00:00:00 --- log: started forth/01.01.19 00:08:16 --- part: adu left #forth 03:28:57 --- join: speuler (l@c38038.upc-c.chello.nl) joined #forth 03:29:14 hi 05:06:15 --- join: Fare (fare@ppp47-net1-idf2-bas1.isdnet.net) joined #forth 05:59:03 --- join: Stepan (stepan@Charybdis.suse.de) joined #forth 05:59:18 Hi :) 06:02:23 sleepy channel 06:03:36 seems like. is it only on the isforth project? 06:05:34 mainly it is but people tend to talk about non-isforth topics 06:35:58 --- quit: clog (^C) 06:35:58 --- log: stopped forth/01.01.19 06:36:39 --- log: started forth/01.01.19 06:36:39 --- join: clog (nef@bespin.org) joined #forth 06:36:39 --- topic: 'work resumed on isforth project - tcn added as developer' 06:36:39 --- topic: set by I440r on [Thu Jan 18 19:26:56 2001] 06:36:39 --- names: list (clog Stepan Fare speuler @I440r @aaronl) 07:02:44 --- quit: Fare (Ping timeout for Fare[ppp47-net1-idf2-bas1.isdnet.net]) 07:27:35 --- quit: Stepan (Ping timeout for Stepan[Charybdis.suse.de]) 07:27:49 --- join: Stepan (stepan@Charybdis.suse.de) joined #forth 07:33:07 --- quit: Stepan (Ping timeout for Stepan[Charybdis.suse.de]) 07:43:45 --- quit: aaronl (The name's X. Bitch X.) 07:44:03 --- join: aaronl (aaronl@vitelus.com) joined #forth 08:30:40 hi ppl! 08:46:33 --- mode: I440r set mode: +o aaronl 08:46:35 --- mode: I440r set mode: +o clog 08:46:37 --- mode: I440r set mode: +o speuler 08:46:47 speuler: u awake ? 08:51:04 * aaronl is 08:51:07 Unfortunately... 08:51:10 i'm sick 08:53:10 ur sick ? 08:53:17 i just got over being sicjk 08:53:20 well mostly got over 08:53:25 hope u dont have what i had 08:53:28 that shit was bad :P 08:53:38 I440r: the author did 08:53:39 I440r: debian got pretty pissed in fact 08:53:44 what was all that in reference too dood ? 08:54:04 oh 08:54:11 you were talking about the tar option change 08:54:13 about bzip2 08:54:20 i just wanted to point out that it wasn't debian's fault 08:54:38 aha yes 08:54:47 i get it! 08:55:13 so it was the maintaners idea :P 08:55:23 well i think debian is the only one that doesnt use I any more 08:55:36 every otehr distros tar uses I to extract bz2 :P 08:55:51 not the debian maintainer's 08:55:58 the latest version of tar changed it 08:56:04 other distros havent updated yet 08:56:13 aha i get it! 08:56:15 wow 08:56:19 debian updated first ? 08:56:23 thats never happned before 08:56:38 i remember waiting 8 months for the new version of windowmaker to get into .deb.form 08:56:41 uh yeah 08:56:46 debian is updated every day 08:57:24 yes but not every package 08:57:26 I hate this... I can't track down this memory corruption 08:57:34 somepacakge maintaners are a little slow 08:57:37 Program received signal SIGSEGV, Segmentation fault. 08:57:37 0x400ca6c3 in free () from /lib/libc.so.6 08:57:43 i'm a package maintainer :) 08:57:46 well 08:57:55 look at all the places wher you are allocating memory 08:58:00 uh huh 08:58:00 and all the places you are freeing it up 08:58:06 or where i'm forgetting to allocate it :) 08:58:10 comment out the free operation 08:58:21 oh 08:58:24 hmmm 08:58:25 it would be nice if it crashed the moment i did something bad 08:58:27 didnt think of that :P 08:58:46 i've narrowed it down to one function call 08:58:52 --- join: Fare (fare@ppp70-net1-idf2-bas1.isdnet.net) joined #forth 08:58:54 but i can't see any bad stuff there... 