#emc-devel | Logs for 2007-02-20

[00:34:10] <cradek> I don't think I've ever seen so much discussion about one line of code!
[00:35:55] <alex_joni> hah
[00:36:21] <cradek> hi alex, haven't seen you around this late for a while
[00:36:25] <skunkworks> :) paul probably didn't want eric to post it.
[00:39:14] <alex_joni> cradek: yeah, not sure why I stayed up
[00:40:38] <cradek> wonder what to do next to emc - it's sure slowed down since 2.1 release
[00:40:48] <alex_joni> * alex_joni feels the same way
[00:41:30] <cradek> maybe it's about done
[00:42:19] <skunkworks> rigid tapping?
[00:42:22] <skunkworks> ;)
[00:42:33] <alex_joni> cradek: I think it's like the dome in Cologne
[00:42:41] <alex_joni> the day it will be done, the world will end
[00:42:58] <cradek> ha
[00:44:08] <jepler> I think that yesterday a half dozen people promised to write new trajectory planners with spline support and infinite lookahead .. you can't say that's no progress.
[00:44:29] <cradek> haha
[00:44:30] <alex_joni> hahaha
[00:44:57] <alex_joni> I promise .. something
[00:45:08] <cradek> to be fair, they didn't promise to do it, they just don't see why it would be hard at all to do
[00:45:10] <alex_joni> (couldn't think of something plausible)
[00:45:35] <skunkworks> cradek: dejavu?
[00:45:48] <alex_joni> skunkworks: each and every day
[00:47:14] <skunkworks> for a while there I didn't think I was going to be able to make it to the workshop - but the house stuff is going well. Plus it is further off than I thought compared to last year.
[00:47:36] <alex_joni> don't remind me :(
[00:47:41] <cradek> yeah it's late this time
[00:47:57] <skunkworks> ;) I have stopped bugging you
[00:52:40] <skunkworks> now it is for sale http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=160087251047&ru=http://search.ebay.com:80/160087251047_W0QQfkrZ1QQfromZR8QQsubmitsearchZSearchQQfviZ1
[00:53:34] <skunkworks> rigid tapping is all I can think of at the moment.
[00:54:28] <skunkworks> more read-ahead would be cool - but it works fine for me the way it is.
[00:56:56] <cradek> yeah tapping would be nice. if I had hardware to use it, I probably would have done it already
[00:57:06] <cradek> kind of boring/tedious otherwise
[00:57:23] <alex_joni> rigid boring?
[00:58:03] <cradek> boring rigid tapping
[00:58:19] <alex_joni> ahh. that kind
[01:55:42] <jepler> jmkasunich: for incremental rebuilds there's also no need to re-run configure -- that is done automatically if configure or a .in file has changed
[02:05:47] <jmkasunich> jepler: good point, I will fix that
[02:06:09] <jepler> jmkasunich: good evening
[02:06:36] <jepler> jmkasunich: if you are staying for awhile, I'd appreciate if you read back to the discussion between me and SWP about specifying durations in HAL
[02:06:54] <jmkasunich> I saw that
[02:07:17] <jmkasunich> these are additions to hal_lib that comps would then make use of?
[02:07:34] <jmkasunich> and the comps would export both seconds and nanoseconds fields?
[02:07:42] <jepler> yes, when it's appropriate
[02:08:02] <jepler> for stuff like oneshot where a long "pulse" may make sense
[02:08:03] <jmkasunich> right - like steplen, stepspace, dirsetup, and dirhold wouldn't use the seconds part
[02:08:26] <jepler> it's much harder to imagine needing a >4 second stepspace
[02:08:30] <jmkasunich> heh
[02:08:48] <jmkasunich> would ns values greater than 1e9 be accepted?
[02:10:08] <jepler> if the times aren't always "normalized" it makes the operations on them a bit harder.
[02:10:17] <jepler> I'm not sure if worrying about that is premature optimization or not
[02:54:08] <cradek> jepler: finally got off my butt and fixed the duplicate message-id problem in the commit emails
[03:01:54] <jepler> whee, hooray
[03:02:06] <cradek> jepler: and as I suspected, the velocity changes are caused by mismatched accels
[03:10:05] <cradek> jepler: http://timeguy.com/cradek-files/emc/biarc.py.png
[03:10:27] <cradek> ^^ with the acute check disabled
[03:10:58] <jepler> is the acute check wrong somehow?
