00:00:00 --- log: started retro/12.03.22 02:03:21 --- join: roarde (~roarde@pdpc/supporter/active/roarde) joined #retro 02:21:29 --- quit: virl (Remote host closed the connection) 03:23:16 --- quit: roarde (Quit: Leaving) 07:11:09 --- quit: SimonRC (Ping timeout: 244 seconds) 07:18:16 --- join: SimonRC (~sc@fof.durge.org) joined #retro 07:45:19 --- quit: karswell (Remote host closed the connection) 07:55:35 --- join: karswell (~coat@93-97-29-243.zone5.bethere.co.uk) joined #retro 09:18:04 --- join: __tthomas__ (~tthomas@slosvr6.aeromecheng.com) joined #retro 09:19:48 --- quit: SimonRC (Ping timeout: 244 seconds) 09:21:35 --- join: SimonRC (~sc@fof.durge.org) joined #retro 09:44:02 <__tthomas__> crc if you see Raystm2 let him know I use bzr on windows.. two seperate computers in fact, I know how to set up ssh certs and can show him the doc on that.. I had to dig a bit but it isn't too bad, have to use putty to generate ssh key and pageant to register it.. 09:54:44 --- join: virl (~virl__@85-127-248-189.dynamic.xdsl-line.inode.at) joined #retro 10:12:56 __tthomas__: will do 10:13:25 * crc does all his work under netbsd and linux 10:14:04 <__tthomas__> when I use linux it is almost always through a ssh session on windows.. 10:14:26 <__tthomas__> usually OS is in a vmware image... 10:14:36 netbsd is in a vmware session, limux is a vpn 10:15:13 I use an iPad as my main computer, and ssh in to whichever is more convienent at the time (moatly linux) 10:15:21 <__tthomas__> my ben nanonote is running wrt linux, and my efika mx is running arm version of ubuntu.. 10:16:06 <__tthomas__> so I do have a few native linux boxes.. 10:17:19 <__tthomas__> and I have my shell account on dune.net, really need to use that more.. I have been wanting to write a mud on it forever, but with only 25 megs of ram have to be careful of how it is programmed.. 10:19:52 <__tthomas__> your the first person I ever met that actually uses their ipad... Some people use it for browsing and watching movies, and always have it on them.. but it is little more than a toy for those uses.. 10:30:44 I use it for almost everything. I have a mac mini, but that is only used to download torrents and provide mass storage for media now 10:31:20 everything else can be done on my ipad, or via ssh+vnc on my servers and work box 10:31:30 --- join: Kumul (~Kumul@adsl-72-50-91-28.prtc.net) joined #retro 10:42:47 <__tthomas__> nice, I should set my net book up simarly or use my nook color for those types of things.. I must admit even though I know several people that work there I won't support that company, so I don't buy any of their products, same with sony.. 10:43:15 <__tthomas__> though sony I am a little more lenient about, I still watch movies and listen to music so they are getting some of my money.. 10:50:04 <__tthomas__> I never could get very good performance out of vnc... Probably from running it over wireless... 11:06:31 I used s nook color with cyanogenmod 7 before getting the ipad 11:07:59 the larger screen on the ipad makes it a it more usable for me, and I couldn't find any android tablets with similar build quality and battery life at the time 11:11:15 <__tthomas__> true, the size and weight were the two things I didn't like about the ipad.. my nook color 7 inch also has cyanogenmod 7 on it.. Also the cost would have factored in but the heft of the thing already turned me against it since I already have a netbook.. 11:12:32 <__tthomas__> tried using my phone for development, but the keyboard is missing some very important keys, can't even play nethack.. :) though I do have terminal ide so I can ssh into my phone and develop java android apps directly on the phone.. Which is great since android dev kit is so huge.. 11:12:56 all I had was a mac mini, and nook color at the time. and I was willing to sacrifice weight for battery life 11:13:32 I have the android dev kit on my linux server 11:13:35 <__tthomas__> does it get more batter life than your nook color? my netbook gets 11 hours, has extended battery.. 