Guest708 is now known as skunkworks_
EMC: 03seb 07TRUNK * 10emc2/src/hal/drivers/mesa-hostmot2/TODO: some more notes on what to do and how to do it
EMC: 03seb 07TRUNK * 10emc2/src/hal/drivers/mesa-hostmot2/ (11 files):
EMC: Rename all the logging macros, so my WARN() doesn't collide with the
EMC: one added in 2.6.27 (problem found by acemi).
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/drivers/ (m5i20.lyx opto22.lyx pico_ppmc.lyx): clean up
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/examples/spindle.lyx: clean up
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/gui/ (image-to-gcode.lyx tkemc.lyx): clean up
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/install/compiling_emc2.lyx: clean up
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/hal/ (11 files): clean up
is richard acosta actaully asking how to use AC motors and VFDs for axis positioning?
until I saw who it was, I assumed it was a question about how to connect a voltage-to-frequency converter to the 7i43 (as some sort of feedback)
that may have been pre-coffee too
hi Matt. did you fix it?
or maybe I should ask if you borke it ...
It runs, by careful setting of stepgen maxaccel, but It's not really right either.
we still need an update from jmkasunich
I think you said the magic words, "It's all FF", rather than all P
for systems with no physical feedback mechanism
because there will never be a "real" error term
sounds interesting - what's the problem?
so, yes, that's what he's asking about
seems SF is adding git repos
* seb_kuzminsky is reading dave gingery
what are you going to make seb_kuzminsky ?
heh, prolly nothing for a long time
i still have a cnc conversion i should be concentrating on
it's just impressive what he accomplished
you gotta wonder how those machines worked...
how well they worked you mean? how precise they were?
the operation of the machines themselves are pretty well covered in "Build your own metal working shop from scrap"
mshaver: you around? any new info on the hm2 stepgen ferror problem?
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/ladder/classic_ladder.lyx: update images
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/ladder/images/ (15 files): update images
build #694 of hardy-x86-trunk-realtime-deb is complete: Failure [failed compile] Build details are at http://emc2-buildbot.colorado.edu/buildbot/builders/hardy-x86-trunk-realtime-deb/builds/694
../docs/src/Submakefile:388: *** Required image file ../docs/src/ladder/images/Bit_Status.png does not exist. Stop.
build #660 of dapper-x86-trunk-realtime-deb is complete: Failure [failed compile] Build details are at http://emc2-buildbot.colorado.edu/buildbot/builders/dapper-x86-trunk-realtime-deb/builds/660
seb_kuzminsky: looks like it takes a dump every time I replace an image with a new one with a different name
BigJohnT: according to http://cvs.linuxcnc.org/cvs/emc2/docs/src/ladder/images/
I don't see a Bit_Status.png
and it is no longer used in the lyx
and I'm totally messed up :/
time to start drinking
* BigJohnT kicks CIA-2
cvs is slow today
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/ladder/images/Bit_Status.png: do not forget to add the images
I'm glad I deleted crack monkey from the commit log
forgot the cvs add?
one of the three...
BigJohnT: the drinking helped!
except for the vapours
now I need to go purchase my drinking buddy some sharp tools for his 72nd birthday party today
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/gcode/tool_compensation.lyx: add info on touchoff thanks for the excellent example Chris
BigJohnT: some pictures to go with that would be best
working on that :)
seb_kuzminsky: short question
is there a way to make buildbot announce a success after a fail ?
oh crap forgot to cvs add the image
alex_joni: no :-(
i wish there was
seb_kuzminsky: CF does that :D
don't sweat it
it's great to have announcements about broken packaging anyways
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/gcode/images/ToolTable-TouchOff.png: add image first!
heh I think I made under the time limit :)
seb_kuzminsky: does a new commit abort an older compile?
or does it trigger a post-build (finish the current build, then start a new one)
EMC: 03bigjohnt 07TRUNK * 10emc2/docs/src/Master_HAL.lyx: move halshow to integrators manual
the builds stack up
but it waits 2 minutes right?
