#emc-devel | Logs for 2007-12-15

[01:29:08] <jmkasunich> hi guys
[01:32:29] <jepler> hi jmkasunich
[01:33:43] <skunkworks> did you know 32bit windows can't see over 2tb?
[01:33:57] <skunkworks> I found that out today :)
[01:37:59] <jepler> skunkworks: hah
[01:38:24] <skunkworks> something I probably should have known.. oh well.
[01:38:49] <SWPadnos> I think disks should be able to be over 2TB, but not partitions
[01:40:58] <fsdafsd> 2TB lol
[01:41:12] <fsdafsd> thats a suicide wish right there
[01:41:51] <jepler> the internet says that ext3 with 1k blocks also has a 2TB limit. disappointing.
[01:42:15] <fsdafsd> when you have 10 TB then you can trade with other farmers
[01:43:14] <skunkworks> All I know is - I have the 5 750gb drives in raid 5 - it finished building and windows didn't see it. I email the company and tried a smaller 'test' raid partition of 1gb which worked. emailed the company and they said - windows 32 bit will not see a drive greater than 2TB
[01:44:54] <skunkworks> so now I am making 2 1.36TB raid groups
[01:45:12] <skunkworks> For Shits and Giggles <-technical term
[01:45:38] <skunkworks> says it should be done in 15 hours
[01:46:04] <fsdafsd> fuck face
[01:46:21] <SWPadnos> nah. that'sfutile. just send me the 750GB drives, and I'l swap them out for 320GB drives that you can actually use
[01:46:26] <SWPadnos> at no charge
[01:46:36] <skunkworks> yeh :)
[01:46:37] <jepler> fsdafsd: excuse me?
[01:47:07] <fsdafsd> gigabyte shmiggabyte
[01:47:22] <SWPadnos> ir in this case, terabyte schmerabyte
[01:47:53] <SWPadnos> s/ir/or/
[01:48:12] <skunkworks> it does make a nice chunking sound now :) like it is doing something
[01:48:26] <SWPadnos> oh, that could be the chunk of death!
[01:49:04] <skunkworks> no - not the click click click (repeat) sound..
[01:49:11] <SWPadnos> heh
[01:49:17] <SWPadnos> I'm still waiting for my check from IBM
[01:49:50] <SWPadnos> I actually signed on to the class action suit against IBM because I had a deathstar drive
[01:50:12] <SWPadnos> it was the 60GB drive that I originally bought in this computer, in 2001
[01:50:32] <SWPadnos> I'm supposed to get a $100 settlement from them, which should buy about 500 GB once I get it :)
[01:51:32] <skunkworks> very cool
[02:01:08] <fsdafsd> so I bought a laser
[02:01:32] <fsdafsd> anyone know how thick of steel I can cut with only 45 watt ?
[02:03:13] <fsdafsd> I might be able to pay 6k for an opensource version of autocad
[02:03:27] <fsdafsd> anyone intrested in an undertaking like that ?
[02:03:39] <fsdafsd> strictly GTK
[02:04:22] <fsdafsd> half upfront half on completion
[02:04:24] <cradek> I think there are websites that help people find open source programmers
[02:04:41] <cradek> that might be a good thing to try
[02:04:43] <jmkasunich> have you looked at qcad?
[02:05:21] <fsdafsd> 10k ?
[02:05:48] <jmkasunich> do you have any idea at all how much programmer time would be needed to make something like that?
[02:05:55] <cradek> that will buy you one copy of autocad, not its entire development
[02:06:11] <cradek> but this is the emc-devel channel anyway
[02:06:15] <jmkasunich> you are talking thousands of man hours
[02:06:24] <fsdafsd> snap to the center of a cirle would not be hard
[02:06:32] <fsdafsd> and array's
[02:06:38] <fsdafsd> opengl makes that stuff easy
[02:07:02] <cradek> jmkasunich: did you see my maxkins pastebin?
