00:00:00 --- log: started forth/03.12.12 00:00:08 you wouldnt have the "data" and "code" in the same page 00:00:36 but i could care less about that heh 00:00:44 It's cache lines that you need to worry about, not pages. :) 00:00:56 so i get processor cache flusses all the time :P 00:01:09 but its based on page size 00:01:28 Not as far as I'm aware. Cache line size for AMD Athlon is 64 bytes, and 32 bytes for Intel CPUs. 00:01:28 if you have data and code in the same memory page the processor has to flush and reload that page 00:01:32 load that page to execute code. 00:01:36 Nope. 00:01:40 flush it and reload it as data to read the data 00:01:54 Only the cache line, not the whole page. 00:01:59 A page is a 4KB chunk of RAM. 00:02:04 yes 00:02:37 i was under the impression that there was only ONE cache and if it has code in it it needs to be reloaded with the data 00:02:45 even if its the same physical memory 00:02:54 Nope. 00:03:01 then ive been lied too lol 00:03:06 There are cache coherency issues to be aware of, sure. 00:03:17 But all modern CPUs are Harvard architecture internally; there are separate data and code caches. 00:03:20 so me 00:03:30 myvariable: 00:03:37 call dovariable 00:03:40 dd data 00:03:49 isnt that bad after all hehe 00:04:14 if call dovariable is in the tail portion of one cache line, and dd is in the head portion of another line, that is true. 00:04:28 Otherwise, you will get a conflict **IF** data is known to have been modified. 00:04:30 define "line" 00:04:45 I just said above: a cache line is a chunk of 32 bytes for Intel CPUs, and 64 bytes for Athlon CPUs. 00:05:02 oh ok 00:05:30 Goodness, I'm not at all sure how I'm going to implement split code/data space though. 00:05:39 i could waste space and align my variables and constants to be more speed efficient 00:05:42 but .. blah 00:06:05 Remember, the conflict occurs only when *modifying* data. Otherwise, code-resident look-up tables would be useless. 00:06:27 kc5 i thunked about it a great deal and came to the conclusion that the overhead in code desing complexityh made it a waste of time 00:06:42 right 00:06:55 I440r: Well, I'm going to think about it anyway. 00:07:04 * kc5tja doesn't want to have to hack around with mprotect() if I don't have to. 00:07:04 an instruction fetch is just a data read 00:07:08 yes 00:07:14 im not discouraging you 00:07:20 you might find a way i missed :) 00:07:30 and if i like it i might steal it heh 00:07:33 ;) 00:08:23 or maybe we can come up with something togehter - tho my mind isnt in gear atm 00:08:35 i think im revving up but im not in gear yet 00:09:26 kc5 if you do it you will be more in-line with how linux expects things to be 00:09:37 im breaking ALL kinds of conventions here heh 00:10:06 neway i gtg zzz 00:10:26 Yeah, me too. 00:10:32 I have to work later today. :D 00:10:38 me 2 heh 00:10:47 I know that there are split-space Forths out there -- that's how ROM-based systems work. 00:10:48 tgif 00:10:52 Not for me. 00:10:57 Friday is the start of my workweek for me. 00:11:02 yes - its harvard architecture 00:11:20 are you direct threaded or indirect threaded ? 00:11:25 Neither 00:11:29 Native code compiling. 00:11:30 sub ? 00:11:34 aha. 00:11:43 Yeah, subroutine threaded, with primitives inlined. 00:11:59 optimizing ? 00:12:03 or "not yet" 00:12:09 Only basic peep-hole optimizations. 00:12:13 how do they work ? 00:12:20 what is "peep hole" 00:12:33 They examine the previously generated opcodes, and determines future actions based on previously emitted code. 00:12:49 ok. doesnt tell me much but then its 2 am heh 00:12:57 Peephole optimizers work by examining a small fragment of the code at a time (hence the name -- looking at the whole source through a peephole). 