<farkuhar>
jaeger: the signature mismatch only occurs if you use the pkgmk that doesn't handle renaming.
<farkuhar>
which i guess would affect everybody still running crux 3.6.1. All the more reason to call 3.7 done and let everyone else migrate to the branch you've been testing for the last several months.
ivandi has quit [Quit: WeeChat 3.5]
ivandi has joined #crux-devel
groovy3shoes has joined #crux-devel
groovy2shoes has quit [Read error: Connection reset by peer]
<jaeger>
Figured it was something like that, I'd prefer not to have breaking changes in a still-current version like that
<beerman>
ah my bad, i had it all in my overlay, and i tried avoiding it, but imo that patch should've been made available in 3.6.1 the moment we had it and decided to merge for 3.7..
<farkuhar>
on the other hand, pkgmeek will accept git urls in the source array, so you could instead do something like this: https://dpaste.com/4FEEALCRH
<beerman>
thats fine if you want to build master..?
<beerman>
does it checkout tags as well or..?
<farkuhar>
I'll have to look into that. At the moment it just recognizes urls that end in .git
<farkuhar>
it would be helpful to have an example of the url format you'd like to use, when checking out a specific branch. then i'd be able to add the appropriate regexp to the pkgmeek function that parses urls.
<beerman>
i don't have any thoughts on it, i think arch does it with a #1.10.0 for example
<farkuhar>
okay, that's something i can work with
<beerman>
i still haven't changed my opinion on pkgmeek btw, i haven't tested it and only will as soon as basically every core maintainer agrees on it (maybe excluding romster because obvious reasons)
<farkuhar>
i agree, you already do so much for the distribution, there's no need to take on extra work for something that might never be adopted.
<farkuhar>
btw, does Fun ever lurk in these channels? i'd like to offer some thanks for the strong foundation he left by writing upkgmk.
<beerman>
i don't expect it, I wouldn't know though
<beerman>
maybe reach out to him on gitlab or something
<farkuhar>
it was an inspiring example of "do one thing and do it well", but i think it was too austere (cutting out too much functionality that maintainers had come to expect from pkgmk). I built upon the upkgmk foundation, trying to add only the bare minimum of code needed to implement the expected features.
<beerman>
i feel like his input is valuable as well, but it might just be me in that point
<farkuhar>
my working idea is to repurpose the 'renames' array, so that if a git branch other than 'master' is wanted, you would write (for example) 'fcft/tag-1.2' in the corresponding slot of the 'renames' array. The slash would serve as a separator between the desired name on disk, and the tag that you want to check out.
<farkuhar>
in keeping with the current behaviour, you would be able to default to the 'master' branch, by writing 'SKIP' in the corresponding slot of the 'renames' array.
<beerman>
it becomes a general modifier
<beerman>
but then, renames is the wrong name for that array?
<beerman>
for source
<beerman>
nit picky, but just my 2 cents without further thought ;)
<farkuhar>
i took the name that Romster suggested in FS#923 and rolled with it.
<beerman>
that might be true, but git is a new feature that doesn't fit the renaming issue we tackled there?
<farkuhar>
right, the name is inextricably linked to the problem we tackled there, but as long as we have it available, why not find creative ways to let it solve other problems?
<farkuhar>
for instance i used it to solve the problem of having to redefine unpack_source() when you don't want specific tarballs unpacked. Just rename them so their file extension doesn't match the patterns that unpack_source() was looking for.
<beerman>
why not find creative ways to let it solve other problems? -> because KISS, this doesn't feel very KISS
<beerman>
it hides in another "feature"
<beerman>
that can't be simple
<beerman>
i would suggest to just change the name renames to something more direct like source_modifier or whatever
<beerman>
then define features of source_modifier to use
<farkuhar>
i'd be fine with naming the array something more general, if we do let it take on a multi-purpose role. Let's settle on something quickly before 3.7 is released and users start populating the portdb with Pkgfiles that use the old name.
<farkuhar>
i propose "save_as". it takes fewer characters to type than "source_modifier", and it will be familiar to anyone who's used desktop office suites.
<beerman>
save_as does not really fit the "checkout that specific tag, commit, ..." objective
<beerman>
imo, at least, maybe it does after a second thought
<farkuhar>
"save_as" actually honors one of the other contributors in FS#923: alan who offered a patch that parsed 'source' as an associative array.
<beerman>
honors are nice, but not everything ;)
<beerman>
if we would do honors we would have still kept the "packager" line in Pkgfiles, which was dropped, because it confuses folks to send their email someplace where they don't reach the actual maintainer
<beerman>
btw, we do honors, old maintainers still have sites up somewhere on crux.nu
<farkuhar>
when was the "packager" line dropped? running `prt-get cache` still generates a file that has 12 lines per port, one of those lines being "packager". Running `prt-get printf %P` will mostly show blank lines, but a handful of ports still provide a value for that key.
<beerman>
yeah those are Romster ports, Romster is kind of a port messy as you can see when you look up his personal ressources on crux, his personal repository etc
<beerman>
i wouldn't count those right now.. he is overdue with cleaning the f up
<beerman>
but looking at it, i have a few in there too
<beerman>
¯\_(ツ)_/¯
<farkuhar>
speaking of overdue, i'm heading outside to enjoy the day before i spend another marathon session staring at a screen. always great to chat with you, beerman!
<beerman>
its still not really anything that is done anymore
<beerman>
sure, i have a few to dos as well, enjoy your sunday
groovy3shoes is now known as groovy2shoes
SiFuh has quit [Ping timeout: 258 seconds]
SiFuh has joined #crux-devel
<jaeger>
Uploaded a new test ISO to https://crux.ninja/tmp/, please report if you see issues
<jaeger>
Romster: any objection to me updating xscreensaver? It looks like the upstream tarball for 6.02 is gone
<jaeger>
It looks like pkgmeek is trying the source URL from the Pkgfile *before* the ones listed in the PKGMK_SOURCE_MIRRORS array in pkgmk.conf - is that intended? It's reversed from pkgmk (and while I can't speak for everyone I prefer going to the mirrors first)
<farkuhar>
I did originally have it going to the mirrors first, the way pkgmk does. Not sure why I changed it, maybe I just didn't want to overload the mirrors during all my testing these past few weeks. :)
<jaeger>
:)
<farkuhar>
Anyway, easy enough to revert. It's on my to-do list now, along with restoring the omitted test that groovy2shoes pointed out on line 195.
SiFuh has quit [Quit: leaving]
SiFuh has joined #crux-devel
<farkuhar>
i just pushed an update to pkgmeek that lets it check out a specific branch of a git project, per beerman's request. Got a successful build of fcft using this: https://dpaste.com/92DNHFNND
<farkuhar>
also incorporated the feedback from jaeger and groovy2shoes. thanks again for testing!
<beerman>
yeah that looks fine, imo
<beerman>
while covering git, I might add that openjdk8 for example would need mercurial support
<beerman>
jaeger: cheers for the iso, i'll try to give it a go