff.. I got nothing.
[20:15:36] <Coren_> I'm *sure* there is a fast test if all you're interrested in is /testing/ for colinearity.
[20:16:01] <watt> Yay.. my patch look much cleaner.
[20:16:40] * Coren_ points out (perhaps belatedly) that you can request that diff ignores changes in whitespace.
[20:17:13] <Coren_> -b iirc
[20:18:24] <Coren_> Though that wouldn't help if you changed indentation of some things.
[20:20:56] <watt> oh crap.. I never uploaded to old patch.. needed to click a checkbox. doh
[20:27:01] <Coren_> Heh. Got it. Try to find the plane going through the tree point and watch for the degenerate case [ 0, 0, 0, 0 ]. :-)
[20:28:20] <Coren_> 6 adds and a dot product.
[20:28:39] <Coren_> err, cross product even
[20:30:33] * SB-X smiles and nods.
[20:31:40] <Coren_> SB-X: Ah, you knew all along and wanted me to figure it out myself? :-)
[20:34:13] <SB-X> I cannot confirm or deny this.
[20:34:20] <Coren_> How Zen.
[20:44:48] <Coren_> Hm. Is is usually faster to test for (fabs(foo)<EPSILON) or (foo<EPSILON && foo>-EPSLION)?
[20:57:30] <watt> I'd assume the double comparison. fabs doesn't appear to be a macro in the new gcc (and I'm not sure about the old)
[21:00:28] <Coren_> It maps to __builtin_fabs, actually (int bits/mathinline.h); but on modern CPUs the register write of fabs is probably more expensive anyways. I was asking in case someone knew for a fact already.
[21:20:16] --> wjp has joined #pentagram
[21:20:16] --- ChanServ gives channel operator status to wjp
[21:35:07] --> Fingolfin has joined #pentagram
[21:35:07] --- ChanServ gives channel operator status to Fingolfin
[21:35:51] <Fingolfin> hiya
[21:52:25] <watt> hi
[21:54:58] * watt apologizes to wjp. "I forgot to actually submit the patch previously"
[21:55:13] * watt curses upload checkbox
[22:44:47] <wjp> I'll try to make some time tomorrow night to take a closer look at it
[22:45:01] <wjp> right now I need to get some sleep, though :-)
[22:45:03] <wjp> night
[22:45:11] <-- wjp has left IRC ("Zzzz...")