00:00:00 --- log: started retro/06.09.04 01:04:26 --- quit: crc (Read error: 54 (Connection reset by peer)) 01:33:51 --- join: crc (n=crc@pool-70-110-129-33.phil.east.verizon.net) joined #retro 01:34:16 --- mode: ChanServ set +o crc 01:39:48 --- quit: nighty (niven.freenode.net irc.freenode.net) 01:48:26 --- join: nighty (n=nighty@66-163-28-100.ip.tor.radiant.net) joined #retro 01:51:00 --- join: nighty_ (n=nighty@66-163-28-100.ip.tor.radiant.net) joined #retro 01:54:55 --- quit: nighty (Connection timed out) 04:37:25 --- join: virl (n=virl@chello062178085149.1.12.vie.surfer.at) joined #retro 06:49:03 --- join: nighty (n=nighty@CPE00119576a9c5-CM0012c90d36fc.cpe.net.cable.rogers.com) joined #retro 08:04:56 --- quit: timlarson (Read error: 104 (Connection reset by peer)) 08:08:35 --- join: timlarson (n=timlarso@user-12l325b.cable.mindspring.com) joined #retro 11:06:41 good afternoon 11:06:57 hi crc 11:09:07 http://retroforth.org/903.tar.gz 11:09:17 this is the start of what will become RetroForth 9.3 11:10:36 I will be using a modified version of the Rx Core in this, merging in some parts of the stuff Quartus has been writing 11:11:45 :) 11:12:26 awesome :) 11:16:23 very cool. 11:16:37 --- join: Quartus_ (n=Quartus_@209.167.5.1) joined #retro 11:29:43 * Raystm2 taking with Jeff Fox (forthlet) in #c4th 11:30:19 awesome 11:30:44 --- join: snoopy_1711 (i=snoopy_1@dslb-084-058-137-106.pools.arcor-ip.net) joined #retro 11:30:44 joined just now...did I miss much yet? 11:31:21 --- quit: Snoopy42 (Read error: 131 (Connection reset by peer)) 11:31:31 forthlet was talking about the new version of colorforth and how green variables have not found any room in the new system. 11:31:44 --- nick: snoopy_1711 -> Snoopy42 11:31:45 I'm waiting for the long-promised chartreuse constants. 11:31:52 hahahaha 11:44:18 crc: can you help me form a url that will pull up a page in my browser for an html file that is in my htdocs/old_site folder? 11:44:43 sure 11:44:53 I thought that http://colorforth.info/htdocs/old_site/file would work but no. 11:45:17 http://ray.rx-core.org/old_site/ 11:45:22 oh. 11:45:25 hehe 11:45:31 don't include the htdocs/ part :) 11:45:36 ah okay. 12:18:33 --- quit: Cheery ("Download Gaim: http://gaim.sourceforge.net/") 13:14:38 --- quit: ChanServ (Shutting Down) 13:16:00 --- join: ChanServ (ChanServ@services.) joined #retro 13:16:00 --- mode: irc.freenode.net set +o ChanServ 13:52:45 --- quit: nighty ("Terminated with extreme prejudice!") 15:01:33 --- quit: Quartus_ ("used jmIrc") 15:22:11 --- join: nighty (n=nighty@CPE00119576a9c5-CM0012c90d36fc.cpe.net.cable.rogers.com) joined #retro 15:36:28 --- quit: virl (Read error: 104 (Connection reset by peer)) 16:14:29 http://retroforth.org/903.tar.gz is now updated. I added in the section stuff and some of the optimizations from Quartus 16:14:56 Downloading! 16:17:07 crc, 'now' is missing. 16:17:12 now? 16:17:17 The word you added for me. 16:17:18 and also default-now? 16:17:26 I renamed them 16:17:52 Oh. The question-mark indicate a flag return to me. 16:19:07 I removed my section stuff, renamed external to expose, and renamed now? to now, and it's crashing horribly. 16:19:36 er, now to now? that is 16:20:02 is: is replaced by just plain 'is' as well 16:20:29 That's good, that's a change I'd made elsewhere myself. 16:23:11 Initially, it appears that even with those name changes, my conditional-compilation words fail. Did you change just the names, or the semantics also? 16:23:40 just the names 16:23:57 * crc looks over NOTES 16:24:40 It's my ignore word that is crashing the system. Did you change the internal workings of lookup word? or >number ? 16:24:46 word? has been replaced by (not-found) 16:24:59 word? itself is now just for displaying an error message in retro 16:25:54 Replacing word? with (not-found) also crashes. Is it simply renamed, or does (not-found) do something new? 16:26:52 just renamed 16:27:11 Then perhaps some other component differs. Revectoring those three routines is now now longer working. 