11:14:01 <__tthomas__> I honestly never use my nook color.. 11:14:36 <__tthomas__> only thing I really like about it is it renders pdf's well.. 11:14:48 I get 10-12 hours from the iPad, and around 4-5 on my nook color 11:16:18 <__tthomas__> 7 inch is a weird size, its great for holding in my hand, but typing on it is awkward, on smaller device you can thumb type.. 11:17:24 I used a bluetooth keyboard with the nook color 11:17:50 <__tthomas__> nook color has bluetooth? I just assumed it didn't so never tried.. 11:18:32 it does, just not officially. cm7 nightlies have the drivers included, though range is limited (1-2 feet max) 11:19:12 <__tthomas__> that is pretty cool, I will probably just throw it on my network with static ip and use it for writing python and java apps for android.. 11:46:15 --- join: Mat2 (5b4085c5@gateway/web/freenode/ip.91.64.133.197) joined #retro 11:46:29 hello ! 11:51:13 --- quit: virl (Remote host closed the connection) 11:59:43 <__tthomas__> hi 12:03:42 hi 12:04:00 <__tthomas__> how is the vm work coming along? 12:04:24 testing phase 12:04:42 <__tthomas__> crc how do you test compatibility and correctness of vm, do you have an image you use for testing? 12:05:53 <__tthomas__> cool, thinking of trying to implement a small app for generating sodoku series, figure be good practice.. 12:07:36 <__tthomas__> I wish I still had it, made a one liner in python a few years ago that could it.. think it was actually more than one line, but was as close as I could get at the time using generators.. 12:09:37 i'm implementing a monitor interface so testing all instruction combinations can be interactively done 12:10:08 <__tthomas__> so 16 * 16 possible combinations? 12:10:20 yes 12:11:25 <__tthomas__> cool reminds me of how zilog was ported to propeller, they realized most instructions were just same instruction with different argument for registers and memory, so were able to make it into a table, so went from 4 to single core.. 12:11:48 <__tthomas__> meant z80, not zilog 12:12:20 that's the fastest way implementing token threading 12:12:58 i'mm reduce dispatch overhead much more because instruction streams can be compiled into new instructions 12:13:12 I'm, sorry 12:13:42 and the vm support stream processing 12:13:48 <__tthomas__> yeah, vm I was working on in retro for ngaro just made an array of quotes, then would check to see if instruction was less than 32 then just index into array at that number and call do.. 12:15:39 __tthomas__: the test/core.rx is all I really have at this point 12:16:41 sadly this check up < 32 decrease performance a lot, look: 12:17:01 it doesn't cover all of the i/o functionality, but *should* fail if any of the core instructions are wrongly implemented 12:19:12 let RDI = pointer into opcode table, RBX = opcode word (16 instructions bundles) 12:19:27 typical NEXT would be: 12:19:47 MOV DL, BL 12:19:54 SHR RBX, 8 12:20:23 <__tthomas__> crc well at least that is something.. :) quite detailed.. 12:20:30 LEA RCX, [RBX+RDX*8] 12:20:37 JMP RCX 12:21:32 <__tthomas__> crc besides no guarantee files are even used for given platform.. 12:22:12 <__tthomas__> Mat2 how fast is c vs. assembler versions.. ? 12:22:51 the c vm is estimated 30% slower (gcc -O6) 12:23:02 depens on the cpu 12:23:13 and compiler 12:23:47 __tthomas__: true. an image for testing purposes is something I'll have ready before the next release 12:23:52 <__tthomas__> yeah, gcc is not a very good compiler.. at least on x86/64 platforms.. 12:24:31 it was a long time before I could verify that the javascript implementation worked properly 12:24:33 <__tthomas__> for others, it is all you have.. so best by default.. 12:26:12 <__tthomas__> crc javascript is kind of nasty in that regard.. chrome has typed arrays now, think w3c is going to standardize on that, which should make things like vm's much easier to implement.. 