08:58:56 * aaronl looks harder 08:59:23 hrm 08:59:26 look for missing * 08:59:28 heh 08:59:42 i even made a single exit point :P 08:59:45 or a for that loops one too many :) 09:00:06 i'm going to comment out parts of this function 09:00:11 and narrow it down even further 09:01:07 eeek 09:01:30 that's bad 09:01:42 correct 09:01:46 i commented out the part of the function that actully does stuff and it still crashes 09:01:50 thats a very good debug method 09:01:57 another one is to put messages in the source 09:02:08 yeah 09:02:08 printf("made it this far\n"); 09:02:12 i do that a lot 09:02:19 but the thing is, i'm running it under a debugger 09:02:22 and it makes it the whole way 09:02:27 but it corrupts ram in the process 09:02:41 heeheh 09:02:45 :( 09:02:50 what debugger 09:02:51 this SHOULD be done in forth 09:02:52 gdb 09:02:57 ugh 09:03:06 thats the biggest piece of shit debugger i have EVER used 09:03:15 i would NEVER trust it for any real debugging 09:03:25 put instruction pointer at start of code 09:03:33 put breakpoint on bottom line of window... 09:03:36 run code 09:03:40 it stops half way down 09:03:42 run again 09:03:49 it stops 3 insttructions past the breakpoint 09:03:50 bah 09:03:53 great debugger 09:03:59 yup 09:04:08 are there good forth debuggers? 09:04:12 i might suggest NOT using it 09:04:21 heh 09:04:24 for fpc hehe 09:04:32 --- mode: I440r set mode: +o Fare 09:04:52 ahha 09:04:59 i think i found it 09:05:23 cool 09:06:04 oh wait 09:06:10 i just commented out the allocation :( 09:06:19 that doesn't help :( 09:06:43 hehe 09:06:50 how big is this code 09:07:18 27k of c 09:07:23 well 09:07:27 the file i'm working on 09:07:34 it's part of a much larger app 09:07:36 ok... dont paste it here heh 09:07:45 heh 09:08:08 gdb is great 09:08:14 803 m->header = ctx->hdrs[i]; 09:08:20 (gdb) print m 09:08:20 No symbol "m" in current context. 09:08:55 hehehehe 09:09:07 what project is this and is it 100% your code ? 09:09:15 no, it's not my code 09:09:23 it's kinda s33krit right now 09:09:33 hehe 09:09:36 shhhh 09:09:40 dont tell ne1!!!! 09:09:40 i'm adding a really cool mail header cache to mutt 09:09:49 cool 09:09:55 so, i found something: 09:09:59 $5 = 1075175432 09:10:03 that SHOULD have been 0 09:10:06 heh 09:10:20 o wait 09:10:24 i had the initializer commented out 09:10:25 HEH 09:10:32 --- quit: Fare (Connection reset by pear) 09:10:43 --- join: Fare (fare@ppp70-net1-idf2-bas1.isdnet.net) joined #forth 09:10:43 :P~ 09:11:06 AHHA 09:11:36 i've narrowed it down to the "evil" function with multiple exit points 09:11:57 static sum_list *sum_get_new_node (sum_hash *h) 09:11:57 { 09:11:57 if (h->usednodes == NODE_BLOCK_SIZE) 09:11:57 { 09:11:57 safe_realloc (&(h->nodes), sizeof(sum_list*)*h->nodeslen+1); 09:11:58 h->usednodes = 1; 09:12:00 return h->nodes[h->nodeslen++] = malloc (sizeof(sum_list)*NODE_BLOCK_SIZE); 09:12:02 } 09:12:04 return &((h->nodes[h->nodeslen])[h->usednodes++]); 09:12:06 } 09:12:10 i'm gonna single-step it 09:12:27 good lukk :) 09:12:28 heh 09:13:10 working on isforth>? 