[03:11:21] <cradek> no, but it causes (at least) the first and last segments to have differen accels from the rest
[03:12:20] <cradek> the first wouldn't have to, except when #1/#2 is planned, #3 isn't in the queue yet, so #2 gets marked acute
[03:13:14] <cradek> so you get one glitch there, and another at the end (segments -2 and -1, there is no third again)
[03:14:05] <cradek> I think there's no fix except always assuming acute (or not acute, which is not safe)
[03:14:16] <cradek> the upside is it only happens at the very beginning and very end
[03:20:54] <cradek> jepler: it could be fixed if a segment could have different initial & final accels - but I'd be scared I wouldn't get that right
[03:21:35] <cradek> you could have an acute join that doesn't affect the nearby non-acutes
[03:22:44] <cradek> * cradek talks to himself
[03:28:46] <jepler> cradek: sorry, distracted over here
[03:28:49] <jepler> and leaving for the night now
[03:28:56] <jepler> I'm glad you understand what's going on at least
[03:28:59] <cradek> no problem, goodnight
[03:29:09] <jepler> even if it's not an easy fix
[03:29:21] <cradek> I don't think I'm going to do anything
[03:29:25] <jepler> fine by me
[13:46:51] <jepler> putting the polish on this spline code is proving harder than I'd expected
[13:48:49] <alex_joni> polish?
[13:48:55] <SWPadnos_> SWPadnos_ is now known as SWPadnos
[13:49:16] <SWPadnos> shininess, not people from Poland
[13:49:34] <alex_joni> ah, ok
[13:49:40] <alex_joni> was thinking about polish notation :D
[13:49:43] <SWPadnos> heh
[13:49:46] <alex_joni> and it didn't make any sense
[13:50:09] <SWPadnos> the only difference between the two words is capitalization :)
[13:50:17] <skunkworks> I was just going to ask that.
[13:50:18] <SWPadnos> well, that and their meanings :)
[14:40:22] <skunkworks> wasn't there a lathe wiki - I found the advanced features but I thought there was a page showing how to currently program threading.
[14:40:49] <jepler> skunkworks: did you try the docs?
[14:40:54] <jepler> the pdf docs, that is
[14:41:50] <skunkworks> oops - thanks
[14:52:56] <cradek> wonder if our -sim package would work on edgy
[14:53:00] <jepler> no
[14:53:05] <cradek> ok
[14:53:53] <jepler> I have an edgy simulator machine, so I could take the time to make the -sim package work
[14:54:09] <jepler> they use different package names for some of the requirements
[14:54:49] <cradek> I don't see spending any time on it - if it just worked, that would have been nice.
[15:16:29] <cradek> jepler: have you tested a spline with all colinear points? seems you might get degenerate arcs
[15:17:30] <jepler> cradek: there are a number of degenerate cases that aren't handled, I'm sure that's one of them
[16:12:16] <SWPadnos> gah - alex, can you configure joomla to not open links in new windows?
[16:12:27] <SWPadnos> things like the manual, site search, etc.
[16:16:38] <alex_joni> SWPadnos: those are specified on each link
[16:16:45] <alex_joni> but yes, one could do that
[16:16:48] <SWPadnos> bummer
[16:17:08] <SWPadnos> I can just middle-click I guess (though that doesn't work with the search button)
[16:17:21] <SWPadnos> it seems random though
[16:18:22] <alex_joni> I set them based on how I felt :)
[16:18:29] <SWPadnos> heh
[16:18:31] <alex_joni> guess I worked more than one day on it :P
[16:34:52] <cradek> jepler: about degeneracy: I'll help test when you're ready
[16:37:35] <jepler> OK
[20:18:47] <mjoyce_> jepler, i am looking to get some custom sheetmetal stamped for the front of a 2u computer chassis
[20:19:20] <mjoyce_> do you have any reccomendations on a sheet metal shop that i could work with? hopefully on the west coast
[20:19:36] <jepler> nope, I don't.