#linuxcnc-devel Logs

Apr 21 2020

#linuxcnc-devel Calendar

01:27 PM seb_kuzm1nsky: there is new activity in micges' mesaflash repo: https://github.com/micges/mesaflash
01:31 PM seb_kuzm1nsky: there is also JT's "mesaflash 2" repo: https://github.com/jethornton/mesaflash
01:31 PM seb_kuzm1nsky: why the fork? what's the plan?
01:31 PM seb_kuzm1nsky is now known as seb_kuzminsky
01:38 PM Tom_L: last i knew jt's was the most current. i dunno what micges has done
01:40 PM seb_kuzminsky: why is there a fork at all?
01:41 PM Tom_L: i'm not sure, maybe micges wasn't around when pcw did some edits
01:42 PM Tom_L: just speculation
02:15 PM jthornton: no access to micges to update so I forked it so we could update it
03:02 PM Tom_L: that's what i thought but wasn't positive
03:35 PM jthornton: now pcw_home has push rights
03:48 PM seb_kuzminsky: oh, that's an improvement
03:49 PM seb_kuzminsky: jthornton: is there still need for a second fork? are there a lot of changes in your repo that are not in micges?
03:50 PM jthornton: https://github.com/jethornton/mesaflash
03:50 PM jthornton: 47 commits
03:52 PM Tom_L: i think he was working on spi wasn't he?
03:52 PM jthornton: hmm micges made a couple of commits yesterday
03:52 PM jthornton: who?
03:52 PM Tom_L: pcw
03:52 PM jthornton: https://github.com/jethornton/mesaflash/commits/master
03:53 PM seb_kuzminsky: micges just merged other peoples' PRs
03:53 PM seb_kuzminsky: he hasn't done any development on mesaflash since 2015
03:53 PM Tom_L: it _is_ a bit confusing having 2
03:53 PM jthornton: ah I see that now
03:54 PM jthornton: tell me about it, but what can you do?
03:54 PM Tom_L: last one i pulled was jt's
03:54 PM Tom_L: have pcw link yours to his site as the official one?
03:55 PM Tom_L: since it's really not a part of linuxcnc
03:55 PM Tom_L: although it may as well be :)
03:55 PM jthornton: yea I never understood why is was not in linuxcnc's repo
03:56 PM seb_kuzminsky: hmm, i realize i have a fork of it too, on github :-/
03:56 PM Tom_L: not everybody needs/uses it?
03:56 PM seb_kuzminsky: iirc we asked micges if we could transfer the repo from him the the LinuxCNC org on github and he declined
03:56 PM Tom_L: pcw pointed me to jt's last time i updated
03:56 PM seb_kuzminsky: it would make sense to have it be a separate repo in the linuxcnc github organization, imo
03:57 PM jepler: I think that would be good as well, assuming pcw_home concurs
03:57 PM Tom_L: i think jt's makes a deb too and i'm not sure micges's did
03:57 PM jepler: several of us have the credentials necessary to move the code under linuxcnc organization. I'd have to check to be sure who all.
03:57 PM seb_kuzminsky: micges' repo does build debs, it's how we've been releasing it on wlo
03:58 PM Tom_L: oh ok
03:58 PM seb_kuzminsky: jepler: we'd need micges to agree to the move, and i think he didn't last time we asked
03:58 PM Tom_L: odd
03:58 PM seb_kuzminsky: which was years ago, maybe it'd be worth it to ask again
03:58 PM Tom_L: worth a shot
04:00 PM seb_kuzminsky: hmm, there's a bunch of stuff in the v3.3 branch that's not merged into the master branch that micges just updated
04:00 PM seb_kuzminsky: i'll reach out to micges and ask again
04:00 PM pcw_home: I'm happy with it anywhere is accessible
04:01 PM pcw_home: s/is/its/
04:02 PM seb_kuzminsky: i want there to be as few forks as possible, and i prefer having a group rather than an individual have push access
04:02 PM jepler: because of reasons, the micges repository and the jthornton repository don't share history. It will take some effort if we need to reconcile the two. I can do at least some of it, if it becomes necessary/valuable.
04:02 PM jepler: seb_kuzminsky: I never got to looking at the rpi-kernel update problem yet, sorry
04:03 PM jepler: I might just pin it in our image and call it a day if I can't figure out and solve the underlying problem
04:03 PM seb_kuzminsky: that sounds fine to me, at least as a (multi-year?) stop-gap
04:04 PM jepler: what's that old rule? do the simplest thing possible, even if it clearly won't work?