09:13:51 yes... well havent done so other than talking to tcn and planning, tcn has made a very nice change to my macros that i like and i was waiting for him to check the new code in 09:13:59 but he was unable to do it. he dcc'd it all to me to check in heh 09:14:10 ill check it in... then check it out (pun intended :) 09:14:23 this function SUX for single stepping 09:15:20 u dont say :P 09:15:30 break it up 09:15:55 x=&(a->nodes); 09:16:03 y=sizeof(......)( 09:16:15 safe_realoc(x,y) 09:16:18 or somethignlike taht 09:16:26 so you- can test EACH part 09:16:40 i'm doing that 09:17:12 k 09:22:03 aaronl: what a horrible language you use! 09:22:22 what's that for? 09:24:46 hehehe 09:24:53 fare awakenz! 09:24:56 --- mode: I440r set mode: +o Fare 09:27:20 --- join: edrx (edrx@200.240.18.61) joined #forth 09:33:32 hmm 09:33:36 i think imight have fixed it 09:33:38 but its uglier now 09:33:51 static sum_list *sum_get_new_node (sum_hash *h) 09:33:52 { 09:33:52 sum_list **l; 09:33:52 if (h->usednodes == NODE_BLOCK_SIZE) 09:33:52 { 09:33:52 safe_realloc (&(h->nodes), sizeof(sum_list*)*(h->nodeslen+1)); 09:33:54 h->usednodes = 1; 09:33:56 l = &(h->nodes[h->nodeslen++]); 09:33:58 *l = malloc (sizeof(sum_list)*NODE_BLOCK_SIZE); 09:34:00 return *l; 09:34:02 } 09:34:04 l = &(h->nodes[h->nodeslen-1]); 09:34:06 return &((*l)[h->usednodes++]); 09:34:08 } 09:34:51 SHIT, it works! 09:34:57 # Mailbox summary file 09:34:57 ^@^@^P^Y^@^@^@^@:Ký¤^èM(Où~W3^@^@^@^E^@^@^@^E^@^@^@^@Aaron Lehmann^@aaronl@vitelus.com^@Re: [Shifty-Eyed-L] mutt setting that may help^@shifty-eyed-l@bad-people-of-the-future.san-francisco.ca.us^@DON'T DELETE THIS MESSAGE -- FOLDER INTERNAL DATA^@Mail System Internal Data^@MAILER-DAEMON@vitelus.com^@New AbiWord release fixes all of these...^@control@bugs.debian.org^@Bigloo^@craig@red-bean.com^@ 09:39:05 what is this function? 09:39:23 null-terminated strings SUCK 09:39:37 (so do &entity; encoded strings) 09:39:58 ? 09:40:20 sum_list 09:40:31 it's insanity 09:41:03 ? 09:41:59 ok, just one last question, as I won't have time to help now... this function is part of what program? 09:42:21 a secret one 09:42:25 :) 09:42:46 bye 09:42:54 --- quit: edrx ([x]chat) 09:52:25 fare i think asciiz is ok 09:52:33 no different from a counted string 09:54:37 not 8-bit clean 09:54:45 ? 09:54:48 cannot be used as a generic data contender 09:55:01 who wantes anything generic 09:55:08 requires several redundant specific datatypes 09:55:11 jack of all trade code is bad 09:55:18 no it ain't 09:55:23 it sure is... 09:55:34 any piece of code should be specific to the task at hand 09:55:49 nmo magic do everything in on application 09:56:00 heh 09:56:10 nmo ? 09:56:23 no 09:56:31 typing in the cracks again 10:24:42 I440r: on x86, is a long word really faster than a word? 10:29:22 erm 10:29:28 its to do with alignment 10:29:30 PLUS 10:29:34 what if you are moving data 10:29:43 pick up one item 10:29:45 put it over there 10:29:49 pick up next item 10:29:51 put it over there 10:30:03 if items = bytes there are more items 10:30:11 if items are words there are half as many 10:30:20 if they are long words there are a quater as many 10:40:59 "summary.c" 901 lines, 30188 characters written 10:41:09 anyway, i have a loop counter 10:41:20 it will never be more than 65000 10:41:25 should i make it a word or a long word? 10:44:07 no need 10:44:13 just make it an int 10:44:18 it doesnt matter 10:44:28 int on