Perl 6 - the future is here, just unevenly distributed

IRC log for #pdl, 2015-02-26

| Channels | #pdl index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:38 sivoais jberger++ for the GitHub org perms. I'm going to get the Travis-CI going
00:43 mohawk sivoais++
00:43 mohawk sivoais, has your commit bit come through? :-)
00:44 sivoais yeah :-) chm++
00:45 * mohawk high-fives sivoais
00:57 sivoais Travis-CI is "Offline for maintenance"
00:57 mohawk oh dear
01:43 mohawk Mithaldu, would it cause a problem to make the PDL git mirror happen every 5 or 10 mins, not every 30?
02:00 mohawk my new "coretest" target builds PDL core and tests it from clean in under 9 secs
02:00 mohawk that's quite useful
02:20 mohawk oh god, Basic/Core/Makefile.PL's postamble is a catastrophe
02:30 madhatter joined #pdl
02:31 mad_hatter sivoais, hello to you
02:33 sivoais hi mad_hatter! o/
02:35 mohawk hi mad_hatter, how goes it?
03:13 opkick [pdl] wchristian pushed 3 new commits to master: http://git.io/ApGc
03:13 opkick pdl/master 2bebe39 Ed J: Implement "core" and "coretest" targets for quick testing
03:13 opkick pdl/master 1a3c4ea Ed J: nmake fixes for de.PL of Core.xs
03:13 opkick pdl/master 9704fb8 Ed J: Fixes for "perl in space"
04:38 mad_hatter joined #pdl
04:43 opkick [pdl] wchristian pushed 1 new commit to master: http://git.io/Apai
04:43 opkick pdl/master a504d4d Ed J: Bring MANIFEST.SKIP up to date
05:43 opkick [pdl] wchristian pushed 2 new commits to master: http://git.io/ApX8
05:43 opkick pdl/master 5a1b33f Ed J: Update .gitignore
05:43 opkick pdl/master 87b63c7 Ed J: Update MANIFEST.SKIP
06:04 drrho joined #pdl
08:31 mohawk i think my superfast "coretest" may actually have been loading already-installed code, oops
08:48 mohawk so building all the stuff in PDL::LiteF actually takes 2 minutes
08:48 mohawk still, quicker than full build
09:13 opkick [pdl] wchristian pushed 1 new commit to master: http://git.io/AhcL
09:13 opkick pdl/master 9f289b6 Ed J: "coretest" to build PDL::LiteF's Core Gen Ops Primitive Ufunc Slices Bad
11:13 opkick [pdl] wchristian pushed 1 new commit to master: http://git.io/Ahd9
11:13 opkick pdl/master 255adcf sisyphus: Basic/Core/Core.pm - avoid negative repeat count warning in PDL::dummy()...
12:55 Mithaldu mohawk: the synch now runs every 5 minutes
15:01 chm joined #pdl
15:02 chm Hi all, catching up from some previous #pdl conversations...
15:04 chm @mohawk: The LLDF branch was a development branch with the hope other developers would assist.
15:05 chm No one stepped up and my PDL cycles ran out last Fall to continue, it was not abandoned.
15:05 chm I had hoped it would go more quickly but got stuck with some of the edge cases in the XS layer.
15:06 chm Again, no one helped out so things only moved forward when I had time.
15:07 chm @mohawk:  I appreciate your folding the commits from core-cleanup and LLDF
15:07 chm Those we made under the assumption that the LLDF work would not take as long as it has.
15:07 chm Again, lack of other developers and more XS-pertise made the last fixes slow to find.
15:08 chm @mohawk:  That is a great tip re the by commit testing.  Please add to the web site.
15:10 chm @all:  Please use a bit more detail in your commit messages, one line is fine if clear and unambiguous
15:10 Mithaldu commit -m "like, stuff"
15:10 chm @all: For more complex commits, please add some more context and info to explain things.
15:11 Mithaldu i found a good format is making commit messages like instructions
15:11 chm PDL developers shouldn't have to read the source changes step-by-step and be on #pdl constantly
15:11 Mithaldu and, unless abundantly clear, spending most of the message on explaining "why"
15:11 chm to understand what the commit is and why/how/...