[02:07:13] <jmkasunich> briefly
[02:07:16] <jmkasunich> didn't study it
[02:07:18] <cradek> I haven't tried it yet but it seems like it might be right
[02:07:42] <jmkasunich> you should make a vismax.py
[02:07:45] <cradek> I might try it as INVERSE_ONLY kins
[02:07:58] <cradek> yeah that would be nice. I could even make it to scale.
[02:08:24] <jmkasunich> one of these years I want to make a to-scale visbport
[02:08:40] <jmkasunich> all I need is a bport to measure
[02:08:41] <cradek> bring your tape measure and head on over
[02:08:50] <cradek> I'll start the coffee
[02:09:09] <jmkasunich> be careful what you ask for, you might get it
[02:09:21] <fsdafsd> take your android too
[02:09:38] <fsdafsd> or else
[02:10:43] <jmkasunich> cradek: you remember the original PC power supplies, with the female IEC power socket on the back for the monitor?
[02:10:54] <cradek> sure
[02:11:05] <SWPadnos> ah, the good old days
[02:11:23] <cradek> the good old days of the burned-in green screens with no power switch?
[02:11:36] <jmkasunich> got one laying around? or one of those female IEC connectors?
[02:11:37] <SWPadnos> I was thinking of the fine amber screens, but yeah
[02:11:55] <SWPadnos> do you need the combo in/out, or just the female?
[02:11:56] <cradek> jmkasunich: do you need the power supply or the connector?
[02:12:01] <jmkasunich> I have bunches of the male ones
[02:12:14] <jmkasunich> the connector - I want monitor power to come out of the electical box
[02:12:32] <jmkasunich> I suppose I could use a regular outlet, but that will take more panel space
[02:12:50] <cradek> do you have the male to female IEC cord?
[02:13:01] <SWPadnos> I have some short male-female IEC cables - you could use something like that if you don't mind a pigtail
[02:13:02] <jmkasunich> I think so (somewhere in this mess)
[02:13:04] <cradek> that would seem harder to find
[02:13:37] <cradek> a single outlet (the little square one) is no bigger
[02:13:55] <fsdafsd> I have IEC cables
[02:14:21] <fsdafsd> some big thick ones
[02:14:23] <cradek> http://timeguy.com/cradek-files/cnc/lathe/DSCN6296.JPG
[02:14:27] <cradek> like these little outlets
[02:14:37] <cradek> you just make a square hole for them to snap into
[02:15:06] <cradek> then you're not screwed if you end up with a monitor with captive power cord or wall wart or something
[02:15:33] <jmkasunich> true
[02:15:49] <jmkasunich> I just dug thru one of my IEC cord boxes
[02:16:16] <jmkasunich> I have a short male to female (about a foot), and a male to bare wires (6 ft or so), no male to female
[02:16:41] <cradek> I think they haven't been made (for mass market) since the AT days
[02:16:42] <jmkasunich> oddles of regular plug to female IEC of course
[02:16:47] <cradek> sure
[02:17:09] <cradek> I may have one or two but that's all
[02:17:23] <cradek> and I got rid of my green screens long ago :-)
[02:19:23] <jmkasunich> the little square 120V outlets are probably the way to go
[02:19:37] <jmkasunich> that panel is a bit down the road though
[02:19:52] <cradek> I even put one of those in the dash in my (last) car
[02:20:02] <jmkasunich> with an inverter?
[02:20:04] <cradek> it was nice to have an inverter available but not in the way
[02:20:06] <cradek> yes
[02:21:06] <jmkasunich> * jmkasunich chops the ends of of some IEC cables
[02:21:30] <jmkasunich> (computer power and monitor power are going to be fed from some DIN rail terminal blocks
[02:22:29] <jepler> looks like IEC female snap-in outlets can be had if you really want them: http://www.mouser.com/catalog/632/961.pdf
[02:23:33] <jmkasunich> ye
[02:23:34] <jmkasunich> s
[02:23:44] <jmkasunich> and cheap enough that scrounging doesn't even make sense
[02:23:55] <jmkasunich> but as cradek pointed out, the cable is the hard part
[02:24:14] <jmkasunich> for some reason I thought I had the cable, but either I was mistaken, or it is buried somewhere
[02:26:17] <jmkasunich> note to self: don't secure coiled cords with masking tape before storing for several years
[02:26:21] <jmkasunich> gummy mess
[02:26:39] <cradek> bleah
[02:26:44] <cradek> I like the removable cable ties
[02:27:17] <jmkasunich> or twist-ties
[02:27:30] <jmkasunich> it was only this one cable, I don't think I was the one who taped it
[02:27:33] <cradek> yeah that too
[02:28:01] <jmkasunich> fortunately goo-gone is cleaning it up nicely
[02:29:23] <fenn> goldmine-elec.com has this wire harness tape that works very well for keeping wires bundled for storage
[02:30:04] <fsdafsd> where is the best place to buy a 100 watt laser tube ?