if you look at the waterfall now, notice hardy-x86-trunk-realtime-rip has 2 builds pending but hardy-x86-trunk-sim has only
* BigJohnT is off to town if no one else wants to test the build failure
that's because there's only one hardy buildslave, and each buildslave is configured to do only one build at a time
speaking of which, if anyone wants to run a buildslave vm, we could get more parallelism in the builds ... <wink>
you have to have an "always on" connection for that right?
it's a kvm/qemu vm, all set up, all it needs is a linux host to run on and the ability to make outgoing tcp connections
always on internet is required, yes
I'm on a string
and two tin cans
bitrate's no problem ;-)
but hogging the line is not liked by my ISP
or my wife
./me has got to get that tower up
talk to you guys later
hey seb_kuzminsky I could do it at my machine shop when I get the linux box going over there...
we have cable internet there
it really doesnt take much resources to set up or to run
ok later :-)
same here.. back in 1-2 hours
seb_kuzminsky: I could run one, but my box isn't always on
do you have one that can be ron on non-X systems?
yes the buildslave vms dont need X
hmm.. then I could probably run one
the version of kvm i'm running is linked against some of the X client libs, but the machine i'm running it on doesnt have an x server, and the kvms are running without X windows displaying anywhere
they're exporting vnc displays for their consoles, but i never connect to them ;-)
seb_kuzminsky: I have a lenny VM with gcc 4.3 on it
hmmm. Jaunty probably uses 4.3 as well
I think even 8.10 might
I Guess I'll find out once it's booted :)
running alpha is pretty funny - there are about 100M of updates per day
they sure did get the boot a bit faster in Jaunty
if you wait a week, you might just download a new CD
yeah, once I get back from Hawai'i
I wonder if that 's contributing to the segfault I get running Axis
(if they build Tkinter / python / whatever with gcc 4,3)
still getting the segfault?
alex_joni: are you making a Jaunty/EMC cd?
I try to keep myself busy with LTS releases
just your previous comment to swp was, "if you wait a week, you might just download a new CD"
huh. only ~12M of updates since midnight or so :)
a new jaunty Alpha/Beta CD
mshaver: a new jaunty CD
should be out on the 25th I think
instead of 100MB daily
or March 1 or something
EMC: 03alex_joni 07TRUNK * 10emc2/src/emc/usr_intf/ (shcom.cc shcom.hh): fix a couple more places where LINELEN was incorrectly used. In later versions LINELEN should be increased, and it should replace all those defines
so, anyone interested in running a benchmark for me (and jmkasunich, even if he's not interested :) ), please download http://www.thothsystems.com/images/dsc_0291.nef,
and let me know the result of "time dcraw -q 3 dsc_0291.nef"
you may need to install dcraw, of course (should be a package available)
(jmkasunich - I'm specifically interested in results on the atom)
take out the comma
what is a nef file anyway?
Nikon raw file
at the moment the atom board is at work
ok, no rush
and I'm recovering
how did it go?
I worked all night thursday, at the 9am friday morning deadline it still wasn't working
SWPadnos: interested in a time from dapper/VM ?
was the comma at the end of the link the problem, or does the download work now?
alex_joni, no, thanks :)
how about native ... ;)
I can follow paths that I already know exist, and I can sometimes detect intersections, etc when I see them (I know what needs to be done to make that more reliable)
(even on Windows - there's a dcraw.exe
but I just didn't have enough time
yeah, "seeing" the maze is a pretty big job
SWPadnos: it seems to work now - I didn't bother saving the file tho
oh, and if you can tell me the CPU type/speed, that would be good
jmkasunich, ok. like I said - no rush. I['m headed for Honolulu on Monday, so I won't care much until March ;)
if you can email me the file (or the URL) and what you want me to do with it I'll be more likely to remember
SWPadnos: got dcraw, waiting on the download for the nef
hmmm. odd. is it dribbling through?
jmkasunich, ok, will do (id I remember)
I wonder how you time something on Windows
those are native?
ie, no extra library loading that's part of the timing ...
