jepler: hitting escape on keyboard and hitting stop on halui.program.stop results same message to be sended to task but with differnet serial number
micges: yes. commit e626d40 don't clash with other guis' serial numbers so easily
I'm trying to locate this issue now
debug on escape: http://pastebin.com/Cc9ep5Ch
debug on hal button: http://pastebin.com/LGRNfDG3
jepler: program linked with emc2-sim-dev is working with rtai
psha: interesting. not for me. and I'm uninterested in making a guarantee going forward that it will work.
[15:43:53] <jepler> http://pastebin.ca/1986401
why don't you tell us how you really feel about that software! http://wiki.linuxcnc.org/cgi-bin/emcinfo.pl?action=browse&diff=1&id=Cam&revision=141&diffrevision=140
(but seriously: yuck!)
jepler: i understand that this must be kept in secret :) but fyi - it's working :)
and there is nothing strange with it - libhal has userspace library interface without any rtapi/ulapi stuff
qq-: i've added segfault fix
nice, in which package , but your site seems down
yes, changing cpu
which packages needs to be d/l for fixing segfault?
wait a bit
psha, my issues here with camview is that my camera is marked off ones added to chain, even the is on (i prestarted this by qv4l2)
i've written v4lconvert plugin too
but first check that it's not segfaulting :)
is it segfaulting still?
no more segfault ;) thanks
try to run LD_PRELOAD=/usr/lib/libv4l/v4l2convert.so camview
and select RGB format in camera output formats
is it working?
RGB format is ok
displaying everything fine?
i have some misterious bugs with v4l2convert preloaded so asking for test :)
display still nothing, RGB format is ok in selection on convert from unknoun..
two plugins in chain?
as said in camview my camera is treated as off
camera and opengl renderer?
try to save this chain and then run camview again with -c chain-file flag
not, only camera + convert from , adding now opengl
now a have a black screen
save this and reload
i'll be back :)
psha, save is ok, reload + (-c ...) is ok , i get last saved config ,and sure the black screen
I have a problem compiling http://www.linuxcnc.org/emc2/dists/lucid/base/source/linux_2.6.32-122.35.rtai.tar.gz
- After 'make modules' runs for several hours, it errors out with, "ld: ubuntu/omnibook/sections.lds: No such file: No such file or directory". I don't really care abour omnibooks, but this is preventing me from making the one little 17kb module I need! Is this archive incomplete?
[18:01:52] <cradek> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/505420
Trying this fix... Thanks!
That worked - compile resumed... Thanks!
cradek: are you using acad 12 for dos in wine?
no, qemu + freedos
I wanted to install an old version of acad12 for windows but had no luck. I am giving up and installing xp in virtualbox
I will use it for otherthings anyway
I'm surprised that something like old AutoCAD wouldn't work, when something like Altium Designer does work (mostly)
yeah, I also would have expected it to work fine.
I am sure I was doing something wrong.. but I didn't find much info on it in the few minutes I searched so I gave up.
After running 'make modules' on the kernel sources archived on linuxcnc.org, diff reports that there are differences between the modules on my system in /lib/modules/linux2.6blahblah/kernel/* and the ones produced by the compile. When I try to insmod a newly compiled module, I get "invalid module format". Ideas?
I've used Icad in wine for years. It's a nearly perfect acad clone.
It runs out of the box on wine 1.0.1
I can send you a link to download it if you like
psha, found that camview use as "capture method read()" (and no show in display) , even when camera is up and streaming , but qv4l2 stream and show images only in "capture method 'memory mapped I/O'"
qq-: camview uses mmap
may you make a screenshot?
of camview ?
mshaver: sure - thanks
I swear xp runs faster in virtual box
I just installed sp3 in less than 15 mins
skunkworks, I wouldn't be too surprised
same results with LD_PRELOAD?
I actually tested some image processing functions on Windows and Linux, using the same computer (dual boot), using the same software (ImageMagick, but compiled for each OS), and the Linux runs were faster, sometimes by a large percentage
disk access seems to be unreal
it actually works :)
SWPadnos_ is now known as SWPadnos
psha, and that is qv4l2 get after camview start what http://picpaste.com/scrot_3-ltoXIHQU.jpg
i think it's ok since you try to query data from one device with two apps
may you try LD_PRELOAD=/usr/lib/libv4l/v4l2convert.so camview?
without v4lconvert plugin
note of the day: remember to specify --install-recommends=no particularly with apt-get build-dep. It might save you 300 megs of downloads when you install emc's build deps
psha, pics are with LD_PRELOAD=/
with it you may choose RGB format in camera, without convert plugin
[20:36:37] <qq-> http://picpaste.com/scrot_4-mV8LNjzZ.jpg
0.0 fps is bad :)
btw you may set CAM_DBG=input LIBV4L2_LOG_FILENAME=/dev/stderr environment variables to see what's happening
psha, see it come with say ~10 , and in ~5 seconds go to zero
fps is shown as average over some time
and stay on zero fps for ever
run with debug environment vars set
and see what it tells you
psha, too high level for me
i not done nothing , b/c i cant get you
place "CAM_DBG=input LIBV4L2_LOG_FILENAME=/dev/stderr" before command
that done " CAM_DBG=input LIBV4L2_LOG_FILENAME=/dev/stderr LD_PRELOAD=/usr/lib/libv4l/v4l2convert.so" m and say nothing
at the end
it's setting environment variables for next command
ah . sry
it'll give you lots of output
lots of output and now ir stay in ~ 11.2 fps ,, still no image
there are repeating DQBUF/QBUF lines?