[02:30:54] <jmkasunich> ebay?
[02:31:03] <fenn> http://www.goldmine-elec-products.com/prodinfo.asp?number=G14925
[02:34:30] <fenn> oh yes and IEC sockets: http://www.goldmine-elec-products.com/prodinfo.asp?number=G15608
[02:36:22] <fenn> or is that the wrong end
[02:37:19] <cradek> that's the common one
[02:37:52] <jmkasunich> fenn: jepler posted a catalog page (mouser) with both
[02:38:02] <jmkasunich> and I've decided not to do it that way anyhow
[02:45:37] <SWPadnos> I like those little square snap-in outlets
[03:13:54] <jmk-st> it has a computer now
[03:15:56] <cradek> yay!
[03:16:35] <cradek> I thought I was supposed to include "rtapi_math.h" but I'm not getting a hypot prototype
[03:16:50] <cradek> [5364862.130307] maxkins: Unknown symbol hypot
[03:16:52] <cradek> oh I see why
[03:16:57] <cradek> oh well that's easy
[03:22:10] <jepler> adding just the prototype fixes it?
[03:22:26] <cradek> I just #defined it the obvious way
[03:22:32] <jepler> oh
[03:22:45] <cradek> motmod requires kinematicsForward() even if type is INVERSE_ONLY
[03:23:00] <jepler> yes -- I don't think that's avoidable
[03:23:23] <cradek> I get following errors turning machine on - I wonder if it's using kinForward()
[03:23:37] <cradek> the joys of debugging in realtime
[03:23:53] <jepler> it worked in --enable-simulator?
[03:23:57] <cradek> yes
[03:24:00] <jepler> :-P
[03:24:08] <cradek> well who knows about FE
[03:24:23] <cradek> I didn't see any obvious position jumps but there's nothing checking it really
[03:26:14] <cradek> errr I mean when going into world mode
[03:28:36] <cradek> and halscope crashed when I clicked "force"
[03:29:13] <jepler> that's not good
[03:33:53] <jmkasunich> not good at all
[03:33:57] <jmkasunich> bitrot?
[03:34:12] <cradek> my fault (maybe)
[03:34:19] <cradek> oh halscope? not my fault
[03:34:55] <jepler> I haven't found it yet, but I broke something when I added 'continuous'.
[03:40:16] <jmkasunich> gimp's "perspective" tool is pretty neat
[03:40:34] <jmkasunich> I couldn't get a decent camera angle to take a pic inside the box, so everything was all distorted
[03:40:39] <jepler> yeah it is neat
[03:40:43] <jmkasunich> gimp corrected it (mostly)
[03:49:53] <jepler> grumble. an annoying bug in the gutsy 'rxvt-unicode' package and they mark it 'Fix Released' based on the fact that the package that will be in Hardy Heron 5 months hence may or may not have the problem fixed.