16:28:27 Revectoring entry also crashes. 16:28:43 I didn't even touch entry 16:29:28 Is 'last' still 'last'? 16:29:32 yes 16:29:36 and :xt ? 16:30:22 yes 16:30:26 :: [ ' entry now? compile ] ; is entry 16:30:32 Is bombing. Is now? the same as now ? 16:30:39 it should be 16:30:54 It appears to be the problem, and if it is different it is the reason the other revectorings fail. 16:31:17 I'll take a closer look 16:34:25 : now? 1+ dup @ 4 + + ; gets me past that hurdle, but something else is bombing 16:35:00 Apparently the wordlist stuff. Ok, I need to know more about what you changed in 9.3. :) 16:38:16 Did you change FILL ? 16:38:39 maybe... 16:38:52 It fails the standard tests. 16:38:57 yes 16:39:04 Didn't used to. 16:39:06 I forgot the check for 0 16:40:29 The wordlist stuff appears to work thus far, but this: :: get-current definitions d: words set-current ; is words 16:40:37 makes WORDS crash hard, which is a complete mystery to me. 16:40:55 I can re-implement WORDS if necessary, but perhaps you've changed something? 16:41:17 * crc is about to upload a version with a fix to 'words' and 'fill' 16:41:27 Ok. 16:41:45 done 16:42:34 version found where? normal download link? 16:42:43 http://retroforth.org/903.tar.gz 16:43:01 Thanks. :) 16:43:11 That broke something else. 16:43:31 * Raystm2 waits for the settling. 16:45:23 Ok. Now it's the ENVIRONMENT? section, which is the first place I use wordlists. It didn't crash until this last binary you just posted. 16:45:32 Have you changed the way the dictionary is implemented? 16:46:00 nope 16:46:14 Nothing in lookup? 16:46:18 no 16:47:06 Just the process of changing the value in last is crashing subsequent lookups. 16:47:30 Maybe it's wordlist itself. I'll check. 16:48:26 Ok. Is 'expose' my 'export' code, or something different? 16:49:14 they are the same 16:49:26 They behave differently. Let me test further. 16:51:33 Hmm, perhaps it's not expose. It looks like if I am creating words in a wordlist that isn't in the search-order, I get a crash. Is there some part of the kernel that relies on dictionary searching, and depends on it being one single wordlist? 16:51:54 In other words, something in the normal process that will break if it doesn't find a certain kernel word? 16:52:25 nothing that I am aware of 16:54:42 It's the wordlist stuff. You see in retro-ans.fs, I create the words environment and end-environment, and put all the environment declarations between them. It's the first 'environment' that's crashing, and I can't for the life of me figure out why. 16:55:24 It's just the wordlist stuff, too. If I comment out the environment declarations completely, the rest of it passes coretest, but the rest of my tests assume the wordlist stuff is there, so I can't proceed until this works. 16:55:44 So I'll back up to the wordlist stuff and see if I can debug it there. 16:57:44 btw I have to use my definition of now? -- the kernel one, I'm not sure what it does. 16:57:50 But it explodes. 16:59:03 Is there some new condition that lookup returns on end-of-line or something? 16:59:31 Because the code makes sense; there must be some hidden internal coupling, some behaviour I'm not reflecting. 16:59:42 nothing that I changed 17:03:09 I'm afraid I'm lost at this point, the wordlist stuff is broken for an unknown reason. I'll have to come back to it at a later point. 17:03:30 ok 17:04:15 can you send me the retro-ans.fs that you modified for 9.3; I'll see if I can track down the cause of the problem 17:14:03 Sent. I wish you had regression-tested this against 9.3 as you built it; it would have been about a hundred times easier to detect which change broke it. 17:20:06 It's the execution of the word 'environment' that sends it packing. 