I think so
[19:30:44] <alex_joni> http://sourceforge.net/project/showfiles.php?group_id=9328
hmm.. no time :/
you may be able to use a batch file that has time /t | dcraw ... | time /t
but I don't know if the /t option is from take command
SWPadnos: got the email, will do
nope, it should be there in a regular prompt as well
[19:34:17] <alex_joni> http://www.pc-tools.net/win32/ptime/
cool, but where's teh screenchot(s) ;)
3 mins eta on the download :/
odd. it's only 8M
my wire is clogged today
it's dribblingin at about 17kB
or 11.6 or something :)
that's about my upload speed
SWPadnos: jmkasunich@mahan:~/downloads$ time dcraw -q 3 dsc_0291.nef
dsc_0291.nef: Corrupt data near 0x13bfc1
corrupt data. interesting
SWPadnos: md5sum it
md5sum it and.
the ppm result is a few random color speckles on a black background
somehow I think that isn't right
the md5sum matches
and I can view the original ned - its a board on a keyboard
although it seems to have a strong bluish tint
is that just because it is raw, no white balance yet?
yeah, it's kind of weird. the ppm would have correct color balance
its also only 160x120
I can't see the NEF file without converting it to something first
for some reason chrome decided to append .tiff to the filename
oh, that's the preview image stuck in there by the camera
alex_joni: does your lenny vm have rtai?
I was just starting to suspect something of that sort
I noticed that the jpegs were blue-shifted, when saving raw+jpeg
SWPLinux: if you run the exact same command you don't get the corrupt data message?
the actual file is 4320x2868
my dcraw is v8.77
let me download it and try on the downloaded version (I changed the name, which shouldn't have an effect)
ISTR I downloaded it about a year back, when I was trying to process pentax raw images
I also STR that I was doing some fiddlefarting around with different versions
I've got 8.86
hmm, dpkg says I have dcraw package 7.94-1ubunu1 installed
I bet I have both - package and something newer that I built from source
you may have downloaded a newer one directly
is there a newer package?
probably not for dapper
(or I would have been updated already)
the atom has hardy on it
here I was thinking that was 8.04
if it ain't broke, don't fix it, is my motto
especially for my main PC
it's 8.80-1 on hardy and intrepid
finally it's doing somehting
yesterday I just dumped everything in the lab and went home disgusted
to top off the 'failure to meet the contest target", I also had "failure to even do a demo run"
I was gonna do a run showing tracking of lines, given pre-knowledge of the maze - I entered the maze config so the code could solve it and run the path
it was a big project. you got a lot further than anyone else I know would have
and then, 5 mins before the demo (just as the real runs were finishing up) a FET in the motor driver board failed shorted
next week, DARPA challenge :)
that's a real lbummer
I had some neat code working
you sure did
is that on a Commodore 64?
for recognizing the intersections, one problem is that an intersection is slightly bigger than the distance from "maximum range of camera" to "obstructed by vehicle's own front bumper"
I suspect the dcraw is badly compiled
I should try it on my Windows PC
so I merged multiple images
the first one I found (optimized) needs some dll's I don't have
on an almost complete unrelated note - found a weird bug? with the atom
the webcam works fine plugged into the back panel USB jacks
alex_joni, a cygwin version would naturally suck
but we didn't want the long cord
what does "const struct foo **blargh" mean? what part of that is const?
so we cut it off and made a connector that would fit one of the motherboard USB headers
SWPadnos: 12.290s for the MS compiler version
that didn't work
seb_kuzminsky: damn good question
I'd think "blarg is a const pointer to a (non-const) pointer to a struct foo
I think it means that the struct can't be modified
btw, I can view it just fine in picassa (without converting)
but it could be that the pointer that's pointed to can't be modified
const makes me confused at times
the "clockwise/spiral" rules says blargh is a pointer to a pointer to a struct foo constant"
which i think is what SWPadnos said
that actually would be more usefull than my interpretation
(which was strictly off-the cuff and without clue)
seb_kuzminsky, that's how I read it - backwards
blarg pointer to pointer to foo struct const
could be wrong though. try writing stuff and see what the compiler error is :)
try writing it, and then modifying various parts, see what the compiler yells about
blarg++; (*blarg)++; (*blarg)->some_member++;
wonder how you write a const pointer to a const pointer to a const struct foo
const struct foo const *const *blargh; maybe?
const pointers aren't much fun tho - might as well just statically allocate things
const struct foo * const *const blargh
In C, you merely shoot yourself in the foot.
