cool, git format-patch will usefully transmit changes in binary files, at least as long as the base64(?) representation will fit within the permitted message size. ex: http://pastebin.ca/1524547
and boy do they not mince words on that mailing list: http://article.gmane.org/gmane.comp.version-control.git/125449
EMC: 03jepler 07master * r35832755d4d5 10/ (3 files in 2 dirs): factor out non-axis-dependent stuff from axis.py
[03:11:55] <jepler> http://emergent.unpy.net/files/sandbox/gremlin.png
(no, it doesn't do anything useful)
**sigh** the tool thrower is working great
someone went to A LOT of work to screw it all up and make it use CAT instead of BT tools
now, I'm attempting to put it back
(if I had known all along, I might have left it alone)
jepler: maybe ability to add preview as a widget to pyvcp?
alex_joni: LINUX_2_2 isn't supported in emc at all?
ok found it ,nm
err.. it was
not sure if it stil works
EMC: 03micges 07master * r8a9604f21de8 10/src/emc/usr_intf/ (emcrsh.cc emcsh.cc): Remove obsolete LINUX_KERNEL_2_2 define
EMC: 03micges 07master * rc9b6226b9093 10/src/emc/usr_intf/usrmot.c: Remove obsolete commands from help command in usrmot
micges_work: I am not interested in that, and imo it doesn't really make sense since pyvcp is for hal but preview is for emc.
Just this once .. here's the log: http://www.linuxcnc.org/irc/irc.freenode.net:6667/emcdevel/2009-08-11.txt
* jepler wondrers if 'master' still compiles on breezy (the only realtime machine I can access at the moment)
I can't get a new enough git to check out master, so I had to copy the source across the network
hm, so no log messages / CIAs for the first push that makes a new branch?
yeah, that's been the pattern so far
EMC: 03jepler 07defer-format * r6f573e3f6997 10/src/ (6 files in 2 dirs): make rtai work again
I'll regret saying this before the rubber meets the road, but that went rather easier than I expected
so.. in simple terms - this is so errors are passed as a number lets say instead of strings? then a lookup table will be used?
skunkworks726: no, not at all
that's a terrible system -- somebody has to pick the numbers, for starters
and when you look in the source code at the place where an error is detected you see the numbers or UNENLIGHTENING_ERROR_DEFINES instead of sensible strings
I guess I don't get the formating after vs before.. (mainly - what is the formating) but it really doesn't matter.
in linux, translation is usually done with a system called 'gettext'. The function gettext(s), sometimes written as _(s), takes a string 's', looks it up in a message catalog, and returns the translated version of s
we use gettext in axis, task, and the interpreter to translate messages
but we can't use it in realtime, because it relies on stuff like reading the message catalog from disk
ah - takes too much time I assume.
you can't depend on it not taking too much time, anyway
so anyway, the code for a specific error message might look like this: reportError(_("Can't jog invalid joint number %d."), joint_num);
originally, the formatting of this into the final string, such as "Can't jog invalid joint number 0.", was performed in realtime
now, there is code that gathers the arguments to reportError -- in this case, a string (still in english) and an integer -- and puts them in a buffer.
task checks the buffer for a new item. If there is one, it pulls out the items, translates any that are strings, and performs the formatting
ok - that makes sense to me suprisingly.
hopefully my implementation is as good as my explanation, then
:) thanks jepler.
I think less 'stuff' happening at the motion layer the better.