17:20:45 What is exceptionally peculiar about that is if I don't call environment, but rather just do 17:20:50 environment-wordlist set-current forth-wordlist set-current 17:20:55 I don't get a crash -- not on that line. 17:21:41 I get no crashes with it 17:21:52 With what? The whole retro-ans.fs file? 17:21:56 yes 17:22:01 I'm using the windows version. 17:22:20 * crc tests... 17:22:27 and it does crash on the windows version 17:22:36 Yes it does. 17:22:39 I'll take a look and see if I can figure out the difference 17:26:32 I assume too you'll want that to be 930.tar.gz not 903.tar.gz 17:26:52 no, 903 17:27:02 929 -> 903? 17:27:11 902 000 17:27:15 000 being 'patchlevel' 17:28:20 it works on linux, bsd, and generic; just not Windows 17:31:38 I uploaded a new tarball that works under wine 17:31:39 Under FreeBSD 4.8, I'm getting a bad result. 17:32:04 I did the 'ld' thing, and retro starts up fine; it coughs up an endlessly scrolling error during one of the initial comments in retro-ans.fs 17:34:13 Ok, I downloaded the new tarball; it crashes under Windows. Takes about 3 seconds longer to display the crash box. 17:36:26 oddly, I am not having a problem on my FBSD 4.8 install, under either the bsd or generic versions 17:36:43 * crc will have to wait until tomorrow before he has access to a Windows machine again 17:37:15 I can't get it to parse the source in any way. The same lines typed at the console work, but -f is bombing out badly. 17:38:00 include appears to be a null-operation in this executable. 17:40:17 very odd 17:42:35 I have include working -- one of the windows includes must not have been updated. But the retro-ans.fs crashes hard after a couple of seconds of nothing. 17:43:37 It crashes on the first definition of a word. 17:50:36 Ok, some progress. My key problem was one mistyped character and rf's complete lack of error trapping. I'm running my tests; my optimization file falls over, but that's ok. The qsort bombs hard. I'll have to figure out why. 17:50:56 I renamed my now? to now so as not to conflict with yours, and for some reason retro said now? now? now? and then crashed, despite my making no reference to now? in the entire source. 17:51:08 Is that some kind of error message? 17:51:24 normally that would mean that 'now' could not be found 17:51:34 * crc has to add back the somewhat more descriptive error message 17:51:42 * crc also has to go eat 17:51:48 Ok. That's new. I wish to GOD the thing would come to a halt on an error. I would actually convert to a religion in order to pray for that. 17:54:52 Ok. Nested includes completely don't work with the windows binary. They just sail past and do nothing. 17:55:53 If only someone would invent error handling, so that it could be used in compilers like this, say, for instance, if include was unable to open or load a file... think how much easier things would be for the hapless users of that system! :) 17:59:22 --- quit: nighty (Client Quit) 18:04:44 Ok. Working with bare rf-windows.exe, nested includes seem to work in a tiny test. So it's back to the wordlist code not working, because include/needs in the ANS layer work with a wordlist to keep track of what file is loaded, and what isn't. 18:05:29 I'll try to get this running in 4.8 again, see what the trouble is there, see if the problem exists still just in the windows binary. 18:08:18 Same result in the 4.8 binary, now that I've got the support files in the right place. The wordlist stuff doesn't crash, but it doesn't work properly anymore. 18:15:41 Ok, latest diagnostic. -f retro-ans.fs -f arcfour.fs loads fine, and the 'needs' in arcfour.fs load fine, which would suggest the wordlist stuff works (at least it doesn't crash). However -- 18:16:04 retro -f retro-ans.fs followed by include arcfour.fs fails as soon as it hits the first 'needs' in arcfour.fs. 18:16:09 So different behaviour between -f and include. 