[03:49:58] <jepler> https://bugs.launchpad.net/ubuntu/+source/rxvt-unicode/+bug/144416
[03:50:11] <jepler> (the package is only missing almost all of its documentation)
[03:51:39] <cradek> once I get a FE in world mode, I can't turn machine back on - I get constant FEs
[03:52:07] <cradek> oh without forward kins, it can't set cmd to equal fb
[04:14:38] <jmkasunich> http://jmkasunich.com/cgi-bin/blosxom/shoptask/pc-cased-and-mounted-12-14-07.html
[04:49:41] <cradek> well would you believe it, the thing works
[04:52:48] <CIA-41> EMC: 03cradek 07w_tool_length * 10emc2/src/Makefile: build maxkins
[04:55:21] <cradek> jmkasunich: here (in rt) teleop nearly works
[04:55:31] <cradek> the accel phase sounds wrong but I didn't research it at all yet
[04:59:15] <cradek> cool, you have the same touchpad I bought. I had the same idea
[05:01:51] <fsdafsd> im too sexy for my car too sexy by farrr im a model if ya know what I mean when i do my little turn on the catwalk yeah
[05:02:12] <SWPadnos> thank you Vanilla Ice
[05:02:33] <SWPadnos> or whoever did that stupid song
[05:02:46] <SWPadnos> ah - Right Said Fred
[14:29:54] <christel> [Global Notice] Hi all! I am about to re-route a few of our European servers, expect there to be a few splits as I move things around. Thank you for using freenode and have a great day!
[17:35:04] <CIA-19> EMC: 03fenn 07TRUNK * 10emc2/src/emc/kinematics/rotatekins.c: allow arbitrary rotational offset - this might as well be useful as well as informative
[17:36:02] <cradek> someone stop me before I hook stepgen following errors to feedhold and call it joint limits
[17:37:25] <SWPadnos> stop!
[17:37:36] <fenn> why would you want to do that?
[17:38:28] <cradek> because joint limits are hard and stepgen etc already knows when a joint limit is being exceeded
[17:39:39] <cradek> well I'm probably going to try it anyway
[17:41:29] <CIA-19> EMC: 03compile-farm 07BDI-4.51 ( * 10emc2head/: build FAILED ; see http://linuxcnc.org/compile_farm/emc2head_slot6_log.txt
[17:42:06] <CIA-19> EMC: 03compile-farm 07Ubuntu 5.10 (breezy) realtime (2.6.12-magma) * 10emc2head/: build FAILED ; see http://linuxcnc.org/compile_farm/emc2head_slot2_log.txt
[17:42:38] <fenn> hrm. why didnt i get that error
[17:42:48] <jepler> probably because you were on sim?
[17:42:54] <fenn> yes
[17:43:04] <SWPadnos> you're using M_PI instead of M_PI1
[17:43:17] <jepler> you could also add a define for M_PI to rtapi_math.h
[17:43:41] <jepler> that header file got only the bare minimum needed to make emc compile, so there are gaps discovered from time to time
[17:53:46] <CIA-19> EMC: 03fenn 07TRUNK * 10emc2/src/rtapi/rtapi_math.h: M_PI seems like a pretty standard thing to use
[17:54:22] <jmkasunich> fenn: what it M_PI is already defined?
[17:54:33] <fenn> then it gets defined twice?
[17:54:45] <fenn> * fenn waits and sees
[17:54:46] <jmkasunich> also, haven't you just broken all the "non-standard" code that used M_PI1?
[17:55:05] <jmkasunich> I think you should grep for M_PI1 and change any you find to M_PI
[17:55:44] <jmkasunich> one last comment, did you drop the L from the end on purpose?
[17:55:45] <jepler> It's M_PIl -- the constant pi in long double precision.
[17:56:02] <jmkasunich> duh, so it is
[17:56:30] <jmkasunich> my point is still valid tho, that one is no longer defined, so unless no code used it, things are gonna break
[17:56:30] <fenn> does the L cast things that use M_PIl into a long?
[17:56:58] <fenn> why is one no longer defined?
[17:57:07] <jmkasunich> duh, sorry
[17:57:21] <jmkasunich> I mentally saw a - in front of the first line in the diff
[17:57:32] <jmkasunich> thought you changed the name instead of adding a name
[17:58:08] <SWPadnos> oh right - and I have one of those fonts where I can't distinguish between 1 and l, so I typoed the variable wrong
[17:58:09] <fenn> is defining something twice bad?
