steve_stallings is now known as steves_logging
steve_stallings is now known as steves_logging
EMC: 03micges 07task_cleanup * rc479e164d98b 10/src/emc/nml_intf/emc.cc: Remove unused NML update functions
why did seb post that patch instead of putting it on the branch?
EMC: 03jepler 07master * r86dbfc89c9f7 10/src/hal/utils/comp.g: add way to get path to Makefile.modinc
EMC: 03jepler 07master * r1064135faa89 10/src/ (emc/usr_intf/emcsched.cc rtapi/sim_rtapi_app.cc): these headers are needed on gcc4.4 systems
cradek: because I asked people not to push changes to v2_3_branch
what's weird is that his patch shows no context
yes that is weird
but if you have a tool database somewhere, then you can just select which number(s) you're using
it's interesting how many endings there are on google for "and if my grandmother had wheels, she'd be ..."
and people tearing the siding off the house. amazing how that kind of noise prevents thinking
jepler: what do you think is a good way to email patches?
last night i just attached the git-format-patch output, just now i used git-send-email
i used "git-send-email --compose" to type in an introduction to the patch ("please apply to 2.3" basically)
I've used git-format-patch
but for me it showed context
on the mailing list, I have your introductory message but not a patch
i fixed the context thing, i had an *old* environment variable giving arguments to git-diff
i also only have the intro :-(
heh - your last email doesn't have an attachment.
the patch is supposed to show up as a separate message, not as an attachment
i'm not sure which is better
probably minimize the number of emails that sf.net gets a chance to delay :-/
[16:29:07] <seb_kuzminsky> http://highlab.com/~seb/emc2/0001-hm2-stepgen-bugfix-for-float-precision-logspew-probl.patch
hi micges :-)
did you use prevent on emc?
[16:37:45] <seb_kuzminsky> http://scan2.coverity.com:9082/view-runs.cgi?magic=dc469294a28f113fa7dc57ade1244270&user=seb&noprevpage=1
did you buy it or is it live check with coverity ?
open source projects don't have to pay
I didn't read carefully
you still have to register and all that, and I think the results aren't made public by default, but they'll do the scans for free
seb: can you describe a little that bug that you patch is for ?
coverity's free-for-floss service is great, mad props, but they run a really old version of prevent on it, and they dont handle branches :-(
so the 2.3 runs and the master runs are all jumbled
micges: i tried to describe it in the commit message in that patch on my website
yes, but is it only applicable to maxvel?
i think it's applicable anytime you compare float arithmetic results to floats in ram
if you do something like if (*foo > bar/baz), you could hit it
I mean hm2 specifically
other params can't hit that trunc problem?
i haven't checked... probably should go audit all the realtime code
note that the "problem" that was fixed is only that dmesg would get spammed
it should work the same, since all valiues are eventually converted to 64-bit floats when they're stored
and a single superfluous store-to-ram, but who's counting ;-)
by the way, we could easily add branches to the buildbot, if people want
joints_axes3 could be
hm, building joints_axes3 with the "-trunk-" builders would be easy but the names would become misleading
making new builders for the new branch makes the waterwall display even wider
renaming the builders to something like "-trunkish-" might be best
grr, submitting a build to buildbot 0.7.9 via the admin webpage bypasses the scheduler branch restrictions
seb: if it is problem forget it
i think it's done, testing it now, i was just too impatient to wait for a new commit so i told it to rebuild an old one
and i got a bit too excited and overdid it ;-)
all the 2.3 builds are going to fail, because they're going to check out master but not run autogen
oh yeah, but after that it should be good, the 2.3 builds are untouched, the trunk builds are renamed to "trunkish" and build both master and joints_axes3
seb_kuzminsky: the builds still happen automatically don't they?
answering my own question: I think so - the hook is there
wow. the 5i23 is a tiny board
it basically extends about 1mm beyond the PCI connector
cradek: should do, yeah, i dont know why they wouldnt
# On branch v2_3_branch
# Your branch and 'origin/v2_3_branch' have diverged,
# and have 516 and 186 different commit(s) each, respectively.
I think I just did something wrong
EMC: 03jepler 07v2_3_branch * r99c26fa12775 10/ (debian/changelog src/hal/drivers/mesa-hostmot2/stepgen.c): hm2 stepgen bugfix for float-precision logspew problem
now I just have to find the motivation to do testing and a release this weekend
hm, on this machin I have io, halui, and hal_manualtoolchange processes left over from as long ago as july 1
I wonder why they haven't exited