Back
[01:32:07] <linuxcnc-build_> build #3301 of 1200.rip-lucid-i386 is complete: Failure [4failed compile runtests] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1200.rip-lucid-i386/builds/3301 blamelist: dummy, Jeff Epler <jepler@unpythonic.net>
[02:08:14] <linuxcnc-build_> build #3311 of 0000.checkin is complete: Failure [4failed] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/3311 blamelist: dummy, Jeff Epler <jepler@unpythonic.net>
[07:25:59] <jepler> /home/buildslave/emc2-buildbot/precise-i386/precise-i386-sim/build/tests/t0/random-without-t0
[07:26:05] <jepler> /home/buildslave/emc2-buildbot/lucid-i386/lucid-i386-sim/build/tests/t0/nonrandom
[07:27:04] <jepler> are these known-flaky tests or did I actually break something? particularly the 2.7-branch failure seems really unlikely to be my code
[08:59:38] <seb_kuzminsky> jepler: that's a known-flaky test
[08:59:55] <seb_kuzminsky> i should say: it's a test that's known to fail occasionally for unknown reasons
[09:18:39] <jepler> seb_kuzminsky: OK
[09:19:41] <jepler> seb_kuzminsky: I think my revised arm-atomic branch is in good shape, but if I'm lucky I'll get review notes from Chris Lesiak
[09:20:00] <jepler> I should probably write at least one more e-mail showing what the object code difference is on x86
[09:33:19] <seb_kuzminsky> i'm going to become rich and famous when i invent a revision control system that's compatible with "git", but called "gi", and all its sub-commands have the same name as git's sub-commands but prefixed with "t"
[10:08:13] <jepler> gi () { f="$1"; shift; git "${f#t}" "$@"; }
[10:08:18] <jepler> $ gi tlog -1 --oneline
[10:08:18] <jepler> 475f8e4 streamer, sampler: fix memory-order bugs
[10:08:25] <seb_kuzminsky> darn he beat me to it
[10:15:54] <jepler> I'm one of those "10x programmers" doncha know
[10:38:58] <pcw_home> ~1.5 days at 4 KHz still 150 usec max servo thread time on C2D 3.16 GHz
[10:41:00] <pcw_home> not sure if this is due to linux 4.1.3-rt3 or hm2_eth startup improvements
[10:48:54] * seb_kuzminsky looks at 4.1.3-rt3 with hunger
[11:12:11] <pcw_home> nice thing is that preemt-rt is trivial to build
[11:30:54] <jepler> an -rt kernel is in debian testing at the moment too
https://packages.debian.org/stretch/linux-image-4.0.0-2-rt-amd64
[11:31:42] <jepler> -rt-amd64 and -rt-686, no arm..
[11:32:02] <jepler> but I don't have any machines that armmp works on anyway :-/
https://wiki.debian.org/DebianKernel/ARMMP
[17:48:09] <JT-Shop> seb always knows where the cool stuff is
[17:55:06] <seb_kuzminsky> my grep fu is strong
[17:57:55] <seb_kuzminsky> our curated set of example configs is pretty comprehensive and cool
[17:58:50] <seb_kuzminsky> and someone like frederic rible can easily digest it and synthesize what they need
[18:00:10] <andypugh> seb_kuzminsky: Do we need to find the bug in the carousel remap before that goes in?
[18:01:15] <andypugh> (I would say that as the bug isn’t in the example config, it could go in, as it doesn’t actually _introduce_ a bug, it just demonstrates it)
[18:03:12] <seb_kuzminsky> i see your point
[18:03:27] <seb_kuzminsky> but i also would love to not publicize/advertise stuff we know is broken
[18:06:18] <jepler> hm the objdumps are too different to directly compare
[18:06:26] <seb_kuzminsky> bummer
[18:06:36] <seb_kuzminsky> andypugh: did your debugging efforts of the bug go anywhere?
[18:07:35] <andypugh> No, because I managed to reformat my HDD and lose all my backups, so I am currently not even close to having a LinuxCNC system to work with.