[17:58:21] <fenn> like, say M_PI is already defined because i'm running sim
[17:58:38] <jmkasunich> I believe your definition will silently override the previous one
[17:58:47] <jmkasunich> in this case that seems harmless
[17:58:47] <CIA-19> EMC: 03compile-farm 07BDI-4.51 ( * 10emc2head/: build PASSED
[17:58:47] <SWPadnos> with doubles, it's often better performance-wise to define a static constant instead of a #define
[17:59:07] <jmkasunich> SWPadnos: not in a header file I don't think
[17:59:11] <CIA-19> EMC: 03compile-farm 07Ubuntu 5.10 (breezy) realtime (2.6.12-magma) * 10emc2head/: build PASSED
[17:59:22] <SWPadnos> only if they evaluate to the same value -I think that's a warning, whereas a redefinition can be flagged as an error
[18:00:01] <SWPadnos> jmkasunich, right - there are problems using static const in a header, but it's still better for performance, even though the memory footprint will increase somewhat
[18:00:45] <jmkasunich> that is a bit of a generalization
[18:01:20] <SWPadnos> which? :)
[18:01:42] <jmkasunich> static var better for performance
[18:01:57] <jmkasunich> if the var is rarely used it makes little or no difference
[18:02:22] <SWPadnos> sort of - it ends up being a cache coherency thing at some point
[18:03:10] <SWPadnos> when you use a double, there's no way to directly load a constant into the FP register, so you have to load two 32-bit registers, store the two words somewhere, and load the resulting double from memory into the FP register
[18:03:37] <SWPadnos> if you have a static const, then the load happens directly, without the int registers being involved
[18:03:41] <CIA-19> EMC: 03fenn 07TRUNK * 10emc2/src/rtapi/rtapi_math.h: resolve any ambiguity in #define order
[18:05:07] <SWPadnos> or something like that
[18:05:34] <fenn> isnt that the compiler's job?
[18:06:03] <SWPadnos> that's how the compiler does things with double constants in the code
[18:06:31] <fenn> and does it do something different with a #define?
[18:06:42] <fenn> (i think you mean longs btw, since doubles are 32 bit)
[18:06:45] <SWPadnos> I don't think it'll automatically make a static var and load it (though it may now, this was a year or two ago that I heard/saw this trick)
[18:06:56] <SWPadnos> a #define is a constant in the code
[18:07:00] <SWPadnos> vs. a const var
[18:07:13] <SWPadnos> no, floats are 32-bit
[18:07:16] <SWPadnos> doubles are 64-bit
[18:07:27] <SWPadnos> and long doubles could be 80 bits on Intel architecture machines
[18:08:00] <jmkasunich> don't the FPUs have a few constants built in?
[18:08:04] <jmkasunich> like pi and e?
[18:08:08] <jmkasunich> or am I confused
[18:08:51] <SWPadnos> could be
[18:09:02] <fenn> #define M_PI acos(-1)
[18:09:03] <jmkasunich> (not that the compiler is gonna recognize that fenn's constant is PI
[18:09:24] <SWPadnos> it's a small issue, especially because FP isn't generally used in the very fast thread
[18:23:49] <cradek> 5? years ago I put some stuff in a box and carefully labeled it. Can anyone tell me where the box is?
[18:24:34] <SWPadnos> in the barn
[18:25:12] <jmkasunich> under another box
[18:25:19] <SWPadnos> jmkasunich, take a quick gander at the pluto-step code - there are some hard-coded constants in there to deal with FPGA limits, and they may also affect how it deals with the position command
[18:25:35] <SWPadnos> I know he's not describing a problem very well, but there could be a problem
[18:25:53] <jmkasunich> I'm not claiming that he is full of shit, I'm claiming that he isn't giving us enough information to go on
[18:25:56] <SWPadnos> I do need to run now - see you
[18:25:59] <SWPadnos> heh - that's true enough
[18:29:12] <cradek> I didn't think to look in the barn... duh
[18:29:39] <cradek> sometimes I forget that to find something, you have to look where you haven't yet looked
[20:40:38] <skunkworks> formatting 1.397TB X 2
[20:41:11] <skunkworks> it has been formatting for about 5 minutes - not even showing 1% ;)