15:11 Mithaldu example:
15:12 Mithaldu in microidium at one point i tweaked the numbers of the camera algorithm
15:12 Mithaldu this resulted in the camera shaking around less in response to the spastic inputs of AI that could change its mind 60 times a second
15:12 Mithaldu the message was:
15:12 Mithaldu reduce nausea creation via camera when under ai control
15:14 Mithaldu also, under "seven rules", this document explains a great format for big messages: http://chris.beams.io/posts/git-commit/
15:16 chm @Mithaldu:
15:16 chm @Mithaldu: Excellent reference and examples.  A specific problem for PDL-2.008 is that many commits are useless for generating Release_Notes
15:17 Mithaldu this repo, yes? git://git.code.sf.net/p/pdl/code
15:18 chm More substantial commit messages can help other developers to participate by indicating remaining issues, problem understanding,...
15:18 Mithaldu yeah, i'm also seeing things like:
15:18 Mithaldu * Bring MANIFEST.SKIP up to date
15:18 Mithaldu * Update .gitignore
15:18 Mithaldu * Update MANIFEST.SKIP
15:20 Mithaldu no offense to the work done, but that should've been rebased into one commit explaining that it cleans up eumm/travis artifacts and removes entries for deprecated files
15:20 chm @Mithaldu:  The verbosity isn't as much a problem as when the 1-line subject omits the fact that a whole directory and Makefile.PL have been deleted.
15:20 Mithaldu haha
15:20 Mithaldu is derek lamb in here?
15:21 Mithaldu chm: meanwhile something that might help you is to use a log display tool that shows action summaries
15:22 Mithaldu my tool has a column where it shows x/4 icons, those icons being: file was changed, file was added, file was deleted, file was moved
15:24 chm Mithaldu: That sounds useful.  Is this your own tool or an opensource or product tool?
15:24 Mithaldu it's tortoisegit :)
15:24 chm oh, I guess I use cygwin rather than native windows so I don't use tortoisegit
15:25 Mithaldu https://www.dropbox.com/s/9qi957ddt87vo80/Screenshot%202015-02-26%2016.24.55.png?dl=0
15:25 chm I'm sure there are other comparable tools out there and I look forward to improving my/our work flow...
15:25 Mithaldu chm: if you actually are on a windows machine, i VERY strongly recommend two tools that might help you:
15:26 Mithaldu http://ghisler.com/ <- that's a tabbed dual-pane file manager with integrated commandline that is the adult to windows explorer's pre-school child
15:26 Mithaldu and tortoisegit integrates with it beautifully
15:27 Mithaldu the only thing that requires a *little* adaption time is the long-right-click to get the context menu
15:28 chm Nice, I'll give it a try.
15:28 Mithaldu if you remember norton commander, it's that
15:29 chm I remember the name but didn't ever use it.
15:29 chm Nice chatting, I have to get to some PDL::NiceSlice work.  o/
15:29 Mithaldu ah, when i was using dos everyone did :)
15:30 Mithaldu have fun :D
15:59 mohawk Mithaldu, problem with squashing all my M.SKIP and gitignore changes into one was that i did in fact do that, then further work revealed further problems
16:00 mohawk and if you want to see not very accurate to what was done commit messages, look on core-cleanup and LLDF :-)
16:02 Mithaldu mohawk: those three commits are literally in sequence
16:02 Mithaldu even if you had already pushed, there's no reason to not force-push
16:15 mohawk force push on master?
16:15 mohawk i don't think so
16:18 mohawk i think part of the answer here is for all of us to start working on topic branches
16:18 mohawk and only merging them (WITH ff-only) when they're ready
16:24 chm mohawk: I acknowledge the mess in the CC and LLDF but I think those are what you call "topic branches" so it would be more productive to set up/educate best practices going forward
16:25 chm mohawk: Given the work you have done with merging in some of the CC and LLDF into master, what would you say to returning the final 64bit support fix to PDL-2.008?