12:27:54 I've dabbled with thped arrays in javascript, but didn't get any significant performance benefits yet 12:30:15 <__tthomas__> the slowness is generally the dom manipulation, and the amount of overhead from included libraries like jquery.. It is small but contains an absolute ton of code... you would see most benefit from things like pixel manipulation in canvas.. 12:30:54 I suggest a good JIT compiler is more important for performance than the typed array feature, it's a dynamic language 12:32:57 must say, I don't have much experience with javascript 12:33:01 <__tthomas__> I think typed array was more for efficiently handling arrays for doing audio and video, not for general speedup.. It will be much faster for specific things.. Like should see a difference in copying memory assuming it doesn't have to go through stack.. 12:35:11 <__tthomas__> yeah, JIT's can be incredible, usually first iteration of loop is slower, but then after can optimize, so if code has lots of tight loops will see huge improvement.. But can also potentially make other code slower, I think in pypy it is recommended to turn jit off for some operations.. 12:36:37 the tradeoff between optimisation time and runtime is critical so in most situations you end up in very ugly (but fast) generated native-code 12:37:54 JIT consumes huge dynamic memory and processor ressouces so for me it is a kind of ugly hack for getting performance out of bad vm designs 12:38:45 <__tthomas__> I personally think runtime profiling is the way to go, then everytime you run your program can potentially optimize using information it gathered from previous runs.. 12:41:13 if you have a fast. interpreted vm you don't need native code for most applications, this could bring you the time for intensive runtime profiling for hotspots where native-code generation would be needed 12:42:50 <__tthomas__> http://www.cs.indiana.edu/~dyb/pubs/nano-jfp.pdf 12:43:20 <__tthomas__> not sure if even need native code, very possible could simplify your paths and fold them in to new vm code.. 12:47:17 True. The next routine above executes in only 2 clocks on modern x86 cpu's. Together with instruction bundling native-code wouldn't be significant faster for some applications 12:48:32 if the instruction-set is well composed 12:48:36 <__tthomas__> I was more thinking have support for platform specific opcodes, you would never use them directly.. 12:51:09 hmm, some microcode like design ? 12:53:01 <__tthomas__> like in ngaro could have space for 64 opcodes, anything above 32 would just hit memory and do a direct call which could just be an exit, or another call if you wanted to emulate them.. but other vm's would actually do things with these opcodes like move value from stack into a register, and use registers for memory to memory copy.. or floating point.. 12:53:18 my vm have basic and composed instructions which executes in two slots and implement quite complex operations like this one: LaIncBsEq 12:54:57 This instruction loads TOS with the value from pointer at stack element n, increment the pointer and branch to adr if SOS (second of stack) equals to the pointer value 12:56:56 this way I can reduce 4-6 basic vm instructions to a single, complex one 12:57:18 I call this instruction fusion 12:59:48 for implementing vector and floating-point operation the vm can be switched 13:00:27 the instruction-set doesn't change but all instruction then perform vector or floating-point instead of integer operations 13:00:49 <__tthomas__> very similar to streamer16 design, only has 8 basic opcodes, but several slots per instruction, gulwig was also like that.. Eric Laforest did a thesis on it.. http://www.eecg.toronto.edu/~laforest/Second-Generation_Stack_Computer_Architecture.pdf 13:01:57 <__tthomas__> sheesh, it is scary how obsessive I am ... 13:02:31 I wasn't aware of the streamer16 design before you mentioned it 13:03:17 it seems we both are a bit obsessive but I think that's not a bad thing :) 13:03:42 thanks for the pdf, will study it 13:03:58 <__tthomas__> That thesis is pretty awesome, he actually got a doctorate for independent studies in forth... 