your box might be 29384765289365492435 bits 10:44:33 it might be 16 10:44:38 tho its prolly 32 :P 10:44:51 ok 10:44:53 u dont realy care in this case heh 10:45:07 bongo wake up :P 10:45:07 hehe 11:33:26 --- topic: set to 'work resumed on isforth project - tcn added as developer (bongo wake up :)' by I440r 11:53:52 --- join: edrx (edrx@200.240.18.101) joined #forth 12:02:48 --- quit: edrx ([x]chat) 12:42:01 --- join: adu (andrew@adsl-63-201-88-194.dsl.snfc21.pacbell.net) joined #forth 12:43:00 --- mode: I440r set mode: +o adu 12:45:11 hey guys 12:48:34 :) 12:49:13 could you help me? 12:49:20 i don't know x86 that well.... 12:49:40 does add and sub use the values in ebx&eax or are they addresses? 12:50:25 eitehr 12:50:29 depends on how you state it 12:50:43 sub eax,ebx subrtracts contents of ebx from the contents of eax reg 12:50:49 like add eax 12:50:54 sub eax,[ebx] 12:51:11 [*] is address? 12:51:11 subs the contents of the address pointed to by ebx from eax 12:51:16 yes 12:51:18 ahh 12:51:18 ok 12:51:21 thanx 12:51:24 that:) 12:51:45 that isnt always the case tho, not everyone codes in that way/.... 12:52:01 o 12:52:02 in my sources it is always obvious when i am talking about a POINTER to memroy 12:52:38 and is "code 1,'+',plus" a gas directive? 12:52:44 yes 12:52:49 ok 12:52:50 thats a macro 12:52:52 not for gas 12:53:00 --- join: SoapForge (flop@203-96-105-22.dialup.xtra.co.nz) joined #forth 12:53:00 this code will not assemble with gas 12:53:00 o its in macro.1? 12:53:00 at all 12:53:03 it requires nasm 12:53:05 yes 12:53:10 code and colon are macros 12:53:17 they create forth headers for the word 12:53:20 nasm? is that available on ppc? i only have gas 12:53:27 erm nto sure 12:53:32 it prolly is 12:53:38 because nasm is written in c 12:53:43 o ok 12:53:43 its even available for dos 12:53:45 HOWEVER 12:53:53 you cant use the normal version of nasm 12:53:55 hang on 12:53:58 let me get you a url 12:54:07 why can't i use gas? 12:54:29 http://home.snafu.de/phpr/e-lhputils.html 12:54:36 because these sources are in nasm syntax 12:54:40 gas is in evil at+t syntax 12:54:59 gas is the most fucked up assembler ever written 12:55:12 it wasnt intended to ass3emble ANYTHING except the output of the c compiler 12:56:57 i'm getting the nasm_0.98.orig.tar.gz, is that cool? 12:56:57 --- quit: clog (Ping timeout) 12:56:57 --- log: stopped forth/01.01.19 12:57:06 --- log: started forth/01.01.19 12:57:06 --- join: clog (nef@bespin.org) joined #forth 12:57:06 --- topic: 'work resumed on isforth project - tcn added as developer (bongo wake up :)' 12:57:06 --- topic: set by I440r on [Fri Jan 19 11:33:18 2001] 12:57:06 --- names: list (clog SoapForge @adu @Fare @aaronl @speuler @I440r) 12:57:15 out put from c? 12:57:18 erm 12:57:18 hmm 12:57:23 it needs to be 98e 12:57:31 uh 12:57:44 the e is for extended... this is an unofficial nasm that has some bug fixes and extra stuff 12:57:46 heeh 12:57:57 in nasm if you do 12:58:01 %define x 1 12:58:05 %define y x 12:58:11 %define x 2 12:58:15 both x and y change values 12:58:25 which is fucking STUPID 12:58:25 ok nasm-0.98e-src.zip, got it 12:58:30 yes 12:58:44 nasm 98e fixes this :) 12:58:53 o 12:58:54 cool 13:00:46 my macros require the ability to save the value of a %define away, redefine