04:05 PM seb_kuzminsky: i'm sure it's soemthing like that
04:06 PM seb_kuzminsky: 'apt-mark hold' should do the trick
04:10 PM JT-Shop: yea if we can get it into github.com/linuxcnc that would be great
04:10 PM JT-Shop: IIRC I just added instructions on how to build it
04:11 PM seb_kuzminsky: all the 7i96 support is in the v3.3 branch of micges' repo
04:11 PM seb_kuzminsky: that's what the 3.3.0~pre0 debs on wlo are built from
04:12 PM seb_kuzminsky: looks like that's included in JT-Shop's master branch, so that's good
04:15 PM seb_kuzminsky: ok, i asked him if we could move it to the linuxcnc org, we'll see what he says
04:16 PM pcw_home: I dont think there's anything significant missing from the JT branch (and much missing from micges)
04:16 PM seb_kuzminsky: JT-Shop: looks like you mostly added travis integration in your repo, no real changes to mesaflash itself
04:18 PM pcw_home: 7I95,7I97,7C80,7C81 support, and new module support (xy2mod,inm,inmd,rcpwm)
04:18 PM pcw_home: slightly improved usage instructions
04:19 PM seb_kuzminsky: pcw_home: jt's master branch looks very similar to micges' v3.3 branch (which has not been merged into micges' master branch for a long time)
04:19 PM seb_kuzminsky: oh yeah, i see the missing boards you're talking about, those are just in jt's repo
04:20 PM seb_kuzminsky: (it's hard to tell since they don't share git history)
04:20 PM seb_kuzminsky: so there's definitely a merge needed here
04:22 PM seb_kuzminsky: JT-Shop: did you offer those new boards and modules to micges as PRs and he rejected or ignored them?
04:23 PM seb_kuzminsky: i don't see any PRs, open or closed, that have that stuff in them
04:28 PM JT-Shop: pcw added that stuff, and no I did not do a PR
04:30 PM JT-Shop: I just provided pcw a way to work on mesaflash
05:53 PM andypugh: I suppose that not all LinuxCNC users need Mesaflash, and not all Mesafash users use LinuxCNC.
06:05 PM andypugh: It’s a bit annoying how you can actually have a 100% clean build locally, and still see 50+ warnings on the buildbot
06:50 PM cerna: What's wrong with forking that repository?
07:02 PM seb_kuzminsky: cerna: there are two problems, i think
07:02 PM seb_kuzminsky: 1. it makes it confusing for users - which one should they use?
07:03 PM seb_kuzminsky: 2. it dilutes developer effort - we have to pay attention to the other repos and port fixes and features from theirs to ours, or risk re-fixing bugs and re-developing features already done elsewhere
07:03 PM seb_kuzminsky: having everyone work on one repo together makes both those problems go away
07:07 PM andypugh: I _hate_ this xhc-wb04-6 driver! It has const me two full days
07:09 PM andypugh: And I never intend using one.
07:10 PM seb_kuzminsky: andypugh: i'm making a buster-amd64 vm now to test your rtai & kernel debs
07:10 PM andypugh: I just can’t get the docs for it to build.
07:10 PM seb_kuzminsky: is it on a branch somewhere? is it in 2.8 (i'm a bit out of the loop)
07:11 PM andypugh: Yes, it is in 2.8. Sadly. http://buildbot.linuxcnc.org/buildbot/builders/1640.rip-buster-rtpreempt-amd64/builds/301/steps/compile/logs/warnings%20%2862%29
07:11 PM seb_kuzminsky: yikes
07:14 PM seb_kuzminsky: well there's a docs/man/man1/xhc-whb04b-6.1 in the repo, and a docs/src/man/man1/xhc-whb04b-6.1.txt
07:14 PM seb_kuzminsky: that doesn't seem right
07:15 PM andypugh: Hmm, the one in docs/man should have been deleted by my most recent commit
07:15 PM seb_kuzminsky: oh, i'm out of date, sorry
07:17 PM seb_kuzminsky: it builds clean on my buster machine, why not on the buildbot?
07:21 PM andypugh: I wish I knew
07:21 PM seb_kuzminsky: oh, it's in the html version of the manpage
07:21 PM seb_kuzminsky: the manpage itself builds fine
07:21 PM andypugh: It might build fine, but it has some terrible formatting issues
07:22 PM andypugh: But I am prepared to ignore that
07:23 PM seb_kuzminsky: yeah, the manpage looks ok, but the html looks like hot garbage
07:25 PM seb_kuzminsky: what kind of name is xhc-whb04b-6
07:27 PM andypugh: It’s the name of the device
07:27 PM andypugh: <shrug>
07:37 PM andypugh: I would have expected the asciidoc tables to translate to HTML better than to manpage. Clearly not.
07:39 PM andypugh: I am tempted to delete the whole driver.