13:04:13 ok, have some work to do, see you later 13:04:21 <__tthomas__> me too, later man.. 13:04:33 --- quit: Mat2 (Quit: Page closed) 13:13:15 hmm, the push to git is broken 13:23:47 git won't let me push commits without a space between username and email. 13:27:16 <__tthomas__> heh, thats lame.. I don't like bzr's merge.. I can't get it to work with winmerge, so have to use kdiff3.. 13:35:11 the mirror to hg (on bitbucket and assembla) still works, I just cant push the git repo to github 13:35:50 <__tthomas__> does that autoupdate, I prefer mercurial, though not really a good reason to switch I guess.. 13:37:58 it's updated at least once an hour, often more 13:38:59 I have a cron job that updates the snapshot on the website and updates the git and hg mirrors 13:39:36 <__tthomas__> Cool, what made you choose launchpad? 13:39:45 though I now need to find a solution to the git issue. (this is not the first time I've run into problems with github) 13:40:22 <__tthomas__> github seems to be the most popular for some reason.. I personally like codeplex.. 13:40:24 ok2 (oleksandr) convinced me to give bazaar a try when I began having prolems with fossil 13:41:07 <__tthomas__> I use fossil for personal project have it hosted on dreamhost via cgi, what problem did you have? I might need to switch myself, though my project is private for the time being... 13:45:45 <__tthomas__> neat looks like codeplex and google code both support mercurial.. 13:50:26 --- join: Mat2 (5b4085c5@gateway/web/freenode/ip.91.64.133.197) joined #retro 13:50:28 when I lst looked at google code, they didn't allow projects to use anything but a very small set of licenses 13:50:46 hi ! 13:50:53 <__tthomas__> I should look what license I was using on google code.. I don't remember.. 13:51:16 with fossil, I repeatedly ran into issues where a merge would create two parallel timelines, with duplicate commits, and no tools to merge them 13:51:21 hi Mat2 13:52:18 crc: have you observed similar problems with your current tool-set ? 13:52:54 no 13:53:14 <__tthomas__> hmm.. for one project was using MIT for other was using New BSD.. on google code.. 13:56:19 <__tthomas__> Mat2 http://www.strangegizmo.com/forth/MISC/msg01721.html I had name wrong it's steamer16 13:58:13 very nice design ! 13:58:28 http://rx-core.org/rx.fossil/timeline?n=200 shows the split timeline 14:01:05 __tthomas__: I had a prototype vm with 8 instruction, 5 slots and LSB as branch bit 14:02:16 <__tthomas__> My current theory about technology is that all technology is just a rediscovery of technology from early 60-70s.. Or it seems so as far as electronics and software goes.. 14:02:39 but the performance advantage was nearly absent because code-pressure is much higher as with 16 instructions 14:04:08 <__tthomas__> CISC vs. Micro Coded MISC, complexity vs. minimalism.. Who will win? 14:04:46 * crc no longer cares ;) 14:05:04 <__tthomas__> heh.. :) 14:05:59 multicore VLIW internal, superscalar and energy hungry cpu monsters with x86 ISA 14:06:57 with exception of china, there MIPS has aredy won the race 14:06:57 <__tthomas__> There is also the one instruction set computer... Subtract Jump if not zero 14:07:45 oh, that's a very interresting topic 14:08:43 <__tthomas__> I really hate C++ 14:09:30 <__tthomas__> 90% of my day is figuring out why something doesn't compile, 10% is actually coding.. 14:09:51 C++ is a mess 14:10:04 <__tthomas__> C it is closer to a 30 70 split and is almost always a typo.. 14:10:34 I don't like the syntax of C either 14:11:26 <__tthomas__> C is almost a good language.. there are a few things I would change.. 14:11:51 alone, that these pointer operator is reused as indirection prefix 14:12:18 then you will found Go attractive, I suggest# 14:12:54 <__tthomas__> I don't like Go since they removed pointer arithmetic a feature I use A LOT.. 14:14:07 ok, that's not a good feature for a system programming language 14:14:15 but the syntax is better 14:14:46 <__tthomas__> I do hate that multiply operator, dereference, and reference are all the same operator, static has 3 different meanings and operator precedence rules don't always make sense, so have to use parenthesis everywhere, I would have preferred left to right evaluation for expressions.. 