it then restore it later 13:01:03 but in the official nasm when you 13:01:18 save %define to somewher else 13:01:26 set %define to new vale 13:01:32 the value saved to 'somewhere else' ALSO changes 13:01:38 fucked up 13:01:43 TOTALLY fucked up 13:07:05 yes ic 13:07:56 :) 13:08:04 x86 isnt that difficult realy heh 13:08:20 tho i have to admit intel processors do tend to have fscked up asm :) 13:08:28 motorola is much easier :) 13:08:58 i think there is a faster way to do abs on ppc 13:09:06 hrm 13:09:07 theres a conditional sub i think 13:09:08 i dunno 13:09:10 dont do ppc asm 13:09:14 i do 13:09:15 hehehe 13:09:16 it gets the absolute value 13:09:40 ya i know 13:10:04 but doesn't subtracting it from zero do the same thing? 13:10:08 or does that take longer? 13:10:17 your trying to optimize right? 13:11:03 hehe 13:11:10 abs does this 13:11:18 sign extend it into edx 13:11:21 i'm looking at it 13:11:32 then subtrat it FROM edx ?? 13:11:33 i forget 13:11:38 its a little obfuscated hehe 13:11:45 mov eax,ebx ;set ebx = sign of n1 13:11:45 sar eax,byte 32 ;propogate sign bit if n1 throughout ebx 13:11:45 xor ebx,eax ;will invert n1 if n1 was negative 13:11:45 sub ebx,eax ;will subtract -1 if n1 was negative 13:11:49 i didnt write it.. but i believe its the most efficient way 13:11:57 thtats it 13:12:03 it just makes the sign bit on 13:12:07 then inverts it 13:12:24 waht ever the sign bit is in eax 13:12:30 it set ALL bits in edx to that 13:12:45 then xor's eax with edx 13:12:53 if the sign was 0 then eax is unchanged 13:13:03 i think i have a Pentium.PDF somewhere..... 13:13:05 if the sign was negative then eax is logically inverted 13:13:09 i.e. made posative 13:13:24 and what happens if you subtract a POSATIVE number from 0 ? 13:13:46 we are not trying to invert the number 13:13:50 we are saying 13:13:55 if (number is negative) 13:13:58 invert it 13:14:04 not (invert nubmer) 13:14:08 thers a diff :) 13:14:22 l440r: you know what two's compliment is don't you? 13:14:27 yes 13:14:32 k 13:14:49 xor with -1 13:14:51 then add 1 13:15:44 * aaronl loves to xor with 1 13:16:06 xor 1 just add 1 or subtracts 1 13:16:07 thats all 13:16:14 xor with -1 13:18:15 l440r: i like to think of it as just enought to make it overflow 13:18:40 like -0x0040 is 0xFF80 13:18:49 or FFC) 13:18:54 wait 13:19:07 its 0xFFC0 13:19:56 if you add the two together then you get 0x10000 13:20:40 ain't dat cool? 13:21:36 :) 13:22:48 i love computers 13:26:00 hey did i ever tell you what i was thinking of calling my distro? 13:26:08 waht ? 13:26:12 BNLSD 13:26:17 ? 13:26:30 Bare Naked Linux Sodtware Distro 13:26:42 hehe 13:26:45 like BNL 13:26:47 hehe 13:26:48 and LSD 13:27:01 bad :P 13:27:06 hehe 13:28:12 my docs say that sar is a signed divide... 13:29:12 * adu loves to xor with aaronl 13:29:49 sar IS 13:30:01 by doing 32 sar's in a row 13:30:13 you propogate the sign throughout the register 13:30:23 sar eax,32 13:30:27 or edx :P 13:30:29 o 13:31:14 so edx in the code aboce will either be ZERO or minus 1 13:31:42 xor of a posative number with ZERO leaves it unchaged 13:31:45 this is probably already done in mathlibs somewhere 13:32:01 xor of negative number with MINUS 1 makes it posative 13:32:11 but i WONT use math libs 13:32:15 i wont use ANY libs 13:32:18