00:00:00 --- log: started forth/04.03.28 00:00:59 --- join: imaginator (~gps@georgeps.dsl.xmission.com) joined #forth 00:14:08 --- quit: kc5tja ("THX QSO ES 73 DE KC5TJA/6 CL ES QRT AR SK") 00:20:11 Hi 00:20:59 hi 00:21:17 I'm having problems implementing a "do word repeat" or "do.n.times" word in assembly for my Forth. 00:21:42 It seems like the obvious way to implement "do word repeat" would be to make do push the address of %eip 00:22:06 and then have repeat take that address from the control stack, but gas doesn't have %eip. 00:22:09 What should I be doing? 00:22:47 well, gas sucks then... 00:22:48 :P 00:23:23 well, you need to push the address of the instruction _after_ do on the return stack 00:23:31 (which, it seems, is the hardware stack here) 00:23:57 That's kind of a problem, because I want most words to be relocatable. 00:24:19 I'm writing to mmap'd memory these instructions. They will basically serve as tasks later on that may be run at any time. 00:25:25 Is there a way to use a raw opcode and movl %esp,%eip even though gas doesn't support it? 00:25:49 well 00:25:51 actually 00:25:54 if you think about it 00:25:54 err movl %eip,(%esp) would be it in gas 00:25:59 thats not what you want :P 00:26:07 * arke made a mistake 00:26:11 Why not? 00:26:22 you just push the current location on the compile-time stack 00:26:50 then, at the repeat, you can do a little pointer math, and do a relative jump =D 00:27:08 How can I get the current location, unless I iterate over all stored words and look for what needs to be relocated? 00:28:53 * imaginator suspects he doesn't understand arke 00:28:56 heh 00:29:13 while you compile it, you store the current dictionary location somewhere 00:29:42 then, at the repeat, you subtract that from the current dictionary location, and do a relative jump back on zero 00:29:51 or rather, not zero 00:35:01 My dictionary setup is much different. I don't think that would work. 00:36:18 --- quit: ayrnieu ("Leaving") 00:37:13 My dictionary isn't linear. 00:37:29 Words are associated with string keys in a trie. 00:39:27 heh 00:39:49 well, then you're gonna have to take the address of whatever its compiling to at compile time 00:46:32 :( 00:46:55 Does an app have to run in a lower ring level than unix puts user apps in to access things like %eip? 00:47:17 * imaginator wonders how to find the opcode. 00:47:22 eh, no, you can always access your own eip lol 00:47:53 call location ; location: .... 00:48:01 after that, location is on the stack 00:48:04 but thats a hassle 00:48:17 It's still better, and should be relocatable. 00:48:23 you don't need eip though :) 00:48:38 I know, but I can make the code relocatable more easily this way. 00:48:55 My then keyword does a relative jump like you were describing by subtracting offsets. 00:49:52 yeah, thats what you want to do here too 00:50:26 * arke is away: sleep 00:54:57 bye arke 00:55:02 sleep well 01:46:36 --- join: fridge (~fridge@dsl-203-33-165-176.NSW.netspace.net.au) joined #forth 02:19:34 --- join: yeoh (~yeoh@219.95.1.19) joined #forth 02:20:00 --- part: yeoh left #forth 03:51:32 --- quit: crc (Nick collision from services.) 03:51:43 --- join: crc (crc@1Cust201.tnt1.levittown.pa.da.uu.net) joined #forth 04:13:22 --- quit: crc ("Goodbye") 04:17:01 --- part: imaginator left #forth 04:59:18 --- quit: Robert ("brb") 05:20:16 --- join: Robert (~snofs@c-2c5a71d5.17-1-64736c10.cust.bredbandsbolaget.se) joined #forth 06:26:07 --- join: hovil (~hovil@CommSecureAustPtyLtd.sb1.optus.net.au) joined #forth 06:26:55 Hi hovil 06:41:09 hi 06:42:22 I work with people who either do nothing (leaving it to me to do it), or do things incorrectly (leaving me to fix it, and then do it correctly) *sigh* 06:49:05 They fall between the following 06:49:12 Stupid and lazy: these don't produce very much, but they don't hurt you. Educate and motivate at your leisure, if possible. 