14:16:27 <__tthomas__> There are a few other things that annoy me about it, but overall the syntax is small enough that I can mostly keep it all in my head.. 14:16:37 I think phyton have a clear syntax 14:17:09 <__tthomas__> yes, until you learn how to use it really well.. Then code becomes very non obvious to someone new to the language.. 14:17:19 <__tthomas__> but it has a very nice learning curve.. 14:17:30 there exist genie, a alternative syntax for vala which roots in phyton 14:18:16 I will rewite my C vm in genie for the next version 14:18:26 <__tthomas__> coffeescript is much nicer than javascript, combines ideas from ruby and python.. and it compiles to javascript, so can use it with node.js which is a nice environment.. 14:19:08 http://live.gnome.org/Genie 14:20:03 coffeescript is nice ! 14:20:41 <__tthomas__> another favorite.. http://boo.codehaus.org/ it is like python but statically compiled for .net... 14:22:41 <__tthomas__> though if I do write a vm anytime soon, will probably be in haxe or processing.. haven't decided which yet.. haxe would be nice since then we would get a c++ and flash version for free.. 14:22:56 I prefer native-code compilation as it's not a good idea to base upon huge runtime-librarys for vm implementations 14:23:31 like CLR 14:23:38 <__tthomas__> *shrug* .net can be faster than native code a lot of times, especially when compiling for 32 bit and running on 64bit.. 14:24:14 <__tthomas__> its jit can produce cpu specific op codes and still be portable.. 14:24:30 probably but the ressurce demands are very high 14:26:50 <__tthomas__> for most applications that doesn't matter though, which is the sad truth for our industry.. otherwise there would be no such thing as modern OS's which take gigabytes of hard drive to install and gigs of memory to run.. 14:27:37 --- join: Mat2_ (5b4085c5@gateway/web/freenode/ip.91.64.133.197) joined #retro 14:27:52 <__tthomas__> You can't even get a phone these days that can run decently with less than 1 ghz multicore processor and half a gig of ram.. 14:28:30 why waste ressorces which can be used otherwise ? 14:28:43 --- quit: Mat2 (Ping timeout: 245 seconds) 14:30:19 but I see your point 14:30:37 --- nick: Mat2_ -> Mat2 14:31:40 I have burned my USB stick 14:31:46 *scratch* 14:32:00 <__tthomas__> if you haven't you should look at rebol, it is amazing how much functionality they managed to getin under a meg.. 14:32:45 yes, rebol is a very clever design 14:33:51 <__tthomas__> I would love to port a lot of their one liners over to retro once we get socket support, could probably already port over the file ones.. 14:34:19 some one had added socket support as I know 14:34:34 should be one of the experimental vm's 14:35:17 but i'm not sure if it in the repro 14:35:23 <__tthomas__> yep, and I added a version as well... though for server we would need either nonblocking sockets or multiple threads, I prefer rebol's nonblocking ports myself.. 14:38:18 we could reserve some ports in ngaro for winsock-library functions 14:39:10 <__tthomas__> yep, I believe already a reserved set of ports, so for testing could just use nonreserved port.. 14:40:59 crc: what about standard nonblocking-socket support in ngaro ? 14:41:49 hmm, this library is a mess ! 14:42:02 <__tthomas__> I was going to wait till we started curses support and see how we handle nonblocking keyboard events... 14:43:57 mat2: I'm open to suggestions on sockets. The only catch is that whatever is done should be doable in more than one implementation. 14:44:34 for experimental work, use any port above 12. 14:44:39 <__tthomas__> crc yeah, should do C# or java, python, and c implementation simultaneously.. 14:45:07 __thomas__: you must set a special flag in the ioctltty function for nonblocking keyboard mode 14:45:37 the tty intercace is as ugly as the socket one 14:45:45 interface 14:46:37 for init nonblocking socket: ioctlsocket(sock, FIONBIO, &buffer) 14:46:59 <__tthomas__> on windows that only works for sockets, you have to use conio library... so to make c code identical could use PDCurses and ncurses.. 