18:16:23 That's all in rf-bsd. I'll try to reproduce in rf-windows.exe. 18:16:52 Exactly the same result in rf-windows.exe. 18:17:11 I think we encountered this in 9.2.something, it wouldn't allow nested includes. 18:17:56 When an include fails, whether it's because of a misnamed file or some other reason like this one, it's utterly silent. It leads to some really bizarre results that make it hard to debug. 18:18:33 In this case, the failing include causes further parsing of the file to just stop silently; looks like the file loaded, but there ain't no thing there. :) 18:22:53 Out of curiosity, why is -f not simply an alias for include (or vice versa)? 18:25:15 It looks like a two-level nested include fails even when the initial include works on a file loaded with -f. So it's include including an include that fails. 21:45:36 * crc updated 903.tar.gz with one that does handle file nesting properly 21:46:32 at least, it works on linux and under wine 21:46:36 Testing. 21:51:28 That binary is crashing at the same spot here. 21:51:53 The 'environment' spot. Preceding the include problem. 21:55:44 It is exhibiting the same include problem under BSD 4.8. 21:58:53 It seems to have gone backward in time. I used the .org address to get the 903.tar.gz package, is that the correct link? 22:02:27 yes 22:03:11 Exactly as it was on the previous revision. It works in BSD except that nested includes do not; it crashes on the 'environment' line in Windows. I've confirmed I'm working with the new tarball, and not the old. 22:03:21 Both downloads were done on completely separate systems. 22:04:02 I am going to have to setup Windows on this box tomorrow :( 22:04:14 But even the BSD one is tripping on the include issue. 22:07:14 ./retro -f retro-ans.fs -f arcfour.fs bye 22:07:16 I haven't been able to replicate this 22:07:20 runs my six arcfour tests just fine. But 22:07:25 ./retro -f retro-ans.fs 22:07:26 include arcfour.fs 22:07:33 how big are the include files? 22:07:35 does nothing, as it hits the needs line at the top of arcfour.fs. 22:07:51 arcfour.fs is 6K. 22:08:07 Modules.fs is 1.5K. 22:08:51 Retro-ans.fs is a bit over 44000 bytes, but I use -f on the command-line for that. 22:09:16 I know that it begins to load arcfour.fs, as I can insert text output instructions ahead of it and they work. 22:09:22 and it works on 9.2.9 or 9.2.10? 22:09:27 Yes. 22:09:41 All my tests ran flawlessly up until 9.3. 22:10:36 I run about 30 of them. The only ones now failing under BSD are those with nested includes. The only ones now failing under windows are all of them except coretest, because I can't switch wordlists again. 22:11:28 * crc tries one more thing; a new 903.tar.gz is up 22:12:38 That works now in BSD, at least for an include nesting depth of one. I'll try windows. 22:14:47 The windows build is still crashing hard. 22:14:50 :( 22:15:21 I'll start doing a comparison between 9.2.10 and 9.3 and see if I can find any changes that would be causing this 22:16:03 It's definitely at 'environment'. You fixed this earlier, was that accidental? 22:18:11 that should *not* be broken 22:18:47 It's broken a lot. I'll just confirm the heck out of the fact that I'm running the code I just downloaded, but I know I am -- I've been overwriting successive builds you've sent, and this one brought the bug back. 22:19:07 09/05/2006 01:17a 16,384 rf-windows.exe 22:19:07 :( 22:19:30 09/05/2006 12:49a 2,272 windows.f 22:19:47 that would be it :( 22:19:58 09/05/2006 01:17a 2,193 rx.dict 22:19:59 what version of Windows? 22:20:02 Win2k. 22:20:18 ok 22:20:54 Something did correct this earlier. 22:20:58 * crc is going to try this at work tomorrow; if I can't determine the cause, I'll restart and build back up step by step 22:21:04 Ok. 22:21:58 for now though, I have to get some sleep; 6 hours until work begins 22:22:04 Ok, tomorrow! 22:48:08 --- join: Cheery (n=Cheery@a81-197-19-23.elisa-laajakaista.fi) joined #retro 23:59:59 --- log: ended retro/06.09.04