00:13:07 Trust me, you wouldn't like it. 00:13:36 lol im not totally averse to optimizations as an experiment. im just opposed to them in production code. 00:13:48 This is production code. 00:13:50 i dislike compilers doing anyting behind your back 00:13:58 YOUR production code :) 00:14:04 i will some day implement an optimizer 00:14:07 for my USERS 00:14:13 For example, when you write 5 + @, the raw code that would normally be produced is htis: 00:14:15 i wont restrict what they do. ill give them the option 00:14:23 sub esi,4 00:14:25 mov [esi],eax 00:14:27 mov eax,5 00:14:35 add eax,[esi] 00:14:38 add esi,4 00:14:42 mov eax,[eax] 00:14:52 However, the output of my compiler will produce this: 00:14:55 mov eax,[eax+5] 00:15:29 (well, it'll produce add eax,5 // mov eax,[eax] to start with, but you get the idea) 00:15:40 yes 00:16:04 neway zzzz... i went thataway --> 00:16:05 :) 00:16:08 night 00:16:09 night 00:16:18 --- quit: kc5tja ("THX QSO ES 73 DE KC5TJA/6 CL ES QRT AR SK") 00:45:31 * warp0x00 is away: run, forest, run! 01:05:42 --- join: schihei (~schihei@blueice1x.de.ibm.com) joined #forth 01:42:11 Hrm. 02:56:38 --- join: karingo (karingo@160.portland-04-05rs.or.dial-access.att.net) joined #forth 03:17:48 --- quit: fridge ("Leaving") 05:12:49 --- join: onetom_ (~tom@cab.bio.u-szeged.hu) joined #forth 05:12:49 --- quit: onetom (Read error: 104 (Connection reset by peer)) 06:09:21 --- join: tathi (~josh@pcp02123722pcs.milfrd01.pa.comcast.net) joined #forth 06:11:37 --- join: aktnot (ident@181.80-202-66.nextgentel.com) joined #forth 06:17:29 --- quit: aktnot ("leaving") 06:17:35 --- join: aktnot (ident@181.80-202-66.nextgentel.com) joined #forth 06:18:25 --- quit: I440r ("work...") 06:18:40 --- quit: aktnot (Client Quit) 06:18:48 --- join: aktnot (ident@181.80-202-66.nextgentel.com) joined #forth 06:57:38 --- join: I440r (~mark4@saturn.vcsd.com) joined #forth 06:59:55 y0 07:00:19 ! 07:00:22 working ... 07:00:25 sorta 07:00:30 :) 07:00:43 this early in the morning? :) 07:09:30 * arke is away: school 07:12:35 --- quit: Robert (Read error: 60 (Operation timed out)) 07:20:35 --- join: Robert (~snofs@c-305a71d5.17-1-64736c10.cust.bredbandsbolaget.se) joined #forth 07:25:18 How do I show what's on the stack in gforth? 07:38:00 I figured it out myself. I have to type .s 07:38:35 yes dot-s is the way 07:41:20 aktnot what platform/os do yo uuse ? 07:49:43 linux/gforth 07:52:06 I prefer bsd, but it's a driver issue, so I use linux on the workstation. 07:55:52 --- join: kc5tja (~kc5tja@66-91-231-74.san.rr.com) joined #forth 07:55:53 --- mode: ChanServ set +o kc5tja 07:57:10 --- topic: set to 'A channel dedicated to the Forth programming language, its implementation, and its application.' by kc5tja 07:57:32 --- topic: set to 'A channel dedicated to the Forth programming language, its implementation, its application, and its philosophy.' by kc5tja 07:57:32 Oh, so that's what we were talking about. 07:57:41 Not about gravity? 07:57:48 I wanted to talk about CVTs today :( 07:57:58 So talk about CVTs or gravity. 07:58:07 But the topic needed to be changed. 07:58:39 * kc5tja should be getting ready for work, but someone decided to start the dishwasher. >:( 07:58:40 awww :( 07:58:46 If you "drop" an infinitly large object, it will drag the earth towards it very quickly, right? Much faster than a marble would fall. 07:58:50 * kc5tja is in a pissy mood today. 07:58:54 Robert: yes! 07:59:08 That's the point I was making this morning... 07:59:15 * kc5tja sighs 07:59:26 --- quit: aktnot (Remote closed the connection) 07:59:28 Folks, take an elementary physics class, please. 07:59:28 What did I do now? :( 07:59:39 kc5tja: so what is gravity? 