IRC Logs for #crux-devel Monday, 2018-06-11

jaegerfrinnst, teK__, jue: serious question - if one of us ever did actually get a rewrite of pkgutils done, would you guys seriously consider switching to it?02:00
jaegerI don't feel like any of us really want to learn/maintain c++ all that much02:00
jaegerAnd while patching small things in the current codebase isn't too tough, it's not ideal02:01
jaegerAlso, when I say a rewrite of pkgutils, I'm not thinking of a straight port to C like tilman's pkgutils6 project, I mean a rewrite/merge of pkgutils' and prt-get's functionality02:01
*** _________mavric6 has quit IRC02:50
*** _________mavric6 has joined #crux-devel02:51
groovy2shoesI know you're not asking me, but fwiw, I like having pkgutils as a separate entity from prt-get, though I couldn't care less about the internal architecture beyond that, though I must ask:  why C?03:41
jaegerBecause none of us are c++ folks03:43
jaegeralso, C doesn't introduce any additional deps03:43
jaegerside note, what's the benefit of having the tools separate?03:44
jaegerPersonally, I think base pkgutils' lack of dependency resolution is a bad thing, for example03:46
groovy2shoesI like having the simpler tools around for when I feel a need/desire to do some "surgery"03:52
jaegerI'm not suggesting any functionality be removed03:54
groovy2shoestbh I prefer C to C++ any day of the week, so I didn't mean to sound judgmental or anything, and tbh I find the attitude refreshing these days, when everyone wants to write everything in Go or decides that it's acceptable to make me install a zillion tiny deps with npm or something else trendy and insane  ¬_¬03:56
jaegerI may be weird but I enjoy the manual care that C needs03:57
jaegergood mental exercise, etc.03:57
groovy2shoesbut C also strikes me as kind of cumbersome for such a tool, where a shell script could work as well with less code—but then there's always that issue of dependencies ¯\_(ツ)_/¯03:57
jaegerI've done some go and I've used python for years, but for something like pkgutils I'd prefer C03:58
jaegerI'd leave pkgmk as a shell script03:58
jaegerbut pkginfo, pkgadd, pkgrm, prt-get stuff, none of that's shell currently03:58
groovy2shoesnah, it doesn't sound weird to me at all, and I get really tired of having to reverse engineer every little packaging ecosystem for every little language when something inevitably goes wrong (most recently haskell, iirc)03:59
jaegerheh03:59
jaegerI took a haskell course in college, not my thing04:00
groovy2shoesI don't know the language very well, seems fine enough to me, and I'm too busy to spend any time learning it, but I wanted to install pandoc...  instructions seemed easy enough:  just run `cabal install pandoc` and wait for a few hours, but no mention of what to do if that fucks up, which it did, of course04:02
groovy2shoes(it's not particularly thrilling to write Pkgfiles for hackage packages, either, so I wound up writing just enough to get a working cabal-install package, use that to get the rest)04:04
jaegeryeah, that can be annoying04:04
groovy2shoeswhew, my texlive patch appears to compile cleanly...  now to make sure it doesn't segfault...04:28
groovy2shoesit's about that time of year for a new texlive release, so maybe it won't be needed anyway...  how long has poppler been out that the devs still find it okay to make breaking API changes willy-nilly?  *sigh*04:30
groovy2shoesnow I don't even remember why I was trying to recompile it in the first place  ¬_¬04:33
jaegerheh04:36
groovy2shoesoh right, cuz the perl update04:49
groovy2shoeswell, the patch works for pdftex, but for luatex it will segfault during garbage collection if compiled with non-default optimization...  I can't figure out why, but after rewriting the problematic code and staring at gdb for an hour, I'm ready to throw in the towel and pretend like it's the FFI's fault  ¬_¬07:33
groovy2shoesinterestingly the luajittex tests all pass either way07:35
groovy2shoespatch is here if anyone wants it:  http://ix.io/1cYD  (it includes the changes from the patch that was already in the port dir)07:40
frinnstjaeger: sure - if it ticks all the boxes that the current pkgutils does08:39
frinnstlike no major dependencies etc08:40
pedjajaeger, if you do pkgutils rewrite, please implement fs#849 :)11:17
pedjahow would pkgutils/prt-get merge work? pkgutils as a prt-get sub-commands, or still separate, just sharing the code base?11:23
*** crash_ has joined #crux-devel12:22
jaegerI imagine something like prt-get (though probably with a new name) that supports all the major commands like install, update, sysup (all of which would handle deps), info, etc.13:50
pedjaI wonder if switching pkgdb from text file to sqlite db would gain anything13:56
jaegerI don't think sqlite is ideal since you can sometimes corrupt it by reading from it :P But I have considered maybe using something else, if it would gain some useful functionality or robustness14:00
*** onodera has joined #crux-devel17:07
darfoi find the pkg db hard to parse but I see it is really easy in perl. see prtcheckmissing. I don't do perl so had to spend a while in the docs to figure that one out.20:59
darfoprtcheckmissing has a bug with symlinks21:00
darfoi don't know perl well enough to fix the while loop logic21:00
darfoLarry Wall's stance the if Perl code needs fixing just rewrite is fine for perl wizards21:02
*** onodera has quit IRC21:33

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!