In C++, you accidentally create a dozen instances of yourself and shoot them all in the foot. Providing emergency medical care is impossible, because you can't tell which are bitwise copies and which are just pointing at others and saying, "That's me, over there."
it's definately that "const foo *a" a points to a const foo
so you can't use *a = ... to change the value
agreed, but "foo *b = a; *b = 666;" works
sure it does
argh. the bastids who built this windows dcraw make it refuse to write to stdout
I noticed on my machine that the disk light is on for 5-10 seconds, so I wanted to pipe to NUL
incidentally, const foo * const * blah could be appropriate
make a pointer which can't be changed or pointed someplace else
if it means "a pointer that I won't modify which points to a pointer I won't modify which points to data I won't modify"
with as many consts as you want
yeah, missed one :P
and it's still better than static allocation
since const in a function is a promise not to modify something
the compiler will error if you attempt to write to something you said you wouldn't write to
\rather than waiting for a runtime error
gcc 4.3.x is really smart ;)
I'm sure that has been there since gcc 1 or 2 :)
(since that's the whole point behind const)
ah, didn't mean const now.. but gcc 4.3.x is really getting smarter
yeah, FORTIFY_SOURCE is good stuff
yeah, I think I agree, but I don't have enough comparative experience to really know for sure
i'm happy to wade through lots of false-positive warnings if there's a true-positive warning that alerts me to a real error every once in a while ;-)
I'm the kind of guy who will spend a lot of time eliminating all warnings if possible
and if it's not possible, then I'm not above using comments and pragmas to kill the warnings in the source file
it's super annoying when there's no way to soothe programs *cough*mandb*cough* about false positives
instead of having to ignore them at build time or set project-wide options that may miss other places where I would want to see it
0 stray cats added!
alex_joni: what's the CPU in that WIndows box you tried?
oh, and where did you get the executable, or did you build it yourself?
the command line should be changed to dcraw -q 3 -c dsc0291.nef > /dev/null
no sense benchmarking disk store time
I don't have /dev/null :)
execution time: 11.392s
SWPLinux: T5300 running at 800 MHz
(was on battery)
seems CPU doesn't make a difference
now running at 1.73GHz, same time
huh. it doesn't increase the clock when under heavy CPU load?
the CPU frequency changed from 800 to 1730 Mhz
I used cygwin to privide /dev/null and time ;)
but the time on the dcrawms didn't change
(the dcrawMS.exe is compiled with MS compiler)
[20:25:53] <alex_joni> http://www.insflug.org/raw/file_download/8/dcrawMS.exe
I guess I'm asking if, when you ran it on battery, the CPU might have been stepped up to 1.7 GHz while it was running
might work faster for you too
ok, I'll check that one
no, I ran it a couple times on battery
it stayed at 800
I got 22.335s on my Athlon 1800 (windows 2000)
does it have an SSD disk?
so did you use cygwin to get "time"?
