#linuxcnc-devel Logs

Dec 03 2017

#linuxcnc-devel Calendar

02:44 PM seb_kuzminsky: dgarr (if you read this): the 285,361 push notification never made it from github to the buildbot, i think due to flakiness on the part of my internet connection
03:23 PM jepler: I just saw https://groups.io/ recommended as a potential replacement for sourceforge mailing lists
03:23 PM jepler: they have a free tier for open source software
03:53 PM seb_kuzminsky: .io is so hot right now
04:00 PM hazzy: jepler: That looks very nice, the interface for browsing/searching the archives also seems much better than sourceforge
04:33 PM skunkworks: so - what would it take to get reverse run into master? I have tested it quite a bit.
04:34 PM skunkworks: like https://www.youtube.com/watch?v=u9QqVx3_RgA&t=28s
04:36 PM skunkworks: Also robs spindle synced motion improvements...
04:38 PM rene-dev: along with my g33.1 improvements ;D
04:38 PM skunkworks: ^ I think that is what I am talking about
04:40 PM hazzy: Also state-tags into master
04:40 PM rene-dev: what are state tags?
04:40 PM skunkworks: :)
04:43 PM rene-dev: what does reverse run do on spindle synced moves? on probes? on toolcahnges? on other remaps?
04:43 PM hazzy: In master and previous linuxcnc versions the various states (active gcodes, ect) may not be accurately reported during a running program
04:43 PM hazzy: https://github.com/LinuxCNC/linuxcnc/tree/feature/state-tags-master
04:44 PM hazzy: State tags tryes to fix those problems
04:45 PM rene-dev: ah, yes. I think I have seen an issue about that
04:47 PM skunkworks: rene-dev, I had tried tool changes - it stops.
04:48 PM Tom_L: skunkworks that would cause a climb vs conventional situation that could cause the cutter to dig into the piece making it out of tolerance
04:48 PM Tom_L: i can see the handiness of it though
04:49 PM seb_kuzminsky: reverse-run should probably stop on spindle-synced moves, like it apparently does for probes
04:49 PM seb_kuzminsky: does the feature branch come with tests?
04:50 PM rene-dev: I did see a test in the diff
04:51 PM seb_kuzminsky: did you read the code? i have not... how did it look to you?
04:52 PM skunkworks: I think the code looks great! lots of semicolons and such...
04:52 PM * skunkworks backs away...
04:52 PM seb_kuzminsky: lol
04:53 PM seb_kuzminsky: it's sure important that there's the right number of semicolons
04:54 PM rene-dev: I quickly scrolled thru the state tags diff... where is the reverse run?
04:55 PM skunkworks: https://github.com/LinuxCNC/linuxcnc/tree/feature/reverse-run-2.7
04:57 PM skunkworks: actually - I am not sure which one is the lastest.. There is a rr master too
04:57 PM rene-dev: yes, there is a test
05:04 PM rene-dev: any particular reason they dont get merged? at least then more people will test it...