16:32 mohawk chm, fear not! i am going to document how to do all these things
16:32 mohawk other examples of less-ideal practice are how eg Basic/Core/Makefile.PL's special cases on windows / different version of EUMM accumulated
16:33 mohawk dozens and dozens of lines of code partly copy-pasted
16:33 mohawk leading to a maintenance nightmare
16:33 mohawk i think going forward that even minor changes, like my coretest or skip/ignore files ones, should be topic-branched, and reviewed by one other
16:34 mohawk that will then stop keen people (like me) pushing prematurely
16:34 mohawk chm, i am broadly in favour of the idea, but since the code isn't currently there, i counter-propose that i extract all the good cleanup / comment-adding stuff
16:35 mohawk then release 008
16:35 mohawk then finish up 64 bit and release 009 in a month or two
16:35 mohawk remember, no-one's charging us for version numbers :-)
16:35 sivoais ^ I agree with mohawk on using review. It's very helpful and helps keep everyone in the loop
16:36 mohawk re topic branches, i have a 3-stage topic branch on the go for a major rework of PDL build (only 1 Makefile.PL!), that i've been tuning and rebasing for a month or two
16:36 mohawk so i do practise what i preach
16:36 mohawk but i need to raise my game and set a better git example :-)
16:36 Mithaldu 17:15:40 (mohawk) i don't think so # look, i don't want to be mean, but that statement is naive, idealistic and ignorant of certain realities
16:37 mohawk Mithaldu, i am saying that force-pushing to a master branch is bad practice
16:37 mohawk what realities are you referring to?
16:38 Mithaldu that whether it's a bad practice depends on how badly the repo is fucked up, and how many people are working on it, and how likely any of them are to see the fuckup, etc. etc. etc.
16:38 Mithaldu there are MANY variables to this
16:38 Mithaldu and it's only a bad practice if it is indeed a practice, and not a thought-out application of an available tool if the circumstances merit it
16:38 Mithaldu in short:
16:38 Mithaldu don't think black/white yo
16:40 Mithaldu note: mst could probably put this better than i do :v
16:47 mohawk i submit that 3 small commits on one similar aim is not really the end of the world
16:47 mohawk and therefore i submit that the circs did not, and do not, merit it now
16:49 Mithaldu now that there're 5 more after them? no
16:50 Mithaldu at the time? yes absolutely
16:50 Mithaldu (ok, 2 commits, but both of them a good bit more serious and one of them from someone else than you, which is the most important factor)
17:14 mohawk i think the inherent cooling-off in getting someone to review will solve the problem
17:15 Mithaldu sure, if you follow that model, it helps
17:15 Mithaldu Moo does that
17:16 mohawk if they agree with me, they must be right
17:16 Mithaldu occasionally (like 3 times over its lifetime) someone still fucks up and a force push is necessary
17:16 Mithaldu i've no idea why you said those words or what you're trying to communicate
17:17 sivoais sarcasm, I assume ;-)
17:18 mohawk a bit of ironic self-deprecation
17:18 Mithaldu sarcasm is a much higher violation of anything than a force push could ever hope to aspire to
17:18 Mithaldu i do carry the hope that mohawk would not stoop that low
17:18 mohawk the serious version is that if a large project does it, then my surmise that it's a good practice is probably right
17:18 mohawk i was not in fact stooping that low
17:19 * mohawk hugs Mithaldu - everything will be ok
17:19 Mithaldu i mention Moo because i work on it and have both seen such things caused and caused them myself :v
17:19 Mithaldu from my POV Moo is a tiny project
17:19 mohawk what do you think of as a large one?
17:19 mohawk (tee hee, innuendo)
17:20 Mithaldu something with more than a handful of regular contributors
17:20 Mithaldu or something that actually has a lot of code
17:20 Mithaldu Moo is basicalle Moose::Tiny
17:20 Mithaldu *y
19:58 drrho joined #pdl
20:33 drrho joined #pdl
21:42 drrho joined #pdl
22:11 mad_hatter joined #pdl

| Channels | #pdl index | Today | | Search | Google Search | Plain-Text | summary