06:49:18 Stupid and motivated: these folks need to be weeded out ASAP, since they actively hurt you. You know, the ones that form "cross-cutting teams" to solve trivial problems. Or add software features that accidentally send mail to everyone in your address book. 06:55:06 --- quit: networm (wells.freenode.net irc.freenode.net) 06:55:07 --- quit: ianp (wells.freenode.net irc.freenode.net) 06:55:07 --- quit: mur (wells.freenode.net irc.freenode.net) 06:55:07 --- quit: cmeme (wells.freenode.net irc.freenode.net) 06:56:13 --- join: networm (~networm@L0625P09.dipool.highway.telekom.at) joined #forth 06:56:13 --- join: mur (~mur@mgw2.uiah.fi) joined #forth 06:56:13 --- join: cmeme (~cmeme@216.184.11.30.unused.swcp.com) joined #forth 06:56:13 --- join: ianp (~ian@69.20.54.28) joined #forth 06:56:14 lol 06:56:38 features :p 07:01:04 --- join: uglymonkey (c-dbvvn-XP@d94.as5.sfld.mi.voyager.net) joined #forth 07:18:47 --- quit: networm (Read error: 113 (No route to host)) 07:19:31 --- join: networm (~networm@L0651P06.dipool.highway.telekom.at) joined #forth 08:29:07 is a variable that's defined with does> read-only? 08:29:41 A read-only variable sounds contradictive 08:29:47 it does, doesnt it 08:30:02 : numvar create 1 cells allot does> @ ; 08:30:33 when i create a new var, numvar test , i cant seem to assign a value to it 08:31:37 i was kind of looking for the behaviour of a constant 08:32:06 Actually many Forths (like IsForth) uses "constants" as variables 08:32:17 And special words are used to change their values 08:33:21 oh 08:33:36 i seem to (attempt to) reinvent the wheel over and over again :p 08:33:42 if you remove the @.. it should work 08:33:53 the @ already fetches what is in the variable 08:33:56 any default word to alter the constant? 08:34:21 yea i know networm, that kind of was the whole point of it... 08:34:35 but if i do, 21 test drop ! it returns an exception 08:35:32 IsForth uses !> 08:35:35 I use -> 08:35:44 So to set X to 123, do: 123 -> X 08:36:13 : numvar create 1 cells allot does> ; numvar test 08:36:21 21 test ! 08:36:25 that should work 08:36:36 oh, you mean, that does> automaticly executes @ ? 08:37:12 it automatically puts the address to the stack 08:37:19 test @ 08:37:24 well yea, but you'd need @ to get the value 08:37:27 ^ that will get back what is stored in the variable 08:37:31 i know :) 08:37:37 but i just wanted ot use test 08:37:40 get rid of the @ 08:38:17 i see.. 08:39:45 not sure that's the best solution.. but you could just make another word for it.. 08:39:59 that would kind of defy the whole point 08:40:09 either i put @ there, or get-variable-value 08:40:11 :p 08:40:55 hm. so the question is, can you get the address where the value is stored inside TEST 08:41:33 no, the question is how to get a variable that behaves like a constant but can be changed like a regular variable :) 08:42:11 well, you need some additional word to change it 08:42:38 something like: 21 change test ! 08:42:46 yea i can live with that 08:42:53 CHANGE should somehow find that address.. 08:43:03 but that brings me back to my first question :) 08:43:09 wasn't there a word to get the datafield address from the XT? 08:43:23 although not really, but why wouldnt it work? 08:43:32 : numvar create 1 cells allot does> @ ; 08:43:34 numvar test 08:43:39 1 test drop ! 08:43:45 oh 08:43:51 because there's no address put on the stack 08:43:56 riiight 08:44:03 yes.. the address is the problem.. 08:44:31 hrm 08:45:11 i guess that makes clear why constants are constants :p 08:47:28 --- join: fdr (fdr@modem-1898.jaguar.dialup.pol.co.uk) joined #forth 08:48:15 21 ' test body> ! 