[18:07:58] <jepler> + c4: e8df f002 tbb [pc, r2]
[18:07:58] <jepler> + c8: 15150209 .word 0x15150209
[18:08:00] <jepler> wow arm is crazy
[18:08:35] <jepler> tbb is "table branch byte". it uses r2 to look up the distance to jump from that .word directly following the tbb instruction
[18:09:02] <seb_kuzminsky> .word
[18:09:41] <jepler> so it can jump to 2*0x15, 2*2, or 2*9 bytes ahead
[18:16:23] <andypugh> jepler: That sounds surprisingly like the Royal McBee LGP-30
[18:22:23] <jepler> there's also an "IT" instruction
[18:22:29] <jepler> .. I haven't read much ARM disassembly yet
[18:23:42] <jepler> standing for "if-then", it makes the next 1..4 instructions after it conditional on one conditional or its inverse. So e.g., ITE lt executes pc+2 if the "lt" condition is set, and pc+4 if it is not set, then continues at pc+6
[18:24:48] <jepler> anyway, mostly the change on ARM is to add "dmb sy" instructions in a few places, which is exactly what is expected
[18:51:09] <KGB-linuxcnc> 03Dewey Garrett 05dgarr/hallib_update 16dfaad 06linuxcnc 10lib/hallib/hal_procs_lib.tcl hal_procs_lib.tcl: proc for finding halfile * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=16dfaad
[18:51:09] <KGB-linuxcnc> 03Dewey Garrett 05dgarr/hallib_update 70c89f5 06linuxcnc 10tcl/bin/emccalib.tcl emccalib.tcl: find halfiles using hallib proc * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=70c89f5
[18:51:10] <KGB-linuxcnc> 03Dewey Garrett 05dgarr/hallib_update 6ff902f 06linuxcnc 10lib/hallib/xhc-hb04.tcl xhc-hb04: use hallib proc find_file_in_hallib_path * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=6ff902f
[18:51:11] <KGB-linuxcnc> 03Dewey Garrett 05dgarr/hallib_update bd4101d 06linuxcnc 10tcl/twopass.tcl twopass.tcl: use proc find_file_in_hallib_path * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=bd4101d
[18:51:15] <KGB-linuxcnc> 03Dewey Garrett 05dgarr/hallib_update f792e0c 06linuxcnc 10scripts/linuxcnc.in linuxcnc.in: support absolute path for halfiles * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=f792e0c
[18:51:19] <KGB-linuxcnc> 03Dewey Garrett 05dgarr/hallib_update 977a2d9 06linuxcnc 10scripts/hal-histogram 10scripts/moveoff_gui scripts/moveoff_gui,hal-histogram: add note * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=977a2d9
[19:15:26] <linuxcnc-build_> build #3303 of 1306.rip-precise-amd64 is complete: Failure [4failed runtests] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/1306.rip-precise-amd64/builds/3303 blamelist: Dewey Garrett <dgarrett@panix.com>
[19:28:58] <linuxcnc-build_> build #3312 of 0000.checkin is complete: Failure [4failed] Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/3312 blamelist: Dewey Garrett <dgarrett@panix.com>
[19:54:32] <jepler> andypugh: these days, reading about computers like Royal McBee LGP-30 feels like reading things that are made up
[19:54:51] <jepler> [to reduce cost, hexadecimal was represented as] 0 - 9 f g j k q w, which was remembered using the phrase "FiberGlass Javelins Kill Quite Well"
[19:56:07] <andypugh> I wonder how that reduced cost?
[19:56:30] <andypugh> I was actualy referring to “the story of Mel” which you probably already found?
[20:01:58] <jepler> ah, gotcha
[20:03:21] <andypugh> But, anyway, the LGP-30 and LGP4000 had a jump option on every instruction.
[20:03:54] <andypugh> And the story is about mis-use of that
[20:22:00] <seb_kuzminsky> well that's a new and surprising failure
[20:23:02] <jepler> Running test: /home/buildslave/emc2-buildbot/precise-amd64/precise-amd64-sim/build/tests/and-or-not-mux.0
[20:23:05] <jepler> halrun: Realtime already running. Use 'halrun -U' to stop existing realtime session.
[20:23:09] <jepler> *** /home/buildslave/emc2-buildbot/precise-amd64/precise-amd64-sim/build/tests/and-or-not-mux.0: FAIL: test run exited with 1
[20:23:12] <jepler> looks like leftovers
[20:24:05] <seb_kuzminsky> but from what? the test before succeeded
[20:24:41] <seb_kuzminsky> is there a shutdown race in uspace?
[20:26:48] <seb_kuzminsky> linuxcnc-build_: force build --branch=dgarr/hallib_update 0000.checkin
[20:26:49] <linuxcnc-build_> build forced [ETA 53m36s]
[20:26:49] <linuxcnc-build_> I'll give a shout when the build finishes
[20:30:39] <seb_kuzminsky> this looks useful:
http://hackaday.com/2015/07/31/waking-up-is-easy-with-this-bed-ejection-system/
[20:40:31] <andypugh> I am not sure I wanted to see Colin Furze’s pants!
[20:41:46] <andypugh> (even though I am a somthing of a fan of his style of engineering and presentation)
[21:17:08] <linuxcnc-build_> Hey! build 0000.checkin #3313 is complete: Success [3build successful]
[21:17:08] <linuxcnc-build_> Build details are at
http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/3313