07:59:39 --- join: aktnot (ident@181.80-202-66.nextgentel.com) joined #forth 07:59:55 I refuse to discuss this topic any further. The documentation is out there, it doesn't take a rocket scientist to see and observe nature. 08:00:01 --- quit: aktnot (Client Quit) 08:00:02 MysticOne: A warp in space-time. 08:00:08 --- join: aktnot (ident@181.80-202-66.nextgentel.com) joined #forth 08:00:24 kc5tja: did you ever get a chance to watch that elegant universe thing? 08:00:30 No. 08:00:36 I've been too pre-occupied. 08:00:39 it was really cool 08:00:57 they talked about all sorts of neat things 08:01:04 though it kinda went over my head once we got into 11 dimensions 08:02:47 I'm sure. 08:02:56 As it is, I have not yet seen it. 08:02:59 if ever. 08:06:09 * kc5tja is upset with two of my roommates right now. 08:06:22 The first for having started the damn dishwasher right when I was going to take a shower for work. 08:06:40 The second because he thinks I should `calm down,' for getting mildly upset with the first. 08:06:55 What, am I no longer allowed to express feelings of dissatisfaction around here? 08:07:36 how long had the dishwasher been running when you went to get a shower? 08:07:51 I have no idea. 08:07:58 Probably about 5 to 10 minutes. 08:07:58 was all the hot water gone? 08:08:13 No, but when it does use hot water, the shower goes *COLD*. 08:08:27 turn off the dishwasher? 08:08:33 I can wait. 08:08:41 I have two hours left to get ready. 08:08:45 oh 08:08:49 But still, he *knew* I get up this time every morning. 08:09:03 He could have just *asked* me, "Hey, after you get ready for work, can you run the dishwasher?" 08:09:12 * MysticOne nods 08:09:29 I would have been more than happy to oblige. 08:09:44 I really would have been much happier if he would just wash his dishes right after using them. 08:09:47 But I digress. 08:12:11 * kc5tja notes that, because of folks not washing dishes after they're done using them, I'm now without a plastic bowl and a fork. Nobody seems to know where they are. 08:12:42 we just wash dishes whenever we have enough to fill the dishwasher 08:13:15 See, I used to do that, and the problem is is it still attracts bugs, it still causes the house to stink up, et. al. AND, it still takes up space. 08:13:29 * kc5tja would rather just get the dishes done right when they're freshly soiled. It's so very much easier. 08:14:43 Anyway, like I said, I'm in a pissy mood at the moment. 08:22:06 * kc5tja sighs -- building a valid ELF header is a pain in the ass. 08:22:32 * kc5tja has the normal ELF header itself, then I need an array of program headers, then an array of section headers, the string table.... 08:24:49 * I440r was thunkin of including debug info on request :) 08:25:07 not easy 08:25:29 I have no choice but to make words that help automate this process. 08:28:22 Also, a single program segment size is restricted to just 64KB? 08:29:11 Oops. I am generating incorrect headers. 08:29:22 _Word fields are emitted as two bytes instead of four, like they should be. 08:29:24 Easy enough to fix. 08:36:08 --- quit: schihei (Client Quit) 09:26:53 --- quit: karingo (""ha"") 09:27:42 --- join: Herkamire (~jason@h000094d30ba2.ne.client2.attbi.com) joined #forth 09:40:53 --- quit: I440r ("Leaving") 09:42:58 --- quit: chandler (Read error: 104 (Connection reset by peer)) 09:56:10 Well, I got the ELF file to include a program header. I need to include information on sections now. 09:56:18 --- join: I440r (~mark4@saturn.vcsd.com) joined #forth 09:56:24 And that means I need a string table, which itself is a section. 09:56:35 I440r: I got a.out to successfully segfault. :) 09:56:48 anyone in here know how to do mysql joins? 09:56:49 I440r: I have it building program headers, but no sections. 