05:20 PM seb_kuzminsky: no good reason as far as i know
05:21 PM seb_kuzminsky: fear of breaking existing functionality maybe
05:22 PM seb_kuzminsky: since i haven't read the patches i have no knowledge of the quality of the changes proposed
05:22 PM seb_kuzminsky: the input from skunkworks and rene-dev is thus valuable to me
05:24 PM rene-dev: In my opinion the point of master is to break things, so they can get fixed. its only used by pepole who know what they are doing
05:25 PM seb_kuzminsky: i disagree, i think users should be able to rely on master almost always working
05:25 PM rene-dev: there are always stable releases if you need to have something reliable
05:25 PM seb_kuzminsky: breakage belongs on feature branches, and it should be fixed there before merging into master
05:25 PM rene-dev: yes, but look what happens. the patch has been sitting there for over 2 years, and no one even knows if it works
05:25 PM seb_kuzminsky: yeah that's also super suboptimal :-(
05:25 PM rene-dev: but no one uses feature branches, unless you need that specific feature
05:26 PM seb_kuzminsky: yeah, imo feature branches are the for the developers and testers of that feature
05:27 PM rene-dev: yes, but its better if more people test :D
05:27 PM seb_kuzminsky: i certainly agree more testers is much better
05:27 PM rene-dev: because feature branch users will only test the feature, any might not notice what else it breaks
05:27 PM rene-dev: another way would be to have a dev branch, which gets everything, and stuff then gets merges into master
05:27 PM seb_kuzminsky: and in many software projects i agree that breaking master is ok
05:28 PM seb_kuzminsky: but i feel different about software that controls machine tools (and other robots, i guess)
05:28 PM seb_kuzminsky: i think main the problem is that we don't have enough volunteer-hours to review incoming PRs :-(
05:29 PM rene-dev: not using it is a good way of making sure it does not break :D
05:29 PM seb_kuzminsky: i don't understand what you mean
05:30 PM rene-dev: look at this: https://github.com/LinuxCNC/linuxcnc/issues/167
05:30 PM rene-dev: or this: https://github.com/LinuxCNC/linuxcnc/issues/164
05:30 PM rene-dev: they have both been sitting there for over 1 year
05:31 PM seb_kuzminsky: yes, i agree that's terrible of us
05:31 PM rene-dev: I understand what you say, but I think the yolo factor of master should be a little bit higher
05:31 PM rene-dev: Im not blaming you...
05:31 PM rene-dev: I know its hard to test this stuff
05:33 PM seb_kuzminsky: it's not super hard to test. we have a pretty good test framework. PRs that add tests that cover the feature or bugfix are wonderful
05:34 PM rene-dev: look at joint-axis. only after it has been merged many pepole started using it and slowly more and more bugs are fixed.
05:34 PM rene-dev: and that was a rather bug change
05:34 PM rene-dev: big
05:34 PM seb_kuzminsky: many, many bugs were found and reported, and fixed (mostly by dgarr) before it merged into master
05:35 PM rene-dev: agree...
05:36 PM seb_kuzminsky: i think it's right that we try really really hard to not crash machines, break tools, and ruin work, even in master
05:37 PM seb_kuzminsky: that said, i share everyone's frustration that we have long-standing PRs that have not been reviewed and merged, and long-standing bugs that have not been addressed
05:38 PM seb_kuzminsky: i think you're saying: we should be more trusting of feature-branch developers, and lean more towards merging incoming PRs
05:38 PM seb_kuzminsky: you may be right.
05:40 PM seb_kuzminsky: it sure seems like what we're doing currently isn't working that well, for users, collaborators (people who send us bug reports and PRs), or for the core devs (us with push access to the repo)
05:40 PM seb_kuzminsky: i'm afk for a bit
05:40 PM rene-dev: so what do you think about my idea of having another branch that just gets everything, for pepole who are willing to do even more bleeding edge testing?
05:41 PM rene-dev: or another way would be to call master 2.8-beta
05:41 PM rene-dev: and then put all the bleeding edge stuff into master
05:41 PM Tom_L: master should be the most stellar code written to date :)
05:43 PM Tom_L: bleeding edge may be a reality on a multi ton piece of hardware
05:44 PM Tom_L: but i suppose those guys wouldn't risk it either
05:44 PM rene-dev: do you have a better idea? and like I said, if you need reliable stuff use the release version.
05:45 PM Tom_L: i'm not sure i completely agree
05:45 PM rene-dev: its not like you run "git pull" and then press cycle start every morning
05:45 PM Tom_L: that's what the test branches are for
05:46 PM rene-dev: but no one tests them
05:46 PM Tom_L: they may not be aware of them to test
05:46 PM rene-dev: yes, have you seen how long the list of branches is?