08:48:56 hm,doesn't seem to work in gforth 08:49:05 that didnt work in winforth either 08:49:12 whateverthehell it did, the value is still 0 :p 08:49:19 what does body> do 08:49:35 get the data address from the XT 08:49:45 --- part: fdr left #forth 08:49:47 doesnt ' already do that? 08:49:49 qFox, you want the TO word. 08:49:54 And VALUE 08:49:55 undefined 08:50:39 value does what i want? 08:51:03 Yep. 08:51:08 TO isn't undefined in Winforth 08:51:14 Do this... 08:51:16 is too :p 08:51:24 Nope, it's not. You're just confused by the error. :P 08:51:32 10 VALUE TEST 08:51:34 TEST . 08:51:38 20 TO TEST 08:51:39 TEST . 08:52:13 ! it worked 08:52:13 TO is a parsing word 08:52:13 !fsee TO 08:52:13 Forth: 4Undefined. 08:52:21 but 08:52:26 its defined when you call value? 08:52:33 ? 08:52:46 Rephrase? 08:52:49 well 08:52:51 it didnt work before 08:52:55 and it did work after 08:52:57 ... :) 08:53:05 No... TO was giving you a confusing error. 08:53:10 If you just type TO... 08:53:13 oh 08:53:15 ic 08:53:15 and then hit CR 08:53:23 It's not finding a word to set anything TO 08:53:29 yea understand 08:53:36 TO 08:53:44 TO essentially has a ' in it 08:53:56 : TO ' >body ! ; 08:54:03 Simple way to define it. 08:54:16 that's just what i tried though.. 08:54:20 ' >body ! 08:54:36 * arke is back (gone 08:04:08) 08:54:40 !fdef : TO ( n -- ) ' >body ! ; \ TO 08:54:41 Forth: 12: TO 3( n -- ) ' >body ! 12; 3\ TO 08:54:53 how do you indicate a parsing word in the stack anyways? 08:55:00 21 ' test body> ! 08:55:04 wonder why that doesn't work 08:55:07 >body 08:55:12 ahh 08:56:59 : value create , does> @ ; 08:57:07 yeah, with >body it works :) 08:57:10 : to ' >body ! ; 08:57:12 Heh, yep. 08:57:24 grrr @ whoever invented body> 08:57:25 cfa>body would be more obvious a word. 08:57:32 definitely 08:57:32 body>cfa 08:57:37 yep 08:57:40 or even body->cfa 08:57:45 body2cfa 08:57:53 -> could mean "store in" though 08:58:01 body-to-the-word's-cfa 08:58:02 --- join: Guest300757 (xc@200.163.162.31) joined #forth 08:58:05 Nice and verbose. ;) 08:58:10 heh 08:58:20 not very fortish 08:58:30 but 08:58:34 --- part: Guest300757 left #forth 08:58:38 isnt value then the same as constant? 08:59:18 Depends on how constant is defined. 08:59:53 the general definition of constant is that, aside from optimization stuff code thingies? 09:00:10 Well, try doing TO to a CONSTANT. 09:00:19 * arke is away: brb 09:00:37 good point 09:00:40 * qFox tries 09:01:02 21 constant ding ok 09:01:02 43 to ding 09:01:02 Error: ding Use TO and +TO on VALUE type words! 09:01:08 evil!! 09:01:12 But yea, if you created a VALUE and a CONSTANT using CREATE/DOES> it would be identical IMO. 09:01:19 it works in gforth 09:01:24 IMO, might as well just use VALUE everywhere. 09:01:26 constant/value seems to be the same there 09:01:36 If you really want a constant constant, then do : FOO 123 ; 09:01:45 see value 09:01:45 : Value 09:01:45 (Constant) , ; ok 09:01:59 WTF! Why can't I use SEE in gForth? 09:02:34 see constant 09:02:34 : Constant 09:02:34 (Constant) , ; ok 09:02:38 lol 09:02:39 works for me.. :) 09:03:07 *grumble* 09:04:30 !fhelp 09:04:31 Forth: This is a words database. Add words with !fdef . See words with !fsee . Use !fhelp for more information. This bot is not a parser, so you cannot use it for running code. 09:04:31 Forth: Only define words in Forth, define words that really cannot be defined that way, as CODE. This is a public bot, dont ruin it. 09:05:09 it should be fed with gforth or winforth initially.. 