09:56:57 MysticOne: Nope. :( 09:57:16 MysticOne: More accurately, not I. :/ jc might, but he seems to be afk. 09:57:30 reading some documentation on how to do it 09:57:33 not sure if that's what I want or not 09:57:45 I have two tables, and I need data from one to perform a query on the other, so trying to simplfy it to one query 09:58:30 * kc5tja nods 09:58:43 * kc5tja is making slow progress in the generation of an ELF executable via FS/Forth's target compiler. 09:58:53 My first implementation, when I'm done, should work, but it'll be ugly. 10:01:55 * kc5tja needs to add data alignment words to the program though. Right now, I'm doing everything on byte alignments. 10:02:24 lol 10:03:47 Which means it'll waste some space in the target file. Oh well. 10:03:51 But for now, ... 10:05:35 Anyway, I have to get to work. 10:05:42 --- quit: kc5tja ("THX QSO ES 73 DE KC5TJA/6 CL ES QRT AR SK") 10:24:37 The mysql.com werbsite documentation area has some good stuff on inner, outer, left and right joins. I don't use those often enough to be able to give an example without just cuttng and pasting from their site. 10:33:57 --- quit: tathi ("leaving") 12:07:30 --- quit: aktnot ("leaving") 13:26:00 --- join: chandler (~darmok@64-145-60-36.client.dsl.net) joined #forth 14:27:38 --- quit: I440r ("Leaving") 15:28:29 * arke is back (gone 08:19:03) 15:38:13 y0 15:39:48 ok 15:39:48 y0 y0 ? 15:45:12 y0 dup 15:58:17 --- join: Robert_ (~snofs@c-305a71d5.17-1-64736c10.cust.bredbandsbolaget.se) joined #forth 16:09:56 --- nick: wUoNrFk -> wUoNrFk-stay-cod 16:10:28 --- join: I440r (~mark4@12-160.lctv-a5.cablelynx.com) joined #forth 16:10:46 --- quit: Robert (Read error: 110 (Connection timed out)) 16:21:02 Herkamire: theere? 16:35:08 Poll: Best MAME game? 16:46:05 hm 16:46:15 rather broad question. 16:46:34 * wUoNrFk-stay-cod can't come to a conclusion on that one 16:47:42 OK... top 5 favorites then? 16:51:20 ... i dunno.... it's stilll.. very dependent on mood 16:53:44 I hear you. Still... without having to pick favorites then, what do you often play? 16:57:06 --- join: cara (~cara@cs6625132-244.austin.rr.com) joined #forth 16:58:20 hey Robert_ 16:58:33 hehe he doesn't remember me 16:58:35 anyway I'm off 16:58:38 --- part: cara left #forth 17:42:33 --- join: kc5tja (~kc5tja@66-91-231-74.san.rr.com) joined #forth 17:42:33 --- mode: ChanServ set +o kc5tja 17:59:40 --- nick: wUoNrFk-stay-cod -> wUoNrFk 18:06:05 --- join: TheBlueWizard (TheBlueWiz@pc5bdn1d.ppp.FCC.NET) joined #forth 18:06:05 --- mode: ChanServ set +o TheBlueWizard 18:06:10 hiya all 18:22:13 arke: OK, I'm back. what's up? 18:23:13 hiya Herkamire....just chillin' :) 18:23:19 hi TheBlueWizard :) 18:24:01 :) 18:48:08 Howdy 18:48:25 * kc5tja will be back though; I'm off to get myself some food. 18:48:39 hiya kc5tja :) 18:49:04 * kc5tja is hopefully going to get his power steering pump finally fixed in the not too distant future. 18:49:15 That leaves only the transmission as its only seriously critical issue. 18:49:27 talking about your car? 18:49:45 Jeez, I maligned that sentence pretty badly. 18:49:47 Yes 18:50:49 mmhmm 18:53:42 * kc5tja will be back. Going to Carl's Jr. to grab myself something quick to eat. 18:53:44 * kc5tja is away: food 18:56:43 --- join: fridge (~fridge@dsl-203-33-164-75.NSW.netspace.net.au) joined #forth 19:00:52 hiya fridge 19:02:48 --- quit: madgarden (Read error: 104 (Connection reset by peer)) 19:02:51 * fridge tips his hat 19:05:26 :) 19:13:34 * kc5tja is back (gone 00:19:50) 19:15:35 * kc5tja thinks he's going to invest in a little space heater. 19:17:12 that chilly, eh? 19:17:24 Yes 19:17:56 It doesn't help that we have only one thermostat in the whole house, five rooms, and each room has its own unique way of leaking heat to the outside. 19:18:39 mmhmm...you live in northern Calif., right? 