~ 120G HDD
I used ptime
[20:28:01] <alex_joni> http://www.pc-tools.net/win32/ptime/
huh. that was 0.4 seconds slower than the other one I had downloaded
and now I don't know where it came from, even though I downloaded it less than 20 minutes ago
oh - firefox instead of Mozilla
huh. the one I have is from the same place, except there's a "3" instead of the "8"
* alex_joni wonders what SWPLinux is talking about
the version of dcraw that I got was from the same site, but I got the non-optimized one since I have win2k
we all wonder that sometimes
* SWPLinux too
SWPLinux: since you have cygwin, there are some other optimized compiles
the MS compiled one refulsed to run unless I piped the output through cat or type to NUL
sure, but I'm not as interested in Windows performance as I am in Linux performance
it's a heck of a lot easier to make a headless Linux machine than it is to make a headless Windows machine
I'm sure it's easier to use the conversion in the camera :D
I'm not sure I'll be using raw files, I'm more curious about relative performance of different processors using these kinds of tools
the next benchmark will be with imagemagick ;)
[20:41:03] <alex_joni> http://www.reghardware.co.uk/2009/02/21/video_sony_ericsson_w995/
-> looks nice
if only it weren't a Sony
one day, the OpenMoko will be usable :)
I love my sony ;)
in fact, I think mine would be if I were in Europe
I don't particularly like Sony the company
the original "proprietary format lock-in" company
those cygwin-based builds are terribly slow
1m5s or so, instead of 22 s
it's actually sony-ericsson
and more important, it ain't nokia :D
which everybody has around here
EMC: 03jepler 07TRUNK * 10emc2/src/hal/drivers/pci_8255.c: add copyright information
EMC: 03seb 07TRUNK * 10emc2/debian/control.in: xsltproc and groff are needed by the configure script now
buildmaster: list builders
Configured builders: dapper-x86-2.2-realtime-deb dapper-x86-2.2-realtime-rip dapper-x86-2.2-sim dapper-x86-trunk-realtime-deb dapper-x86-trunk-realtime-rip dapper-x86-trunk-sim hardy-x86-2.2-realtime-deb hardy-x86-2.2-realtime-rip hardy-x86-2.2-sim hardy-x86-trunk-realtime-deb hardy-x86-trunk-realtime-rip hardy-x86-trunk-sim lenny-x86-trunk-realtime-deb lenny-x86-trunk-realtime-rip lenny-x86-trunk-sim
well let's see if it works
[23:08:12] <alex_joni> http://emc2-buildbot.colorado.edu/buildbot/one_box_per_builder
<- getting long ;)
buildmaster: notify failed
try 'notify on|off <EVENT>'
buildmaster: notify on failed
The following events are being notified: ['failed']
heh, that initial checkout takes a bit
yeah, getting pretty long
build #0 of lenny-x86-trunk-sim is complete: Failure [failed environment-report configuring autotools] Build details are at http://emc2-buildbot.colorado.edu/buildbot/builders/lenny-x86-trunk-sim/builds/0
gotta install some depends
and I have a fix for the debian/configure part
can you change the command that gets run?
<- how is that triggered ?
alex_joni: the buildmaster has "builds", each of which is a sequence of build-steps
the environment-report is the first build-step of all the builds
just as a sanity check more than anything
I noticed on debian there is no /etc/lsb_release
(talk about standards..)
0 seb@lenny-rtai-x86 /home/seb> lsb_release
No LSB modules are available.
lsb_release -a ?
I mean, try running it with -a
the buildbot does run with --all
see the env-check output from the second lenny builder
yay it worked!
wow, big pile of warnings from rtai 3.6.1...
libpth-dev is needed for sim but not for rt
i think that means libpth-dev should be part of Build-Depends:
all in favor say "aye"
I think so
wonder what's up with all those rtai_lxrt warnings
build #0 of lenny-x86-trunk-realtime-deb is complete: Failure [failed compile] Build details are at http://emc2-buildbot.colorado.edu/buildbot/builders/lenny-x86-trunk-realtime-deb/builds/0
forgot to install fakeroot ;)
but that's not a Build-Depend
i added it to the buildslave-admin-guide ;-)
EMC: 03seb 07TRUNK * 10emc2/debian/control.in: need libpth-dev to build sim
dpkg-shlibdeps: warning: dependency on libz.so.1 could be avoided if "debian/emc2/usr/bin/halscope debian/emc2/usr/bin/classicladder debian/emc2/usr/bin/halmeter" were not uselessly linked against it (they use none of its symbols).
my work on this planet is finished
i must now return to my home
on the planet Zarkon
you're from Zarkon?
then we're pretty much neighbours
* alex_joni is also from the Agoraea cluster
seb_kuzminsky: should be that debian/configure puts it in the generated debian/control -- unless you've changed all that around while I've been away
(it = libpth-dev)
jepler: i haven't touched it
at least not intentionally
oh i see
i installed the build-deps for realtime but not for sim
i'll clean up the mess i made...
EMC: 03seb 07TRUNK * 10emc2/debian/control.in:
EMC: revert my previous checkin...
EMC: only sim needs libpth-dev, and it gets it when you run "debian/configure sim"
thanks for setting me straight jeff