#emc-devel | Logs for 2009-04-16

[00:23:08] <cradek> I am surprised to learn that any system is smarter than the basic merge
[00:44:23] <cradek> "A branch in darcs is a repository (repo). A repo can contain many tags, but only one branch."
[00:45:00] <cradek> I don't like this scheme.
[00:46:37] <cradek> http://lists.osuosl.org/pipermail/darcs-users/2004-July/002356.html
[00:47:01] <cradek> #1 here is why I don't like it - this work of keeping track of branches should not be left to the user
[00:47:32] <jmkasunich> when did darcs get added to the confusion?
[00:49:33] <cradek> 17:49:07 CDT
[00:50:04] <cradek> but for me, it's out of the running already
[00:50:14] <jmkasunich> good
[00:50:19] <jmkasunich> that way I don't have to think about it
[00:50:37] <cradek> heh, maybe you trust my kneejerk reactions too much.
[00:50:59] <jmkasunich> nah, I trust my own knee-jerk reactions
[00:51:05] <jmkasunich> at least I've heard of git and svn
[00:51:07] <cradek> btw, I did decide to come
[00:51:14] <jmkasunich> cool
[00:51:28] <jmkasunich> speaking of trips - did you decide if you are taking the bus to wichita?
[00:51:41] <cradek> yes I am pretty sure I will
[00:52:04] <jmkasunich> I'm trying to decide between KC and rent a car vs wichita
[00:52:13] <jmkasunich> I suppose pricing the rental would be a good data point
[00:52:50] <cradek> seems like they are always $20-25/day for a week, for a crappy little car
[00:53:25] <jmkasunich> cheapeat at budget is 184 for the week
[00:53:54] <cradek> huh, more like 30 then once you pay the extra BS fees
[00:54:02] <jmkasunich> strange - "standard" is cheaper than "economy" or "compact"
[00:54:15] <cradek> bizarre
[00:54:25] <cradek> people are still worried about gas I guess
[00:54:47] <jmkasunich> heh, then they try to sell you GPS ($50 for the week) and LDW ($24/day)
[00:55:08] <cradek> what's LDW?
[00:55:15] <jmkasunich> loss damage waiver
[00:55:21] <cradek> oh
[00:55:26] <cradek> bogus insurance
[00:55:40] <jmkasunich> yeah
[00:55:53] <jmkasunich> I realised I made a mistake - said I wanted it at wichita
[00:56:08] <jmkasunich> if I get it in KC, the cheapest is 238 for the week
[00:56:20] <cradek> hm
[00:56:30] <jmkasunich> and economy and compact are both cheaper than standard
[00:56:36] <cradek> you could get a better price with priceline if you don't care what car or company you get
[00:57:24] <cradek> maybe you could convince me or jepler that KC's not far out of the way (it worked once before...)
[00:57:24] <jmkasunich> eww - priceline makes you look at shatner
[00:58:01] <jmkasunich> best is alamo, still $189 for the week
[00:58:14] <jmkasunich> ftw
[00:58:25] <jmkasunich> 189/week, then in much smaller print $300 total
[00:58:38] <cradek> uh?
[00:58:43] <jmkasunich> total includes taxes and fees - but $111 in taxes and fees?
[00:58:57] <cradek> that's even more BS fee than I expected
[00:59:35] <jmkasunich> RANSPORTATION FACILITIES CHARGE $12.00
[00:59:36] <jmkasunich> CUSTOMER FACILITY CHARGE 3.00/DAY $18.00
[00:59:36] <jmkasunich> CONCESSION RECOUP FEE 11.11 PCT $22.42
[00:59:36] <jmkasunich> DOWNTOWN ARENA FEE 4.00/DAY $24.00
[00:59:36] <jmkasunich> VEH LICENSE FEE RECOVERY 1.99/DAY $11.94
[00:59:36] <jmkasunich> STATE TAX $22.19
[00:59:57] <cradek> haha
[01:00:19] <cradek> well you don't expect to get your CONCESSION RECOUPs for free do you?
[01:01:35] <cradek> it would sure be nice if not everything was a scam.
[01:01:52] <jmkasunich> keep dreaming
[01:02:57] <jmkasunich> is jeff driving to wichita?
[01:03:38] <jmkasunich> trying to decide if having a car during the week would be a valuable thing...
[01:04:01] <cradek> yes he will drive separately if I take the bus.