19:19:08 Only in mirror-world. :D San Diego is right on the Mexican border. 19:20:51 I thought you live near San Francisco or something like that 19:20:57 nope 19:21:01 I don't have a heater, it forces me to get intimate with my wife in the colder months ;) 19:21:04 * TheBlueWizard knows where San Diego is 19:21:22 I thought San Diego tends to be warm even in winter time 19:21:46 TheBlueWizard: Compared to NY state, it's swimming weather. :D 19:21:50 I440r, you around? 19:22:02 However, we just got done with 4 days or continuous rain, so it's pretty chilly and windy here. 19:23:10 I see 19:32:35 Herkamire: whats rj's email addy? do you kknow? 19:32:40 y0 all 19:32:45 and 19:33:04 * arke is away: f0000000000d 19:35:24 Who is rj? 19:35:29 * TheBlueWizard takes out a few 0's from arke's f00000000d...being fat ain't healthy, y'know... :) 19:35:59 Nope. 19:36:13 And the funny thing is, I'm clinically obese, yet, I'm thin. :) 19:36:31 huh???? 19:36:36 * kc5tja weighs about 220lbs; I should weigh 175 for my height. 19:37:03 * TheBlueWizard weighs only 135 lbs...should've weigh more 19:37:22 But I'm not fat. Even the nurses at the doctor's office were rather shocked. :) 19:38:55 * kc5tja gives TheBlueWizard two shots of Tequilla. :D 19:39:01 hmm...muscles are dense matter, so if you're muscular, you can weigh more 19:39:13 * kc5tja isn't very muscular. 19:39:24 And I do have excessive body-fat. 19:39:39 What I'm saying is that I look like every other normal, non-obese person on the planet. 19:39:40 hmm... 19:39:43 "I'm not fat! I'm big boned!" 19:39:47 Heh 19:39:53 :) 19:40:08 Though the muscles in my legs are getting noticably bigger from biking around town a lot. 19:40:57 So I wouldn't be surprised if I actually gained weight since the last medical weighing. :D 19:41:32 If you have a counted string, is it safe to say "mystring 1+", or is it best to say "mystring count drop" to skip the length byte? 19:42:01 Or is there a better way, in case 16 or 32 bit counts are supported? 19:42:07 To be maximally compatible, count drop is the preferred solution, because it permits your code to work with systems that use 16-bit counts (or more). 19:43:28 And it remains efficient, since count doesn't really count, just gets the length byte and returns the address. 19:44:06 : count dup (n) + swap (c)@ ; 19:44:28 where (n) is the size of the length field, and (c) is optional. :) 19:44:58 K. Just confirming. 19:45:03 So, for an 8-bit counted string system, : count dup 1+ swap c@ ; 19:59:44 * MysticOne smacks kc5tja with a hot water eating up dishwasher 20:05:26 * kc5tja smacks MysticOne back with a dishwasher eating up hot water. 20:05:41 hehehe 20:06:12 you both living in same house? 20:06:24 Nope. 20:07:26 MysticOne lives in the panhandle area of Florida. 20:07:48 we're practically next door though 20:07:51 just like 3k miles apart 20:08:18 yeah...on the same latitude ;) 20:09:11 :) 20:25:17 Heheh :D 20:33:54 * kc5tja is responding to a post on the ColorForth mailing list. 20:34:03 Some good comparisons going on between classical and Color/MachineForth. 20:35:50 how long have you been forthing kc5tja? 20:36:03 too long 20:36:05 he's old and bitter 20:36:06 you seem extremely knowledgable on the subject 20:36:08 * MysticOne hides 20:37:43 MysticOne: I AM NOT!! 20:37:48 stupid newbie... ;D 20:37:51 :) 20:38:03 fridge: I've been using Forth for at least 6 years. 20:38:16 Much of my knowledge has been gained from writing my own Forth systems. 20:38:30 kc5tja: old and bitter! 20:38:46 You not only learn the language that way, you also learn implementation details, and an appreciation of how certain designs affect performance and size tradeoffs. 20:39:07 MysticOne: I'm only 30 years old. though I do have arthritis in one of my joints. 