so it's actually capturing
but with no luck :)
it seems so
i'll exchange my working cam with non-working tomorrow and see what's happening
it's not that stange as yours but with unsupported format to
[20:50:42] <qq-> http://paste.debian.net/99395/
i'll try tomorrow with bad cam
and report back
and now is time for pushing fixes upstream :)
psha, it stream only we don't use opengl, or opengl is off ..
thats sure opengl cut down fps to zero
it stream == we see normal fps activity , even with opengl , is is marked as streaming but no fps
[21:18:06] <qq-> http://paste.debian.net/99400/
< here I started qv4l2 as memory mapped I/O capture method (working case for qv4l2 )
s/I started / I started before
qv4l2 + read () capture method I get http://picpaste.com/scrot_8-MVeTWdW4.jpg
yes, only one mmap user is allowed
i meant en read () they works both ..
work - but without any image?
yes , ftp* is working
so let me take a look at another gspca camera with JPEG (non-MJPEG) format
it was asked already but what type of licence RTAPI have? GPL or LGPL?
in rtapi.h there is note about lgpl but in rtai_rtapi.c there is gpl notice
Clearly the headers and code are licensed differently...
(is that even possible?)
you must include headers in your code if you want to use the library, so those have to be LGPL (to allow proprietary use of the interfaced defined in the headers)
the library itself can't be used in proprietary code, so that's GPL
the distinction is that you can use the shared library (at runtime) by including the header, but you can't appropriate the actual library code itself
so realtime drivers could be non gpl ?
(well, including the header lets you use the library in your code anyway)
yes, that was the original intent
I can't say that it's still true though, you do need to look at the licenses
yes I'm looking now
I think the idea was that if you have some proprietary widget, and the driver needs to have some proprietary knowledge in it, then you should still be able to write a HAL driver
without having to show the world your proprietary knowledge
Or, I could use it to prevent people laughing at my code.
right. they'd have to laugh at the results only
Talking of which, has anyone been bored enough to look through the 8i20 stuff?
I hate to see a patch languish, but I'm entirely unqualified
Well, I am still working at it anyway, I managed to introduce some new null-pointer problems tonight. I seem to have a real talent for that.
I haven't looked at it (and I don't have any of that hardware either)
No, I am not sure many people do. I have serial numbers 2 and 3.
(well, A00000000002 and A000000000003)
I do have a 3x20 and associated cards now though, which is cool
This is an embarassingly basic question.
Given an element in a struct u32 *command_reg_write
nevertheless someone will be able to give you a wrong answer...
And a pointer to the particular instance of the struct
hm2_sserial_instance_t *inst = &hm2->sserial.instance[i];
What writes to the "value" of command_reg_write?
I have *inst->command_reg_write = 0x0;
But even reading that gives a null pointer problem
(Perhaps it is just uninitialised)
as in a=inst->command_reg_write
Sorry yes, I get a Vector 14 with HM2_DBG("*c_r_w %x\n", *inst->command_reg_write);
ok, that's different :)
(no * in mine)
try DBG with HM2_DBG("*c_r_w %x\n", inst->command_reg_write);
which will print the value of the pointer, not what it's pointing to
Gives me 0
that's a problem then, since it's a null pointer
(pointing to 0, which is null)
So, I have the level of indirection right, but a null pointer, not a good pointer and too much indirection?
you are correctly accessing a pointer which is null (points to location 0), which isn't allowed
HM2_DBG("&c_r_w %x\n", &inst->command_reg_write); gives me fdd6b4b0
that's the address of the pointer
andypugh: are you aware that *a->b means *(a->b), not (*a)->b as you might expect?
which isn't too helpful in this case
-> has higher precedence than *
cradek: Yes and no. I think that it does what I expect, and I am now wondering why I should expect otherwise,
what is supposed to cause command_reg_write to point to something useful?
ok, so tram_register should be called with &inst->command_reg_write
(as the pointer to whatever tram_register is registering)
I have spotted the problem. I don't call tram_register_write in setup mode.
oh. excellent :)
I wonder if there are any motors at MPM that could be used to test this code (assuming the right Mesa stuff magically appeared there)
I have been re-jigging a state machine to do communications handshaking, forgettting I need to do it two different ways.
Any 3-phase motor with any sort of feedback will work, to an extent.
There is still a fair bot of work that needs to be done on the bldc hal components, though the patch makes the bldc_sine component work to an extent (I am struggling with homing, but suspect problems with my resolver converter).
andypugh: I could bring a resolver - would that help?
surely we could find a 3 phase motor of some kind.
(are you coming to fest?)
I actually looked at the airfare today, as I want to buy a motorcycling suit from Duluth (and $100 shipping + $200 tax is worth avoiding)
But no, I won't be at the fest.
My test setup is my 8i20 driver, my bldc_sine component and my Arduino Resolver converter. ALl of which are largely untested, so problem solving is challenging.