[01:04:13] <jmkasunich> so thats at least one car
[01:04:30] <cradek> he would strongly prefer to have his own car though, so he can have a different schedule.
[01:04:43] <cradek> so I think at least one of the rest of you ought to have a car
[01:05:01] <jmkasunich> yeah, one is nowhere near enough
[01:05:08] <jmkasunich> two min, 3 better
[01:05:35] <jmkasunich> jon e is coming, and seb, and swp (maybe, part of the time), and who else?
[01:05:52] <cradek> johnt? rayh?
[01:06:08] <cradek> I haven't heard from rayh for a while.
[01:06:19] <jmkasunich> nope
[01:06:25] <jmkasunich> hope all is well up north
[01:06:35] <cradek> yeah, me too.
[01:08:25] <jmkasunich> sockets are very handy as press accessories
[01:08:41] <jmkasunich> no matter what size the thing you are pushing happens to be, theres probably a socket just a hair bigger
[01:08:50] <jmkasunich> to use as an anvil
[01:09:13] <cradek> yep, I do that all the time too
[01:09:34] <jmkasunich> now I need to rig up a stop at the right depth
[01:13:44] <jmkasunich> I love shrink fits - so much more civilised than force fits
[01:14:51] <cradek> fun - what is it?
[01:16:09] <jmkasunich> sleeve onto spindle (provides shoulder for front bearing)
[01:16:14] <jmkasunich> then pulley onto sleeve
[01:17:07] <jmkasunich> looks like the pulley-sleeve push stopped about 0.030 sooner than I wanted (but acceptable)
[01:18:00] <cradek> I don't like that one-chance stuff
[01:18:41] <jmkasunich> I'm not crazy about it either
[01:18:52] <jmkasunich> but there isn't really a better way to do this kind of thing
[01:18:59] <jmkasunich> strong, concentric, compact
[01:22:58] <jmkasunich> I should have made my nut out of bronze or brass
[01:23:15] <jmkasunich> drill rod nut on drill rod spindle wants to gall
[01:23:27] <jmkasunich> I made the threads quite snug
[01:24:47] <cradek> tiniest bit of anti-seize?
[01:26:37] <jmkasunich> maybe
[01:27:03] <jmkasunich> I think pressing on the back of the spindle distorted the first thread or two
[01:27:15] <cradek> dang
[01:27:16] <jmkasunich> either enlarged the diameter or compressed the pitch, can't tell
[01:27:34] <jmkasunich> with oil and a wrench I can put the nut on
[01:27:50] <cradek> you need one of those split dies
[01:29:39] <jmkasunich> yeah
[01:29:51] <jmkasunich> but 10mm dia x 1mm pitch isn't the most common thread ever
[01:30:13] <jmkasunich> sort of like NF for inch threads - NC (10mm x 1.5) is far more common
[01:30:39] <cradek> bbl
[01:34:24] <jepler> cradek: I didn't know there had been any doubt you were going to wichita
[01:54:49] <jepler> git merge does indeed do what the darcs guy says on that example. git-rebase gives one of two different results, depending what is rebased on what (if you rebase the two commits on top of the one commit, it "replays" them both as separate commits, which is apparently what darcs does by default).
[03:15:15] <cradek> jepler: that's not it - I'm going to cleveland/toledo/NAMES this weekend
[03:38:28] <CIA-2> EMC: 03cmorley 07TRUNK * 10emc2/src/emc/usr_intf/pncconf/ (pncconf.py pncconf.glade):
[03:38:28] <CIA-2> EMC: Redo naming of internal parport pins to simplify convertion to signalname.
[03:38:28] <CIA-2> EMC: fix parport so pins 2-9 can change be changed from input to output instead
[03:38:28] <CIA-2> EMC: of 1-8. Fix the method that converts internal data pins to HAL phrases
[03:38:28] <CIA-2> EMC: to now include all parports. Need to finish this to put proper endings on
[03:38:31] <CIA-2> EMC: The HAL phrases such as -out or -in for parport pins.
[05:40:20] <garage_seb> cradek, are you around?
