#linuxcnc-devel | Logs for 2015-07-22

[07:40:01] <jepler> I suppose you're best off putting the read-request that you expect to be largest in bytes, first
[07:41:37] <jepler> but yes, to get the benefit of overlapping the turnaround time for packets, you have to use the read-request functions
[07:41:54] <jepler> I didn't come up with a clever way for linuxcnc to do it automatically
[10:02:55] <pcw_home> Manual is fine, I'll have to play and see which is best, it does seem like the biggest first would be best but I'm not sure
[10:34:59] <KGB-linuxcnc> 03John Thornton 052.7 9310bc2 06linuxcnc 10(19 files in 4 dirs) Docs: rename spanish docs to follow lower case dash convention * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=9310bc2
[10:35:00] <KGB-linuxcnc> 03John Thornton 052.7 ad4e984 06linuxcnc 10(68 files in 11 dirs) Docs: delete out of date untranslated spanish docs * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=ad4e984
[10:35:00] <KGB-linuxcnc> 03John Thornton 052.7 603eae8 06linuxcnc 10docs/src/Submakefile 03docs/src/gui/gmoccapy-hu.txt 04docs/src/gui/gmoccapy_hu.txt 10docs/src/index.tmpl Docs: add hungarian translated document to build system * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=603eae8
[10:35:18] <jthornton> 3 things left on the list
[10:36:47] <seb_kuzminsky> cool
[10:49:19] <seb_kuzminsky> jthornton: francis tisserant (the french translations guy) and i have been experimenting with po4a for managing the docs translations, it's in the (very old now) branch called seb/master/po4a, do you have any input on that?
[12:58:02] <andypugh> carousel_demo apparently gives a “duplicate O-word label” error on the second toolchange, then never again. There isn’t one. Any ideas?
[13:01:49] <archivist> I wondered if he was using it with some of his own gcode
[13:04:36] <andypugh> No, I see the same issue, but put it down to asking for a tool that didn’t exist
[13:09:24] <andypugh> But it still happens (once-only) when the tool exists.
[13:09:39] <andypugh> I suspect it’s a subtle remap bug :-(
[13:10:24] <andypugh> Unfortunately it isn’t harmless, it aborts the tool chinge with the head down.
[13:19:46] <seb_kuzminsky> yuck
[13:20:00] <seb_kuzminsky> i suspect there are plenty of subtle remap bugs lurking
[13:20:51] <jepler> it's unfortunate we didn't test remap harder before it landed in our master branch
[13:21:12] <andypugh> I need to head off for dinner with my coleagues now, but I would appreciate it if some other folk could try to reproduce it. Open the config, home machine, then MDI T1 M6 / T2 M6
[13:21:15] <jepler> wasn't it one of mah's projects? You might go diving in their git to see if they have fixed any bugs in it
[13:22:27] <andypugh> Hmm, I think I burned those bridges.
[13:22:59] <cradek> I pull their master every so often. In all this time I've seen about two fixes to the core as we know it
[13:25:53] <cradek> just pulled again, I don't see anything
[13:26:17] <cradek> there may be work in other branches I don't see
[13:26:25] <andypugh> OK, back later, I will scroll back.
[13:45:51] <jthornton> seb_kuzminsky, I don't see that branch on git.linuxcnc, after I'm finished building I'll do a git branch -r
[14:31:29] <PCW> With the read-request in hal, running 2 Ethernet cards at 3KHz through the switch!
[14:31:47] <jepler> PCW: excellent
[14:32:32] <PCW> still has the startup issue (doest start at 3 KHz every time)
[14:33:07] <jepler> humbug.
[14:33:21] <skunkworks> In trajectory planner testing.. I found another bug in a 2 axis only config. (acc violation of a 1/3rd or so) so - almost
[14:33:25] <jepler> fails to start, or appears to start but gives following error?
[14:34:12] <PCW> failed to start once ( I dId not use your latest startup patchea so its expected)
[14:35:08] <PCW> almost cuts the servo thread total time in 1/2
[14:35:19] <PCW> (to use read-request)
[14:35:54] <PCW> ~600k clocks vs 980K
[14:38:37] <jepler> that is nice
[14:42:37] <PCW> as expected the second card read time is almost 0
[14:45:05] <PCW> so, Neat!
[14:55:50] <PCW> I'll try setting up 4 cards on the H97 when I get a chance
[15:06:57] <seb_kuzminsky> skunkworks: keep em coming!
[15:07:27] <skunkworks> rob has been squashinng them quite quickly. I think that is the last one. (until we find more...)
[15:09:01] <seb_kuzminsky> oh good, it's always nice when you find the last bug
[15:10:51] <jepler> every time
[15:10:57] <jepler> .. I keep feeling like that wrt hm2_eth
[15:11:41] <skunkworks> to bad it isn't like finding your car keys - always the last place you look...
[15:27:37] <mozmck> problem I often have is, I find a bug, fix it, and the fix creates 3 new ones :)
[16:06:30] <KGB-linuxcnc> 03John Thornton 052.7 0dbc60c 06linuxcnc 10docs/src/gui/gladevcp.txt 10docs/src/gui/gladevcp_fr.txt 04docs/src/gui/images/emc-actions.png 03docs/src/gui/images/vcp-actions.png Docs: rebranding update * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=0dbc60c
[17:15:27] <andypugh> OK, so this only remained connected for 10 mins after I left.
[17:15:48] <andypugh> Did I miss anything in particular?
[17:21:59] <Tom_itx> zlog
[20:52:48] <cmorley> andypugh: I emailed MAH about another tool remap problem - he felt that the queue buster commands were probably the biggest problem...is the remap with python or gcode commands? he recommended calling emccanon commands directly....
[20:55:47] <jepler> the next interesting thing about store ordering on arm is that apparently my hal_stream rewrite made it much easier to trigger in sampler, it takes 2-4x as long to trigger it in 2.7 branch
[20:59:18] <PCW> probably a good thing to make random bugs more frequent
[21:05:03] <PCW> two card hm2_eth setup has been running all afternoon at 3KHz
[21:05:05] <PCW> I'll let it run a few days then try 4 cards
[21:12:54] <jepler> .. 46 million iterations with my newly re-fixed sampler code and no glitches
[21:57:54] <pcw_home> Thats the ARM problem?
[21:59:44] <jepler> right
[22:10:34] <cradek> newly refixed differently?
[22:13:57] <jepler> yeah, two differences. first, it's not based on the hal_stream branch; and second of all I took the advice to use C11-style atomics
[22:16:29] <cradek> oh cool, someone on the list was helpful
[22:16:33] <KGB-linuxcnc> 03Jeff Epler 05jepler/arm-atomic 5148b15 06linuxcnc 03docs/man/man3/rtapi_atomic.3rtapi 03src/rtapi/rtapi_atomic.h rtapi: implement subset of <stdatomic.h> * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=5148b15
[22:16:33] <KGB-linuxcnc> 03Jeff Epler 05jepler/arm-atomic 2e3d3c0 06linuxcnc 10src/hal/components/sampler.c 10src/hal/components/sampler_usr.c 10src/hal/components/streamer.c 10src/hal/components/streamer_usr.c streamer, sampler: fix memory-order bugs * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=2e3d3c0
[22:22:50] <jepler> whee
[23:57:26] <seb_kuzminsky> woo