#linuxcnc-devel Logs

Sep 09 2018

07:47 AM jepler: Jin|away: auto
07:49 AM jepler: make was not suitable for building kernel-type modules without too much banging.
07:49 AM jepler: another possible breakage from linuxcnc 2.7 -> master is that we unbundled our version of yapps. perhaps fedora's version of yapps is just incompatible.
07:49 AM jepler: $ git ls-tree -r origin/2.7 | grep yapps
07:49 AM jepler: 100644 blob 1bb8bf6d7fd4c8d09aea89b47de20fb8bbb61626 lib/python/yapps/__init__.py
07:49 AM jepler: ...
07:49 AM pcw_home: this is weird just upgraded home PC OS now LinuxCNC does this:
07:50 AM pcw_home: emc/task/emctask.cc 397: interp_error: EOF in file:/home/peter/linuxcnc-dev/share/axis/images/axis.ngc seeking o-word: o<reset_state> from line: 0
07:53 AM Jin|away is now known as Jin^eLD
07:55 AM Jin^eLD: jepler: the yapps thing would be a pain if this incompatibility is indeed the case...
07:57 AM Jin^eLD: allthough I'll probably start messing around with the build once I update my system to Fedora 28, I am still on the older 27 right now
07:58 AM Jin^eLD: I helped my friend yesterday to compile 2.7.14 on F28 which required some makefile patches for the rpc stuff, otherwise it did build
08:07 AM Jin^eLD: on a slightly different topic.. I'd appreciate some hints: my comp is almost feature complete and I am thinking of ways to test it within a simulation
08:07 AM Jin^eLD: I saw that there is a full sim with gmocappy
08:07 AM Jin^eLD: the comp "only" does the gear switching
08:07 AM Jin^eLD: does it sound feasible to reroute the gmocappy sim in a way that it would pipe the speed requests through my comp?
08:08 AM Jin^eLD: then I could see how it behaves within a complete linuxcnc environment, i.e. within a milling test for instance
08:08 AM Jin^eLD: ?
09:59 AM jepler: so what does it take to get rid of xdr rpc? probably awful floating point precision problems, but nothing our testsuite distinguishes https://github.com/LinuxCNC/linuxcnc/commit/6b585965b8803ee77ae135b2bbab2496a8f5b402
10:00 AM jepler: Jin^eLD: sounds reasonable to me
10:00 AM Jin^eLD: well, I did not really have to get rid of it as such, they implemented a switchover to libtirpc
10:00 AM Jin^eLD: glibc-headers on F28 do not provide /usr/include/rpc anymore
10:01 AM Jin^eLD: so CXX/CFLAGS need an include of /usr/include/tirpc/ and for linkage -ltirpc must be added where appropriate
10:03 AM Jin^eLD: but otheriwse the functionality is the same, drop in replacement
10:04 AM jepler: sure, but I'm not aware of anyone who's offered a patch that doesn't break working systems
10:05 AM Jin^eLD: well, I can't test this on F27, so I can only come up with a configure patch when I get back fro mholidays and update to F28
10:05 AM Jin^eLD: I have limited traffic here
10:05 AM Jin^eLD: ideally configure should check if rpc/rpc.h is usable and fallback looking for libtirpc
10:05 AM Jin^eLD: sholdnt be a big deal, I can look into it in october
10:07 AM Jin^eLD: I'm more worried about not being able to build the master branch on F27 (dont know how F28 would do), I know python generally, but not well enough to be aware of python2/python3 syntax differences
10:07 AM Jin^eLD: and well, halcompile is a lot of code, probably easier fixed by someone who knows how it works
11:08 AM jepler: you have to use python2, or put in 100-1000 hours fixing the python2 / python3 incompatibilities.
11:11 AM Jin^eLD: well, the thing is that I was using python2, I think thats what linuxcnc picks by default?
11:12 AM Jin^eLD: the python3 compatibility commit that you identified seems to break python2...
11:15 AM jepler: did you actually determine that it was that commit/merge? I identified another possible 2.7 to master difference (bundled yapps -> OS yapps)
11:16 AM jepler: I don't have fedora and don't plan to have fedora so it's sort of on you to figure out whta is broken and what the fix is, in this case
11:18 AM Jin^eLD: I did not look into it too closely, I added a traceback to see what actually trigered the exception, but got lost not knowing the python2/python3 differences, allthough at that point I did not even know if that was the problem or if something did not work out for other reasons
11:19 AM Jin^eLD: and then when I was going to retry at some point I hit the yapps issue
11:20 AM Jin^eLD: but yes, I understand that you are not going to fix fedora specifically :) and I would only start messing around with it after updating to the most recent one, because it does not make much sense to fix it for an out of date distro
11:21 AM Jin^eLD: I think they're already on the way to F29 while I still have F27
11:22 AM Jin^eLD: luckily its not a blocker for me right now, because the 2.7.14 release compiles and on the actual machine my buddy is anyway using the linuxcnc distribution
11:33 AM Jin^eLD: btw which of the available simulations would you suggest for trying to hook into?
04:17 PM Jin^eLD is now known as Jin|away