jepler: found it!
a mesa card
we don't have any more bugs, they're all fixed
the servo drive card>? for jepler?? :)
did someone write the manual on all the unexpected features then? :)
EMC: 03cmorley 07TRUNK * 10emc2/src/emc/usr_intf/pncconf/ (pncconf.py pncconf.glade):
EMC: Add spindle signal names to stepper signal names variables.
EMC: fix charge pump signal name so it works
EMC: fix encoder index HAL pin numbering error
EMC: fix servo open loop test so it works - need to fix Estop button
EMC: maybe make the test controlled by keyboard too.
jepler: cvs server is down
micges1 is now known as micges
git patch series to get rid of hal and rtapi error constants and use unix errnos instead: http://emergent.unpy.net/files/sandbox/error-constants.patch
wow, that looks like it was a huge pain
nah, not really
are you unsure about committing it?
mostly done witregular expressions
could you correct the spelling on "successful"? :)
no, probably not
it's funny. I started to look at doing the same thing, but I couldn't find the kernel errno definitions
I think I looked at it when someone had an error loading a module, and I couldn't tell them what it meant
it looks like my proposed change mostly gets rid of the "successfull"s
that's the spelling error you mean, right?
it's there on the HAL_SUCCESS line
ah, see, you didn't read all the way to the end
in the last commit, all those defines are removed altogether
here I am in CVS mode
but you're right, maybe I should add a commit before all the others to correct the spelling error
probably not necessary :)
I may be overdoing the commit splitting, to be honest
did you get a chance to try the new hm2 stepgen driver & firmware?
still waiting for hardware to test (as usual)
I'll surely let you know
ok, i'll stop nagging you ;-)
aeb: another thing: today I was playing pwmgen from mesa
you not nagging me it's the same important to you and for me
and for some values .velocity from encoder was NAN
that's a new one...
version from emc2.3 beta but encoder.c didn't change since then
all velocity estimate code is in encoder.c ?
for the encoder, yes
stepgen has its own velocity feedback
I have nan once for minute and I switched to ddt's and everything works ok
the encoder vel estimator should be much better than ddt of position
this is from a pre-2.3 version of encoder.c? do you know how old?
I'll later try find hal from my test machine
seb: last commit to encoder.c was 3 months ago
yes, the encoder velocity should have been good since Jan 31 2009
but ok I update to 2.3.1 tomorrow and recheck
what was the condition when it reported NaN for velocity? going fast or slow? changing speed? anything funky going on?
slow and changing speed to -slow
if you update to the CVS top-of-tree in the 2.3 branch you'll get the stepgen fixes too
ok, i'll try to reproduce it tonight when i look at the encoder.reset thing chris morley reported
I noticed that, thanks
seb: I've looked into source and there is very little space to generate nan
seb: that was on encoder.2 and scale was set to 1000
velocities from +-20 to +-200 changed by setp
seb_kuzminsky: I notice that you declare locals of type 'hal_float_t'. You should probably declare them as 'double' instead. You don't want the 'volatile' qualifier, and you probably don't want the reduced precision for temporaries in the case that hal_float_t is narrower than double.
jepler: oh yeah, good idea
my night is filling up...
problem is, i'm not sure I'd accept that change in 2.3, but that makes it near-impossible to automatically apply bugfixes from TRUNK to 2.3
ugh I hate everyone
heh, it's near impossible anyway with CVS, so nothing lost there ;-)
* seb_kuzminsky swoons
(can't you cherry-pick, or whatever it's called, in bzr and then export to cvs?)
yeah that's what i do currently
bzr unfortunately doesnt merge-track cherrypicks :-(
but it works ok
I don't see how you can
cherrypicks aren't merges, so how can you track them as merges?
they're sort of partial merges
doesnt git track cherrypicks?
the "git cherry-pick -x" option suggests not :-(
"git rev-list --left-right --cherry-pick" and "git cherry" have some way of determining that a commit "introduces the same change as another commit on the 'other side'"
it seems to use this thing called a "patch id"
wow, there seems to be nearly a 10x speed difference running 'git filter-branch' between these two machines (old p4 2.4ghz slower and new phenom 2.3ghz faster)
(I'm trying to make an alternate copy of the emc history without any of the bitfiles, to see how much smaller it is)
one commit every two seconds vs multiple commits per second
which is nice because I keep missing some subtletly
and there are 17kcommits to do
Rewriting emc2's history without the .bit and .BIT files trims about 10 megs off; .git/objs goes from about 38 megs to about 28 megs.
I don't think this is a big enough benefit to justify destroying history
peter is supposed to mail me a replacement for every .bit that has stepgens in it...
micges: I added you as a developer - welcome! If you would try getting a devel checkout now, we can make sure it works.
micges: what is your sourceforge login?
the same here
next round of beers is on micges
devel checkout is working
now I am busy assigning all the bugs on sourceforge to you
cradek: there's an auto-assign feature
oh well... ;)
my Jaunty RTAI kernel has been up for a day now
jaunty is so yesterday
I haven't changed it, but I also haven't run any RT programs
yeah, I should be using Karmic Koala now, huh?
at least you'd be the first one for a change :P
micges: I also added you to the bug and feature request trackers so you can open/close/assign things to yourself
especially new ones :P
I mean the ones you report :P
git marks all commits with a full name and e-mail address. For the cvs history import to git, should I use Michał Geszkiewicz <email@example.com> or some other address instead?
actually, that makes me think I should post this to the developers list
EMC: 03micges 07TRUNK * 10emc2/src/emc/usr_intf/axis/extensions/emcmodule.cc: remove unused fields in nml
EMC: 03micges 07TRUNK * 10emc2/src/emc/nml_intf/ (emc.cc emc_nml.hh): remove unused fields in nml
wow it works :)
good night all
jepler: I see where line is and comment is wrong
thanks for advice
good night all
jmkasunich: I dunno if we've talked about this before, but what do you think about getting rid of hal and rtapi error numbers and using unix errnos instead?