20:39:16 hehehe 20:39:37 hopefully not your knees if you're cycling =\ 20:39:42 Nope. 20:39:54 My right-hand, middle finger, knuckle closest to the finger tip. 20:40:08 amputate! 20:40:09 Fortunately, it won't affect my aikido either. 20:40:13 Heh, no. 20:40:14 :) 20:42:52 At any rate, I'm thinking of coding my Dolphin OS's kernel in FS/Forth. 20:43:24 Among several other projects. 20:43:47 Open source folks will totally hate me for choosing Forth though (especially one as custom implemented as mine. :) ). 20:44:02 but you're old and bitter, so you don't care, right? 20:44:26 Right. :) 20:44:35 :) 20:44:37 I'll put it right on the front page of the developer documentation: 20:44:44 http://www.stud.ntnu.no/~shane/stasj/pics/humor/div/248.html 20:45:52 "Dolphin is written in a dialect of MachineForth. DEAL WITH IT. NO we do not intend on switching to ANSI Forth in the future. NO we do not intend on switching to C (though, performance-critical routines may be coded in C). NO we do not intend on changing to Lisp/Scheme. If you do not agree with this, go away -- you'll only hinder the progress of those who are actively coding and supporting the kernel." 20:46:27 kc5tja: I think "ABANDON ALL HOPE YE WHO ENTER HERE" is much more succinct 20:47:23 fridge: "Beyond this point, there be dragons!" 20:48:39 Gosh, I really need to finish FS/Forth. 20:48:47 Get it producing working ELF executables under Linux. 20:48:58 I want to compare its benchmarks against gforth, bigForth, and gcc. 20:49:16 * kc5tja knows it won't be the fastest kid on the block, but I predict it'll pretty universally be between gforth and bigforth. 20:53:20 jc sorry im arround now :) 20:53:25 whussup ? 20:55:14 hiya I440r!!! :) 20:55:27 tbw lol i didnt even notice ya heh 20:55:39 kc get that program header in 20:55:46 I have a program header. 20:56:03 But it now crashes when I execute the program, because it's not allocating space for the code segment. 20:56:03 aha and its still segfaulting ?> 20:56:06 Yes. 20:56:20 ok show me the elf and program headers 20:56:27 ummm.... 20:56:28 :) 20:56:32 Hexdump OK? 20:56:36 yea 20:56:44 I build the program headers programmatically, because they're too damn hard to do it by hand. 20:57:04 program headers PLURAL? 20:57:25 401B63A8: 7F 45 4C 46 01 01 01 00 - 00 00 00 00 00 00 00 00 .ELF............ 20:57:25 401B63B8: 02 00 03 00 01 00 00 00 - 00 00 01 00 4C 00 00 00 ............L... 20:57:25 401B63C8: 00 00 00 00 00 00 00 00 - 34 00 20 00 01 00 00 00 ........4. ..... 20:57:25 401B63D8: 00 00 00 00 - .... 20:57:28 No, singular. 20:57:32 That's the ELF header. 20:57:39 * kc5tja needs to find the address for the program header currently. 20:57:57 the prpgram header should go immediatly after the section header 20:58:05 between it and the text section 20:58:09 401B63F4: 01 00 00 00 34 00 00 00 - 00 00 01 00 00 00 01 00 ....4........... 20:58:09 401B6404: 18 00 00 00 00 40 00 00 - 07 00 00 00 01 00 00 00 .....@.......... 20:58:43 Well, I just got done saying, (a) it crashes because it's not allocating memory for the text section, and (b) I don't have any section headers yet, thus it's not allocating memory for the sections. :) 20:58:51 your p_offset is $34 - mine is $00 20:59:23 objdump -x reports a proper ELF header, and gives proper values for the various fields, so I know mine is correctly formed. 20:59:38 your file size is different from your mem size too 20:59:39 thats wrong 20:59:40 I am placing all the headers (except ELF) at the end of the file, because it's easier to do that way. 20:59:47 No, it's perfectly valid. 