alex_joni, swpadnos: Would you like to give the 2.3 smp packages a home on linuxcnc.org?
ehj666: pm me the login info
Do you remember the login for my ftp site?
they're up at: http://linuxcnc.org/experimental/hardy/smp/
ehj666: feel free to reply to the list with the addy
k, thanks alex
alex_joni: I recive 404 on http://linuxcnc.org/experimental/hardy/smp/
micges: odd, it works here
I can't download any of those files
ah, the folder works, the files don't
try now ;)
cradek, if ran loadusr halui from the halconfig window would that load halui for the session
I wonder if there's any trouble that would be caused by doubing the hal shmem size (again), or if it could be runtime configurable somehow
I think hm2 has many more pins than all the other drivers, so we're going to see this problem now as more people use it
i have two cards also
not just 1 like most
that's the killer, I bet
cradek: I must change hal to 160k size since Januray ;)
micges: no problems after doing that?
no problems at all
even when size=260000
can get cvs ok and complite it no worrys
rob: follow section 2.1.2 and then 2.2 where it says how to get RELEASE_2_3_0
rob: then increase (double?) that memory size, then compile it with --enable-run-in-place, and test for us
I don't think any of the devels have two mesa cards
perhaps we could double it in 2.3.1 - we will need to experiment and see if anything breaks first.
rob: simple build instructions in section 2.4 of that same page
need change both hal_size values?
cradek: I have 11 mesa at work and pc with 5 pci slots ;P
I thought we increased it for 2.3.0
I can run some tests
alex_joni: last time we increased it was 2006/05
way before hm2 :-)
yeah, but we talked about it shortly before 2.3.0
a user with 2 x 5i20 + halui reported out of memory
heh, sounds very familiar
can it be runtime configurable?
dunno.. it's a constant in HAL somewhere
yeah I know
[19:40:13] <cradek> http://cvs.linuxcnc.org/cvs/emc2/src/hal/hal_priv.h.diff?r1=1.15;r2=1.16
maybe jmkasunich will comment later
if someone sends me three 5i20 cards I'll happily test
(I bet 5i22 is much worse, with its third connector)
I think the guy with the 2 x 5i20 tested a bigger size without problems
the only thing is that we need to bump HAL_VER if we increase the HAL_SIZE
loadrt hostmot2 debug_idrom=1 debug_module_descriptors=1 debug_pin_descriptors=1 debug_modules=1
^ from rob's config
I think jmk looked into it, and said it's ok, or something like that
does turning off those debugs remove some pins?
beats me :)
with a pogo stick
rob: how it's going ?
hi micges i have not yet tryed, ill give it a go early morning i think tomorow
its only the last define line i have to edit correct? define HAL_SIZE 131000
to say, define HAL_SIZE 160000
set it to 260000
thx for quick pointers tho micges & cradek
rob: bump version too
HAL_VER should be a couple lines above
it's 0x0000000A or something
make it B
ok alex_joni thx
* jepler finally get around to his wichita fest room reservation
dgarr: it looks like your e-mail client misformatted your patch; it doesn't apply
patch: **** malformed patch at line 28: };
in several places it looks like line wrapping was performed
I think I fixed it up
but maybe you should check your e-mail client settings for the future
EMC: 03jepler 07TRUNK * 10emc2/src/emc/usr_intf/emcrsh.cc: -h / --help to show usage (thanks dewey)
jepler: ok -- thanks ( i pasted into mutt) maybe it's better if i use filebin: http://filebin.ca/daroe/diff.emcrsh.cc
another patch: for hal_input params-to-pins, cradek suggested holding until after 2.3 so maybe now is ok to consider, i've tested with a joypad only: http://filebin.ca/qugmwy/hal_input.py.diff
dgarr: if using vim with mutt, ":set paste" before pasting helps a lot
cradek: good to know -- maybe a mutt attachment would be better also
cradek: any interest in modifications to ttt.c? http://filebin.ca/xxfsov/ttt.c
i've been using this for a while but not sure if there would be others interested
wild, I noticed that isalnum bug just today!
this looks great - thank you
about ttt - unfortunately I think another package/program is called that already.
that's why truetype-tracer (formerly known as ttt) got its long name.
the rest of it looks great.
good -- the raw format is purposefully simple to apply in alterate applications that don't want to parse dxf or use gode
that's a good idea.