[06:19:17] <garage_seb> it looks to me like the suggestion of not limiting stepgen's vel or accel is working
[06:19:29] <garage_seb> my notes here: http://highlab.com/~seb/emc2/stepgen/
[06:43:33] <CIA-2> EMC: 03seb 07TRUNK * 10emc2/src/hal/drivers/mesa-hostmot2/stepgen.c: allow maxaccel=0, meaning no accel limit
[06:43:49] <CIA-2> EMC: 03seb 07TRUNK * 10emc2/docs/man/man9/hostmot2.9: document new maxaccel=0 behavior
[06:45:50] <CIA-2> EMC: 03seb 07v2_3_branch * 10emc2/src/hal/drivers/mesa-hostmot2/stepgen.c: allow maxaccel=0, meaning no accel limit
[06:46:03] <CIA-2> EMC: 03seb 07v2_3_branch * 10emc2/docs/man/man9/hostmot2.9: document new maxaccel=0 behavior
[06:48:28] <micges> good morning
[07:13:00] <alex_joni> seb_kuzminsky: thanks
[07:13:09] <alex_joni> I'll let you know how it works
[07:21:16] <seb_kuzminsky> hi alex_joni
[07:21:57] <seb_kuzminsky> goodnight alex_joni
[07:24:37] <alex_joni> see you seb, don't stay up so late ;)
[07:31:10] <SWPadnos_> SWPadnos_ is now known as SWPadnos
[07:40:03] <SWPadnos_> SWPadnos_ is now known as SWPadnos
[07:47:28] <SWPadnos_> SWPadnos_ is now known as SWPadnos
[12:21:02] <skunkworks_> does the message from zhangj make sense?
[12:24:05] <jepler> I kind-of understand what he's trying to ask (he tried to derive one of the formulas in the source code but got something else)
[12:24:14] <jepler> but I don't know whether he's right or wrong
[12:24:30] <jepler> and I didn't read the scrollback in #emc where they were discussing this earlier, either
[12:27:04] <skunkworks_> oh - interesting. I should read that.
[12:42:08] <alex_joni> I didn't quite understand what he was getting at
[12:42:24] <alex_joni> I mean.. how he got there
[12:52:39] <jepler> a little voice in my head says that 2* has something to do with the discrete approximation of the integral
[12:52:45] <jepler> but .. I don't know the details
[12:54:05] <jepler> one question to ask: if you change it like you propose, how does emc work? better? the same? worse?
[12:59:10] <jepler> no, you get 0.5 from the integral I was thinking of
[13:07:51] <cradek> jepler: that's the question I'd ask first, too
[16:10:11] <dgarr> Re: emc.so -- i get a segv in my application for "package require" on some packages: http://pastebin.ca/1394025
[16:13:44] <jepler> dgarr: digesting
[16:17:03] <jepler> this behavior is new since the introduction of the tcl/emc.so?
[16:17:50] <dgarr> i believe so but I had not done a checkout in a week or so -- i could be mistaken
[16:24:12] <jepler> good (?) news, I get the segfault too
[16:24:48] <dgarr> using my example or another way?
[16:26:15] <jepler> using your example
[16:29:35] <jepler> the axis one
[16:30:37] <jepler> so tkemc didn't work at all for you? it worked for me..
[16:31:05] <CIA-2> EMC: 03jepler 07TRUNK * 10emc2/tcl/bin/emctesting.tcl: fix incorrect path to emc.tcl
[16:31:47] <dgarr> tkemc worked for me " tkemc does not segv when running"
[16:31:58] <dgarr> 'package require'
[16:32:17] <jepler> you mentioned a fix to emctesting.tcl
[16:32:43] <dgarr> yes that was just to get it to run, nor related to "package require" (i think)
[16:33:03] <jepler> odd, tkemc runs for me without that change
[16:33:22] <jepler> so all the scenarios you ran into involve axis + "package require"?
[16:33:41] <dgarr> correct
[16:34:22] <dgarr> but only for some packages: segv for Hello, thread,Img,... nosegv for msgcat,BWidget,...
[16:34:39] <jepler> I think msgcat and bwidget may not have any parts that are shared libraries
[16:35:04] <dgarr> i agree -- seems to always be _Init in a .so
[16:38:34] <SWPadnos> jepler, one question. it looks like your last commit just removes that line, it doesn't seem to "correct" it ...
[16:38:46] <jepler> SWPadnos: look further up the file, above the context of the diff
[16:38:52] <SWPadnos> ok
[16:39:09] <SWPadnos> I'm (obviously) only looking at the diff email - thanks
[16:44:01] <jepler> and now I have a much-reduced test case
[16:46:07] <jepler> http://emergent.unpy.net/files/sandbox/package-require-crash.py
[16:46:52] <SWPadnos> that's pretty small :)
[16:47:00] <jepler> I'm sure I'll get it chased down soon
[16:47:34] <dgarr> impressive reduction
[16:49:12] <jepler> executing those same two commands in wish doesn't crash, of course
[17:05:19] <seb_kuzminsky> jmkasunich: did you see the stepgen change cradek suggested?