09:05:17 i'm lazy 09:05:23 and winforth is just code. 09:05:31 MINFORTH 09:05:32 except for the words you least expect it from 09:05:41 yes, minforth then, it has most things in forth 09:05:46 * arke is back (gone 00:05:24) 09:06:13 !fhelp owner 09:06:13 Forth: This is a words database. Add words with !fdef . See words with !fsee . Use !fhelp for more information. This bot is not a parser, so you cannot use it for running code. 09:06:13 Forth: Only define words in Forth, define words that really cannot be defined that way, as CODE. This is a public bot, dont ruin it. 09:09:52 !fsee dup 09:09:52 Forth: 12: dup 3( n -- n n ) whocaresitsabloodyprimitive 12; 3\ duplicates the value on the stack 09:10:09 heh 09:10:11 networm> me. 09:10:25 oh, heh 09:10:35 :) 09:11:21 networm, still gonna make a bot out of your forth VM? 09:11:22 --- join: blockhead (default@dialin-694-tnt.nyc.bestweb.net) joined #forth 09:11:57 still planned. 09:12:37 Hey, if we had enough Forth bots in here, we could have Forthbot wars. :) 09:12:49 heh 09:13:11 is someone runnign a forth bot here? cool! 09:13:12 Sounds great. 09:13:36 how would they fight? try to kick the others? 09:13:55 hahaha this bot is a stupid bot though :) 09:13:59 : foo recurse ; foo 09:14:00 probably call each other names and contradic teach other on how [ and ] work :D 09:14:01 just database style 09:14:06 lol 09:14:42 i'm gonna change the !fdef trigger soon so it'll be a bit more lenient to definitions 09:16:07 networm, well they'd probably each have to implement a certain wordset that they would use to battle with. Making the bot exit or freeze would be too easy. :P Better to make them play some kind of game. 09:16:57 something like C robots should work 09:17:04 Yea. 09:17:14 but if wouldn't matter then which bot runs, but which script it runs 09:17:27 Yep. 09:17:42 Could have a server bot and then different fighting bots. 09:17:54 heh 09:18:01 They'd /msg back and forth with the server for info and commands. 09:18:16 and the server would /msg important stuff to the channel :) 09:18:28 * networm bot has won the match 09:18:30 haha 09:18:35 Hehe. 09:19:29 Hmm... and perhaps the bots could load programs from the server in mini-blocks. 09:19:32 --- join: Herkamire (stjohns@h000094d30ba2.ne.client2.attbi.com) joined #forth 09:19:38 Hi Jason :D 09:19:48 hi Chris :) 09:20:27 :) 09:20:29 (: 09:28:28 * arke is away: shower + wanking 09:30:57 Win32Forth is actually pretty neat. 09:32:13 * blockhead installed it awhile ago but never got around to trying it out :D 09:33:20 I'll have to actually try doing some Win32 programming in it. 09:33:32 It's cool how you can TURNKEY apps with it and stuff. 09:33:56 turnkey is kinda good! 09:35:21 Nice cool Forthy alternative to commercial dev environments like MSVC. 09:36:09 * blockhead really should take a look at it 09:38:05 i got a window running 09:38:14 now trying to create some code that will create a filename 09:38:35 there are probably a whole set of standard words for it already, but w/e :) 09:40:34 Win32Forth has a LOT of words. 09:40:44 it does 09:41:07 and the editor is a bit bugy when it comes down to undo/redo 09:41:22 which can be quite annoying sometimes 09:45:13 I have only used the editor to browse thus far. 09:51:56 --- join: kc5tja (~kc5tja@66-91-231-74.san.rr.com) joined #forth 09:52:02 --- mode: ChanServ set +o kc5tja 09:52:36 I just unleashed a 55-gallon container of liquid white phosphorus in the ColorForth mailing list. 09:53:07 umm, you flamed someone? 09:53:22 you claimed winforth ruled all others? 09:54:16 hey all 09:54:36 and if you didnt, can i give it a try? ;) 09:54:46 qFox, is the winforth gui written in winforth? 09:54:56 i think so.. 09:55:00 the editor is 09:55:02 is it OSS? 09:55:06 oss? 09:55:10 open source 09:55:20 oh, um i dont think so, perhaps on the website 09:55:34 http://groups.yahoo.com/group/win32forth/ 09:55:54 oh wait, wrong link 09:56:01 i had some great ideas for the factor IDE 09:56:15 it will have a presentation-based UI with html, irc, console and gui backends 09:56:47 and probably more words than winforth :) 09:56:52 hum i seem to have misplaced my winforth homepage link 09:56:57 thats a lot of words 09:57:15 3,181 Words in Application dictionary 09:57:15 1,689 Words in System dictionary 09:57:15 4,870 Words total in dictionaries 09:57:15 15,513 Windows Constants available 09:57:18 yep. 09:57:27 ooof 09:58:34 so you can write gui apps in winforth? 