14:47:33 no chance for ncurses because of its incompatibility with tty interface 14:48:06 and I don't see this functionality in the library 14:48:21 but a workaround may be possible 14:50:57 here is a solution: 14:51:01 http://stackoverflow.com/questions/448944/c-non-blocking-keyboard-input 14:51:53 <__tthomas__> curses, I believe getch() is nonblocking.. 14:52:15 <__tthomas__> you just check to make sure getch() != ERR 14:52:15 WINDOW *w = initscr(); cbreak(); nodelay(w, TRUE); 14:52:55 this ensures getch() is nonblocking 14:53:26 <__tthomas__> making ansi getchar() nonblocking is way trickier.. 14:53:42 I've never tested the image on a vm with non-blocking input 14:53:46 seems so 14:54:24 <__tthomas__> I was shocked when I found out getchar() didn't return after a keyboard press, only returns when enter key is hit.. 14:56:06 <__tthomas__> Hmmm.. I thought our windows code set socket to nonblocking for sockets, but looks like we just change the timeout on socket to 1ms instead.. 14:57:19 why don't we use these conio.h port to linux ? http://jakash3.wordpress.com/2011/12/23/conio-h-for-linux/ 14:57:24 <__tthomas__> crc I imagine would need to add a 20 ms sleep in wait opcode so as not to be in a constant burn loop.. 14:57:45 <__tthomas__> same reason we don't use cygwin and termios? 14:58:11 ok 14:58:50 <__tthomas__> I assume ansi C99 version was a goal.. curses and console output would be an option.. 15:01:48 mat2: there's no license information listed on that conio code 15:02:16 I used to have a vm implementation usng curses 15:03:02 <__tthomas__> If you give me a linux version I can #ifdef to make it compile on windows and linux, someone would still need to do the work to make sure it runs on BSD's and MacOS-X 15:03:45 ok, I will test it 15:03:49 <__tthomas__> I am pretty sure PDCurses is almost completely compatible with linux curses.. 15:04:26 ncurses have some extensions but these are to specific to be useful here 15:04:48 and there exist some incompatibilities with PDCurses also 15:05:28 <__tthomas__> for tty can just use escape sequences and standard IO per normal.. 15:05:51 <__tthomas__> which would work with sockets and telnet.. 15:07:09 <__tthomas__> eventually when there is a canvas version of terminal, should be generic enough to work with javascript, SDL surface, and opengl.. 15:07:40 so ANSI escape sequences seem to be the way to go 15:08:15 and a single function for nonblocking input 15:08:24 <__tthomas__> eh? depends, if on tty only platform could just have ports generate ansi escape sequences.. 15:08:31 <__tthomas__> or use a different library.. 15:10:28 crc: I don't know the copyright situation in the US but in europe source-code in pblic forums is common, if not otherwise declared 15:10:33 <__tthomas__> eforth is meant to work over serial, so uses nonblocking IO, they just have two ports, one is flag for IO set, other is to read IO.. pretty easy and simple enough to support, for platforms that block, flag would just always read true, since anytime check IO it is going to have a value... 15:11:02 good idea 15:11:19 <__tthomas__> In the US the original author is able to relicense at will... 15:11:34 <__tthomas__> so if no assumed license author could still sue.. 15:12:07 __tthomas__: I just committed a quick stab at a curses based vm. The port 5 stuff for screen size needs to be adapted for curses, but the input and display stuff is switched over 15:13:05 <__tthomas__> awesome.. :) when I get a chance will have a look, I am technically working right now.. *I work from home* 15:13:54 __tthomas__: Crazy ! 15:15:57 need some sleep, ciao ! 