[18:05:20] <CIA-2> EMC: 03jepler 07TRUNK * 10emc2/src/configure.in:
[18:05:20] <CIA-2> EMC: Fix for the crashing bug reported by dgarr
[18:05:20] <CIA-2> EMC: apparently there's a bad interaction in Python's Tkinter when stubs and
[18:05:20] <CIA-2> EMC: non-stubs Tcl extensions are combined (on at least dapper and hardy).
[18:05:20] <CIA-2> EMC: To fix this, use a stubs-type Tcl extension. Stubs doesn't work before Tcl
[18:05:20] <CIA-2> EMC: 8.1, so remove some configure support for that, but 8.1 about ten years old,
[18:05:24] <CIA-2> EMC: so I can't imagine people still using 8.0.
[18:05:26] <CIA-2> EMC: 03jepler 07TRUNK * 10emc2/src/emc/usr_intf/emcsh.cc:
[18:05:28] <CIA-2> EMC: Fix for the crashing bug reported by dgarr
[18:05:30] <CIA-2> EMC: apparently there's a bad interaction in Python's Tkinter when stubs and
[18:05:32] <CIA-2> EMC: non-stubs Tcl extensions are combined (on at least dapper and hardy).
[18:05:34] <CIA-2> EMC: To fix this, use a stubs-type Tcl extension. Stubs doesn't work before Tcl
[18:05:36] <CIA-2> EMC: 8.1, so remove some configure support for that, but 8.1 about ten years old,
[18:05:38] <CIA-2> EMC: so I can't imagine people still using 8.0.
[18:07:35] <jepler> dgarr: please try that and let me know
[18:08:58] <CIA-2> EMC: 03jepler 07TRUNK * 10emc2/src/configure: regenerate
[18:12:33] <jepler> "With the 8.1 release, Tcl became the first scripting language to have all the facilities needed for mission-critical enterprise applications. "
[18:13:03] <SWPadnos> like the ability to use real mission-critical code written in other languages?
[18:13:29] <dgarr> jepler: Compiling emc/usr_intf/emcsh.cc
[18:13:29] <dgarr> emc/usr_intf/emcsh.cc:22:17: error: tcl.h: No such file or directory and more
[18:13:29] <dgarr> do i need to make clean or something?
[18:13:48] <SWPadnos> configure changed, you should probably re-configure and make clean
[18:13:52] <jepler> that's unexpected :-P
[18:21:39] <alex_joni> SWPadnos: it should do that by itself
[18:21:45] <SWPadnos> ok
[18:21:46] <alex_joni> of couse if one doesn't cvs up in the proper dir...
[18:22:09] <alex_joni> it doesn't know (yet!) that there are new files in CVS which you don't have :)
[18:30:07] <dgarr> jepler: so i just finished make clean; configure --enable-run-in-place --enable-simulator and get segfault starting axis with confis/sim/axis.ini
[18:30:29] <dgarr> emc: line 634: 11294 Segmentation fault $EMCDISPLAY -ini "$INIFILE" $EMCDISPLAYARGS $EXTRA_ARGS
[18:30:47] <dgarr> this is without "require package stuff"
[18:30:53] <jepler> ugh really?
[18:31:09] <jepler> sigh
[18:31:10] <dgarr> emc tkemc.ini started up though
[18:31:18] <jepler> I guess I didn't test axis...
[18:31:28] <dgarr> maybe my debugging has left some debris locally
[18:32:32] <jepler> I didn't "make clean", either, so I have files built with different cflags than you
[18:33:17] <jepler> I bet I'll get the crash now, after this "make clean"
[18:37:21] <jepler> "our hardware vendor tells us that the Core2 series processors are
[18:37:21] <jepler> reaching end of life"
[18:37:29] <jepler> ugh, CPUs change too fast these days
[18:38:53] <alex_joni> jepler: yup
[18:39:14] <alex_joni> although I think you can still get them at least 2-3 more years
[18:46:28] <jepler> yes, after a clean and rebuild I segfault too
[18:52:43] <CIA-2> EMC: 03jepler 07TRUNK * 10emc2/src/emc/usr_intf/axis/extensions/ (_toglmodule.c seticon.c): .. and that means these 'Tcl/Tk extensions' have to use stubs too
[18:52:45] <jepler> how about now?