09:58:43 Oh, yeah, you can kind of say that I flamed someone. 09:58:53 After they proceeded to tell me how much "I don't know what I'm doing." 09:59:09 oopsie 09:59:12 Even despite my rather influential list of credentials in the areas *precisely* that I'm talking about. :) 09:59:15 kc5tja, will the boxos be any kind of 'color forth', or just plain text? 09:59:20 what DID you do 09:59:23 :o 09:59:56 slava: Well, now that I'm considering the addition of proper color, the thought has crossed my mind. But it's too early to tell. I'm thinking 'no' at this point, because it'll be more marketable to a wider audience that way. 10:00:22 qFox: I'm not sure if the archives ahve the message yet. If so, I'll see if I can forward you the link. 10:00:45 what kind of bomb did you drop? 10:00:58 any survivors :D 10:01:00 are those archives still "alive" anyways? i mean, the language is kind of ancient and all? 10:01:01 ? 10:03:48 http://www.strangegizmo.com/forth/ColorForth/ <-- Towards the end of the page, "DARPA Takes Aim at..." 10:04:09 That remains to be seen. 10:04:34 You have to read the thread from the beginning to get the full understanding of how this guy came to think several of us on the list "don't know what we're doing." 10:05:09 who is 'this guy'? 10:05:14 qFox: ?? What do you mean the 'language is kind of ancient and all'? 10:05:25 forth 10:05:26 old 10:05:33 hovil: John. R. Strohm 10:05:40 everything thats being said about it, has been said by now? 10:05:49 qFox: What? 10:05:56 qFox: Where did you get that idea from? 10:06:06 it was a question 10:06:08 hahaha 10:06:09 That's like saying, "Everything that can be said about Lisp has been said already." 10:06:13 Clearly it's not true. 10:06:35 everything than can be written must have already been done so 10:06:45 I mean, gads! writing systems have been around for ages 10:07:03 there are only 12 notes so all the music that can be written is already written :D 10:07:05 hovil: That's been said before. Come up with something NEW...geez... :D 10:07:18 kc5tja, I've never heard it before 10:07:26 hovil: Sorry, was a stupid joke. 10:07:32 I live the simple life 10:10:14 --- join: bitwise (bitwise@69-166-74-248.atlsfl.adelphia.net) joined #forth 10:11:31 --- part: bitwise left #forth 10:12:08 kc5tja, I thought you meant it was a famous quote or something =P 10:12:42 I wish I had bought my assignment to work 10:12:55 it is quite slow now, after the initial rush 10:15:55 though, being 4am, my mind is quite slow too 10:16:08 kc5tja> you kurt? 10:20:19 --- quit: blockhead (Read error: 104 (Connection reset by peer)) 10:20:35 it is quite slow now, after the initial rush 10:20:39 arke, RUSH!!! 10:21:25 --- join: blockhead (default@dialin-704-tnt.nyc.bestweb.net) joined #forth 10:23:59 qFox: No, I'm Samuel A. Falvo II 10:24:05 kc5tja> hm, ok, before i read a whole bunch of posts that say all the same, when does it get intresting? :) 10:24:18 all teh same, or the opposite from the previous post... 10:24:31 I already said, towards the bottom of the page. 10:24:55 that was before i knew who you were :) 10:33:23 --- quit: uglymonkey (Read error: 110 (Connection timed out)) 10:59:15 hi kc5tja 10:59:32 * arke is away: I'm busy 10:59:50 * arke is back (gone 00:00:00) 11:03:27 * kc5tja is reading up on consistent overhead byte stuffing... 11:03:43 It's a sweet system; I proposed something before, but it wasn't as systematic and thorough as Cheshire's work. 11:04:39 ? 