21:00:04 large paste coming up 21:00:08 create elf_header 21:00:08 $7f c, $45 c, $4c c, $46 c, $01 c, $01 c, $01 c, $03 c, 21:00:08 $00 c, $00 c, $00 c, $00 c, $00 c, $00 c, $00 c, $00 c, 21:00:08 $02 w, \ e_type executable 21:00:13 $03 w, \ e_machine intel 80386 21:00:15 1 , \ e_version 21:00:17 e_entry , \ e_entry 21:00:18 As long as file_size <= mem_size, it (should, at least) work (according to the spec) 21:00:19 $34 , \ e_phoff offset to program hdr 21:00:23 here $00 , \ e_shoff section header offset 21:00:27 $00 , \ e_flags 21:00:29 $34 w, \ e_ehsize 21:00:31 $20 w, \ e_phentsize 21:00:33 $01 w, \ e_phnum 21:00:35 $28 w, \ e_shentsize size of a section hdr! 21:00:37 $03 w, \ e_shnum # section hdr entries 21:00:39 $02 w, \ e_shstrndx 21:00:47 \ program header starts here 21:00:51 1 , \ p_type 21:00:53 0 , \ p_offset 21:00:57 $8048000 , \ p_vaddr 21:00:59 $8048000 , \ p_paddr 21:01:01 here -1 , \ p_filesz 21:01:03 -1 , \ p_memsz 21:01:05 7 , \ p_flags 21:01:07 $1000 , \ p_align 21:01:15 constant p_filesz 21:01:17 constant e_shoff 21:01:19 21:01:33 You're setting p_filesz and p_memsz to $FFFFFFFF? 21:01:58 And you do realize that the p_align field you have is 2**65536, which is an insanely huge number? :) 21:02:19 your e_entry is $0100000 21:02:26 I'm not talking about e_entry. 21:02:28 i do not think linux will like that 21:02:33 I'm positive it does. 21:02:43 * kc5tja did a hexdump of /bin/ls 21:02:52 no i patch those two later 21:03:08 it depends on how large my code + header space is 21:03:33 which is why i have a constant p_filesz 21:03:47 e_shoiff is also patched later 21:04:15 i relocate headers to "here" then where headers end is where e_shoff is set to 21:04:16 * kc5tja nods -- My code builds the headers differently 21:04:22 then i put my section headers in there 21:04:36 what do you have in your string table 21:04:47 I don't have a string table yet. 21:04:48 you need .text and /.shstrtab 21:04:56 * kc5tja is going to work on that next, because the section headers are next. 21:04:57 create string_table 21:04:57 0 c, \ because some idiots dont understand 0 based 21:04:57 ,' .text' 0 c, \ or they assume we dont 21:04:57 ,' .shstrtab' 0 c, 21:05:15 i just copy that to where it belongs and set the elf headers pointer to it 21:05:28 create text_header 21:05:28 1 , \ sh_name offset in $ table to name 21:05:28 1 , \ sh_type progbits 21:05:28 7 , \ sh_flags read write executable (dejavu) 21:05:28 e_entry , \ sh_addr where this section lives 21:05:29 $80 , \ sh_offset file offset to start of section 21:05:31 here 0 , \ sh_size how big is the section (dejavu) 21:05:33 0 , \ sh_link 21:05:35 Actually, the reason for the first 0 is to ensure that a 0 index refers to a proper (null) string. 21:05:35 0 , \ sh_info 21:05:37 $10 , \ sh_addralign 21:05:39 0 , \ sh_entsize 21:05:47 constant sh_size \ point to above entry 21:06:05 i know - but its so they dont have to deal with zero based 21:06:21 they want the first real item to be the ONEth item 21:06:23 lamers 21:06:40 No, I have to disagree. 21:06:43 create st_hdr 21:06:43 7 , \ sh_name offset in $ table to name 21:06:43 3 , \ sh_type sht_strtab 21:06:43 0 , \ sh_flags 21:06:44 0 , \ sh_addr 21:06:46 here 0 , \ sh_offset 21:06:48 15 , \ sh_size 21:06:50 0 , \ sh_link 21:06:52 0 , \ sh_info 21:06:54 1 , \ sh_addralign 21:06:58 0 , \ sh_entsize 21:07:00 The string table index is an offset into the string table. The second item is not 2, it's something like 7 or 8. 21:07:02 constant st_off \ ofset from start of file to string table 21:07:04 thers the string table header 21:07:33 * kc5tja nods 21:07:42 Liek I said, I'm not quite there yet. 21:07:53 I have to implement the string section first. 21:07:57 ya 21:08:18 the spec i have specifically states that the string section is optional 21:08:29 and actually i lied, linux is quite happy with its absence 21:08:32 but gdb isnt 21:08:37