05:47 PM Tom_L: i saw your post on the reguar channel a few days ago for a test but i didn't have anything working to test with
05:47 PM Tom_L: maybe some should be weeded out
05:47 PM rene-dev: even I dont know which ones still need testing, and which feature is where
05:48 PM Tom_L: i'm not sure anyone does except the ones that make the branch
05:54 PM rene-dev: thats one of the problems
05:56 PM skunkworks: on the new TP I solicited testers... I just don't have time right now.
06:06 PM skunkworks: but that was because I really really really really wanted it
06:11 PM rene-dev: whats the application?
06:11 PM rene-dev: wire edm?
06:13 PM skunkworks: that is the main one yes.
06:14 PM rene-dev: retrofit or homemade?
06:16 PM skunkworks: sort of home made. have a sinker edm that I would like to add wire to play with.
06:17 PM rene-dev: its also useful for sinker edm
06:18 PM skunkworks: sure
06:19 PM skunkworks: right now the sinker part (hydraulic) seems to work just fine
06:19 PM skunkworks: Hey - your the stmbl guy!
06:20 PM seb_kuzminsky: hmm, the github irc integration is broken
06:21 PM skunkworks: I tried to build some PWM amps - then some AMC drives showed up on ebay and I grabbed them. Never looked back. I would like to try yours though. Smart serial interface is awesome.
06:21 PM seb_kuzminsky: github says "Last delivery was not successful. Service Timeout."
06:22 PM seb_kuzminsky: the Travis CI and Email are working, and the buildbot webhook
06:22 PM rene-dev: it is. full digital communication to drives is so nice to have.
06:25 PM rene-dev: ah, hydraulic. I was thinking 4 axis sinker :D
06:27 PM rene-dev: seb_kuzminsky I will make a branch with all those feature and PRs that are done, but just need more testing.
06:27 PM rene-dev: but that is where the problem continues, there is really not a list of all those.
06:28 PM seb_kuzminsky: the list that matters is the list of github PRs
06:28 PM seb_kuzminsky: any other branches are either obsolete/historical, or incomplete (possibly abandoned) feature branches (this is what i believe)
06:30 PM seb_kuzminsky: i think we as a project should talk about if/how we should change the way we merge PRs
06:32 PM rene-dev: I think the way to go is to have a branch with all of them, so people can test it, or change the master to be like that, and release more often.
06:32 PM rene-dev: and when pepole use that branch, and their machine catches on fire, its their fault.
06:33 PM rene-dev: but how else will you find out?
06:33 PM rene-dev: I sometimes have to blow up stmbls to see what the hardware is capable of.
06:35 PM Tom_L: if you toss them all together it may be hard to know which one broke it
06:35 PM rene-dev: I think one important part of testing a feature is not only testing the feature, but also checking everything else still works as expected. like the probe-jog bug I found.
06:36 PM rene-dev: Tom_L well, that is what git is for. do you think robs spindle improvements will get any better by sitting there for another year?
06:37 PM rene-dev: Im talking in particular about those 3 PRs, because they all go together...
06:45 PM seb_kuzminsky: i think the most helpful thing any volunteer can currently do is inspect individual PRs, read the code and check it for correctness and general quality/hygiene, write any missing tests, and report to the author of the PR and to the general community of developers what they've learned about that PR
06:46 PM seb_kuzminsky: but of course i haven't made time in my life to do those things, so my opinion shouldn't count for much here
06:47 PM seb_kuzminsky: i'm very reluctant to break master - if it sometimes crashes peoples machines, i think people will not run it and we'll not get the benefit of their testing, and instead we'll get a reputation for crashing machines
06:48 PM rene-dev: I understand that. But I will still make the branch, because I need those 3 related features next week :D
06:48 PM rene-dev: while im complaining about stuff, this can be closed :D https://github.com/LinuxCNC/linuxcnc/issues/28
06:48 PM seb_kuzminsky: i've got to run, i'll look at it when i get back
09:42 PM jepler: thanks seb_kuzminsky