11:04:41 hwats that? 11:06:00 You first need to understand how packet data is exchanged on a serial link. 11:06:18 Generally speaking, you have a source address, destination address, some protocol ID, the data, and then some kind of checksum. 11:06:31 Well, that's all good and fine, but how do you identify where a packet starts and ends? 11:07:03 Typically, one would use "framing" information -- SLIP uses $C0 as a start-/end-of-frame flag byte, 11:07:12 $7E is used by PPP and most other HDLC-derived protocols. 11:07:35 The problem is, RS-232 and other (a)synchronous serial chips aren't aware of what a packet is. 11:07:40 It just knows to send a byte whenever it has one. 11:07:55 So how does one differentiate the $7E-as-packet-flag-byte from the $7E-as-data-byte? 11:08:29 Well, traditional solutions involve bit-stuffing (best implemented in hardware), where after five consecutive '1' bits in the data, a '0' bit is conveniently and automatically injected, to prevent any accidental occurance of the flag in the data. 11:09:00 Problem is, in software, this is horribly inefficient to pull off. Plus, it adds 20% network overhead for worst-case data streams (on average, it's about 1% over the life of a single connection, but still). 11:09:10 Another, faster solution is to use byte stuffing. 11:09:31 Basically, everywhere you see a flag byte in the data, replace it with an escape character, followed by an indicator of what to escape. 11:09:54 This is fast to do, but since it now takes two bytes to represent what used to be only one, packet sizes can grow to 200% their original size. Ouch! 11:10:19 What COBS does is basically a variation of RLE-encoding. 11:10:31 run length encoding encoding? 11:10:38 Let's say, for the sake of argument, that $00 is our flag byte. We choose this byte because of conveniences you'll see later on. 11:10:40 don't answer that, I'm being a pest 11:11:02 So whenever we see $00 in the data stream, we know unambiguously that a new frame is starting, and any previuos frame is done. 11:11:27 If the received byte is $01, then this encodes a raw $00 byte. 11:11:46 If we see $02, then we encode ONE data byte, which is IMPLICITLY UNDERSTOOD to have a $00 byte following it. 11:12:03 So $02 $41 received on the wire will cause $41 $00 to be placed in the buffer. 11:12:13 $03 $41 $42 will cause $41 $42 $00, etc. 11:12:20 oh, thats cool 11:12:22 I like that 11:12:30 Thus, given a prefix byte N, we have N-1 bytes of pure data, which is implicitly followed by a zero byte. 11:12:55 For sequences longer than 254 bytes, a special exception is used: $FF, which means 254 raw bytes of data, but NO implicit zero byte. 11:13:11 This means, best case, 1:1 packet sizes are possible. Worst case: 1:1.04. :) 11:13:34 AND, this kind of encoding is utterly trivial to do for microprocessors. 11:13:37 And check this out: 11:13:39 $00 == Flag byte 11:13:43 $01 == implicit zero 11:13:49 $FF == 254 non-zero-terminated bytes. 11:13:57 Note that all these surround $00. 11:14:04 =D 11:14:11 Thus, in 6502 machine language for example, we can *very* rapidly check for these special cases: 11:14:15 LDA INPUT 11:14:19 BEQ NewFrame 11:14:26 DEC A 11:14:32 BEQ AppendZeroByte 11:14:55 err, oops, wrong order of instructions. 11:15:03 And let me use X register instead of A, so it'll work on all 6502 clones. 11:15:05 LDX INPUT 11:15:10 INX 11:15:25 BEQ TwoFiftyFourBytesFollow 11:15:27 DEX 11:15:33 BEQ NewPacketFollows 11:15:37 DEX 11:15:41 BEQ AppendZeroByte 11:15:50 sounds good for packet radio 11:15:59 ; At this point, X contains number of raw bytes that follows. Just append $00 when you're done. 11:16:20 Or, alternatively and perhaps even faster: 11:16:22 LDX INPUT 11:16:31 gahh 11:16:32 LDA INPUT 11:16:48 SHL A 11:16:49