07:49 PM jepler: I think I fixed the 2.8 debian package problem after many permutations. by moving update_ini to where it gets fixed during the build to whatever configure --with-python= used
07:50 PM jepler: andypugh: looks like some of those warnings are only seen in a uspace build, I guess you are probably building rtai
07:50 PM jepler: I'll have a look, I love stomping out warnings
07:51 PM andypugh: I think I already have them
07:52 PM andypugh: Just waiting for runtests to finish afore pushing
07:52 PM jepler: cool
07:52 PM jepler: what did you determine about memset(task)? that one has me scratching my head, and I wrote it..
07:52 PM andypugh: Though not the “missing section” warnings in the docs.
07:53 PM andypugh: Well, untested, I am casting it to a char* so that the compiler doesn’t notice that it is a complex type.
07:54 PM jepler: oh, I think there's a deeper problem there :-/
07:54 PM andypugh: As that worked for the very similar “memcpy” case
07:55 PM andypugh: The real fix is for the classes to define their own clear and assign behaviour. But that requires an inderstanding of C++
08:00 PM jepler: Yeah in this case I think that memset should just be deleted. I'll push such a change to 2.8.
08:00 PM jepler: I'll wait until after you merge yours so we don't collide
08:01 PM jepler: so cool to have so few compiler warnings in the C(++) code now.
08:02 PM andypugh: I was hoping for _none_ but that seems unachievable.
08:02 PM andypugh: Anyway, I pushed, so you can have free rein
08:02 PM jepler: thanks!
08:09 PM jepler: https://gwolf.org/debian/raspberry/2020/04/21/fresh-daily-built-debian-images-for-your-raspberries-finally.html
08:09 PM jepler: not that I'm likely to switch my script from raspbian, it's still nice that people are working on Proper Debian
08:10 PM andypugh: Why does git ignore docs/man/man1/streamer.1 but not docs/man/xch-wh04b-6.1 ?
08:11 PM andypugh: Hmm, maybe it’s because it used to exist?
08:12 PM jepler: git grep streamer -- "**/.gitignore"
08:12 PM jepler: docs/man/.gitignore:man1/halstreamer.1
08:13 PM andypugh: I did a search of .gitignores and missed that one.
08:15 PM andypugh: Thanks
08:19 PM Tom_L: andypugh, fwiw there are at least a couple wanting to use the xhc-whb04b-6
08:20 PM andypugh: But they are wrong and misguided
08:20 PM Tom_L: we may all be a little misguided
08:21 PM Tom_L: just saying, it _would_ get used
08:21 PM andypugh: Not with these docs it won’t
08:21 PM Tom_L: heh
08:22 PM andypugh: http://linuxcnc.org/docs/2.8/html/man/man1/xhc-whb04b-6.1.html
08:22 PM andypugh: Not very helpful
08:23 PM Tom_L: it should be somewhat close to the xhc_hb04 one
10:10 PM jepler: did the toolchanger tests get flaky? Two builds ago of 2.8, all was OK. Then in the next two builds there were 4 failures out of all the builds. two different tests failed, and it wasn't consistent.
10:11 PM jepler: http://buildbot.linuxcnc.org/buildbot/builders/1306.rip-precise-amd64/builds/6681
10:11 PM jepler: http://buildbot.linuxcnc.org/buildbot/builders/1640.rip-buster-rtpreempt-amd64/builds/303
10:11 PM jepler: http://buildbot.linuxcnc.org/buildbot/builders/1306.rip-precise-amd64/builds/6682
10:11 PM jepler: http://buildbot.linuxcnc.org/buildbot/builders/1500.rip-jessie-i386/builds/3306
10:28 PM dgarr: in this: http://buildbot.linuxcnc.org/buildbot/builders/1610.rip-stretch-rtpreempt-i386/builds/1698/steps/runtests/logs/stdio
10:28 PM dgarr: i notice in 2.8 tests/motion-logger/startup-gcode-abort has a skip file
10:28 PM dgarr: but in master tests/motion-logger/startup-gcode-abort has NO skip file
10:29 PM dgarr: i have not been able to reproduce the other errors
10:36 PM dgarr: the skip was removed at 1680847a9 so maybe problem with stags merge? i dont know
10:58 PM dgarr: to be specific, i cannot reproduce the motion-logger/startup-gcode-abort on master (noskip file) nor the tool-info/* on 2.8 (tried buster and wheezy (vm))
11:12 PM seb_kuzminsky: andypugh: can you add virtio devices to your buster rtai kernel config?
11:33 PM seb_kuzminsky: oh, and don't rename the files, .deb and .dsc packages have structured filenames that are ugly but important