15:16:07 --- quit: Mat2 (Quit: Page closed) 15:19:52 --- join: virl (~virl__@85-127-248-189.dynamic.xdsl-line.inode.at) joined #retro 15:42:01 --- join: CIA-98 (~CIA@cia.atheme.org) joined #retro 15:43:25 retro: 03crc  * rcrc@questor-20120322224302-001ctzza7j1xsuvg 10rx/RELEASE_NOTES.rst: mention retro-curses in release notes, along wth double' vocabulary 15:44:17 all termios stuff is removed from the retro-curses now 15:44:38 <__tthomas__> I was wondering, I was just removing stuff.. 15:46:49 <__tthomas__> crap, I deleted that file, how do I re-pull a revision? 15:54:59 <__tthomas__> hmm.. I prefer what svn does, if you delete a file it just pulls the newest.. 15:55:17 <__tthomas__> looks like in this case I am screwed till commit again.. 15:59:33 <__tthomas__> cool, that was easy.. 16:05:56 <__tthomas__> it works, but enter no longer causes it to parse, now only space does.. 16:06:49 I'd do bzr revert filename 16:07:07 <__tthomas__> I got it working now, no need to 16:07:18 <__tthomas__> I had to download it from web and copy it over my file.. 16:07:56 enter is working on my linux box 16:08:29 <__tthomas__> oh, it does work, just have to hit it a few times.. 16:08:54 <__tthomas__> I know why, on windows its \r\n, not just \n 16:11:02 the image should translate \r into \n 16:11:16 <__tthomas__> yeah, but windows sends both.. 16:12:03 I'll have to look nto that 16:13:23 <__tthomas__> hmm.. that didn't fix it.. oh well.. 16:16:25 <__tthomas__> I tried to change 16:16:28 <__tthomas__> c = getc(vm->input[vm->isp])) == EOF 16:16:29 <__tthomas__> to 16:16:33 <__tthomas__> c = getc(vm->input[vm->isp])) == KEY_ENTER 16:17:17 <__tthomas__> oh, heh, it isn't reading using curses, it is using stdio.h 16:18:10 I noticed that, but fixing it will require a bit more effort (the function has to jandle both the keyboard and the file input stack) 16:19:43 <__tthomas__> hmmm.. couldn't use iostream, vs. filestream in ports, or am I missing the issue? 16:21:08 Ive not thought about it much 16:21:26 <__tthomas__> got it.. 16:23:06 <__tthomas__> do you have doc showing ports for cursor and color? I could play around with it tonight if you do.. 16:24:31 not yet 16:25:37 <__tthomas__> it's exciting, we are very close to having a game of snake/nibbles working in realtime.. 16:25:59 I was planning to draft the additions to the ngaro spec over the weekend 16:26:26 <__tthomas__> no problem.. :) 17:28:46 I have a branch for the enhanced console bits setup at https://code.launchpad.net/~crc-x/retro-language/ngaro-port8 17:29:20 the spec on this is far from complete, but any comments would be appreciated 17:30:28 <__tthomas__> Oh cool, no problem, I was kind of wanting to do a small write up of just some things I have been thinking of anyways.. Though the purpose and requirements for 32bit vs. 16 bit version is quite different in my mind.. 17:32:18 <__tthomas__> how do file names work for file operations, is that direct access to memory to get filename, or does it do it one cell at a time? 17:35:20 the c vm copies the filename to a native char array, then uses that. pyon and ruby do similar things. Im not sure about go or lisp 17:59:32 <__tthomas__> ah.. I was doing it different for my sockets which is why never committed.. I was sending an address, then when I would write would just let python copy bytes directly.. so DMA like.. 17:59:52 <__tthomas__> I committed the experimental python if want to look, got to go meet the wife for food.. 17:59:57 --- quit: __tthomas__ (Quit: Leaving.) 19:05:03 --- quit: karswell (Read error: Connection reset by peer) 19:10:15 --- join: karswell (~coat@93-97-29-243.zone5.bethere.co.uk) joined #retro 19:21:19 retro: 03crc  * rcrc@questor-20120323022056-oncym2zzyysebewf 10rx/vm/experimental/ (4 files in 2 dirs): pull tthomas tweaks to retro-curses, experimental python vm w/sockets support 19:23:25 retro: 03crc  * rcrc@questor-20120323022303-dl0p2ejb2g2d8wbw 10rx/vm/experimental/python_sockets/socket.rx: add missing Host: in header of python-sockets test code 20:09:58 I've moved the git repo to google code: http://code.google.com/p/retro-language/ 20:27:15 --- join: __tthomas__ (~Owner@c-24-23-234-140.hsd1.ca.comcast.net) joined #retro 21:05:18 --- quit: Kumul (Quit: gone) 23:59:59 --- log: ended retro/12.03.22