[18:52:52] <dgarr> 0xb7975a82 in install (s=0x0, arg=0x835c250)
[18:52:52] <dgarr> at emc/usr_intf/axis/extensions/_toglmodule.c:21
[18:52:52] <dgarr> 21 if (Tcl_PkgPresent(trp, "Togl", TOGL_VERSION, 0)) {
[18:53:25] <dgarr> oops, i need a few minutes to get and test that
[18:55:04] <jepler> sure, no problem
[19:01:05] <alex_joni> dgarr: seen the latest post on the list?
[19:02:06] <alex_joni> it's similar to what you do..
[19:04:21] <seb_kuzminsky> alex_joni: do you think the maxaccel=0 thing is the right thing to do?
[19:06:04] <SWPadnos> hmmm. here's my thought for the day (just before I head out for a meeting :) )
[19:06:49] <SWPadnos> when the TP is calculating the next position, does it take accel into account, or does it assume that the motor will be able to snap to the new speed (which was calculated with accel in mind)?
[19:08:01] <jepler> seb_kuzminsky: if it doesn't seem to hurt following (compared to specifying a few percent headroom), I say go for it in sample configs.
[19:08:21] <jepler> infinite accel in the step generator itself makes lots of sense, and 0 is the sensible way to specify it
[19:08:55] <dgarr> jepler: ok-- the test with tst.py&tst.tcl, axis.ini, my application, with the test .so package, now i get: libhello.so: undefined symbol: Tcl_CreateNamespace
[19:09:10] <jepler> we (that is, stepconf) still specify stepgen accel with headroom based on an old report from skunkworks_ that this gave slightly better behavior than specifying infinite velocity, but I forget the details
[19:09:18] <jepler> dgarr: I think you have to start using stubs too now
[19:09:54] <dgarr> ok -- it was just an example, i'll have to learn about stubs --- thanks for all your help
[19:10:23] <jepler> ok, if the examples in the files I modified today aren't enough for you to get your code working again I'll try to help out
[19:10:59] <jepler> honestly I don't understand this 100% myself either
[19:11:10] <dgarr> my code is all working -- the example for hello was just something i found to test packages in .so files
[19:11:21] <jepler> ah ok
[19:11:35] <jepler> I didn't know if you had a self-written tcl extension .. you have lots of fancy toys
[19:11:56] <dgarr> s/b my code is all working on simulator, i'll wait a while before testing in the garage for real
[19:12:15] <jepler> as always, thanks for the good reports with lots of info
[19:13:43] <alex_joni> seb_kuzminsky: no report yet how it works
[19:13:50] <alex_joni> but I think it would be the best thing to do
[19:13:57] <seb_kuzminsky> ok cool :-)
[19:14:12] <alex_joni> if something is wrong, users can still specify a HM2_STEPGEN_MAX_VEL/ACCEL
[19:14:23] <dgarr> alex_joni: i had communications with the poster about 8 months ago and encouraged him to emc:)
[19:14:24] <seb_kuzminsky> fair enough
[19:14:30] <alex_joni> so it's only a configuration fix, rather than a new version of emc2
[19:14:31] <jepler> did you look at velocity and accel during cruse phases and starting phases?
[19:14:40] <alex_joni> dgarr: cool
[19:14:54] <seb_kuzminsky> jepler: yes, with infinite accel it's a bit bouncy
[19:15:00] <seb_kuzminsky> but mostly good
[20:35:45] <jepler> I'm updating my doc building scripts to build trunk and 2.3, instead of trunk and 2.2
[20:36:23] <jepler> soon I hope to work on making the lyx->html conversion work with the native hardy version of lyx, and we'll just drop pdf documentation building on hardy post-2.3
[20:36:55] <jepler> er,
[20:37:00] <jepler> drop pdf documentation building on dapper post-2.3
[20:37:09] <seb_kuzminsky> had me scared there
[20:37:14] <jepler> I don't think there's any reason to be supporting dapper post-2.3
[20:37:24] <seb_kuzminsky> i agree
[20:44:29] <alex_joni> dappers end of life is June 2009 for desktop and 2011 for server
[20:44:50] <alex_joni> June 2009 is surely unlikely for 2.4
[20:45:07] <jepler> sure is
[21:31:24] <seb_kuzminsky> lol