IRC Logs for #crux-devel Tuesday, 2016-05-17

Worksterah cool so i wont need to take on more tasks that would be great.01:24
Worksterlast i read you said maybe not :P01:24
jaegerAt the time I was rather annoyed about the amount of work it would take. I ended up doing it anyway.01:24
Worksteryeah it is quite a mess. and i hit a few quirks over the span i am using docker.01:25
Worksterso it's picked up more dependencies i see01:26
jaegeryeah. Had to update go, add runc and containerd01:26
Worksterwhat's with the certs?01:26
jaegergo doesn't seem to work properly with our curl-style certs01:27
Worksterdoes that replace the existing ca-certs or append to it?01:27
jaegerIt works alongside01:27
Worksterand you're hosting containerd, was thatone of them horrible github archives?01:28
jaegeryeah01:28
jaegerteK_ said something about fixing that soon, though01:28
Worksteri'll test it out later, i rather like knowing where the original source is so i #source=() or something.01:28
Worksterarch has a nice way to rename archives, but it adds more to Pkgfiles01:29
jaegerwe'd do something similar, I imagine. something like source=(wtf-version.tar.xz:=https://github/uwotm8/v0.2.3.tar.xz)01:30
Workstergo introduces some weird paths on them symlinks01:30
jaegergo does a lot of weird stuff, heh01:30
Worksteronly problem is that'll break other scripts like my pkg-url i'd have to rewrite it to handle that.01:31
Worksterand how many other scripts in prtverify and friends.01:31
jaegerIt makes the most sense, though, in my opinion. Not like we can fix upstream all the time01:31
Worksterand additional checks for consistency01:31
Worksterthough it wont take much to fix any of those.01:32
Workstermust of pulled some hair out on these Pkgfiles...01:34
jaegerTook a while to get things right, definitely01:35
Worksterthat ca-certs-alt be merged in wiht core/ca-certs provided it all works?01:43
Worksterwith*01:44
Worksterand i'll test it as soon as i get the chance.01:44
jaegerNo technical reason they couldn't, though I doubt jue would like it01:45
*** _________mavric6 has quit IRC02:38
*** _________mavric6 has joined #crux-devel02:39
*** Workster has quit IRC06:25
*** Workster has joined #crux-devel06:25
frinnstwhy not just include ca-certs-alt with the go port (its go, right?)09:14
*** onodera has joined #crux-devel11:53
*** Nomius has quit IRC13:44
jaegerbecause it isn't go-specific, other things could use it too if they don't support the other style13:59
*** Nomius has joined #crux-devel14:10
*** Romster has quit IRC14:55
*** Romster has joined #crux-devel14:55
*** Nomius has quit IRC14:58
*** Nomius has joined #crux-devel15:05
*** Nomius has quit IRC16:25
*** Nomius has joined #crux-devel16:26
*** Romster has quit IRC16:52
*** Romster has joined #crux-devel16:52
*** Romster has quit IRC17:28
*** Romster has joined #crux-devel17:29
*** Romster has quit IRC18:37
*** Romster has joined #crux-devel18:38
frinnsthas anybody tried grub 2.02b3?19:47
frinnstnoticed it exists.. "i wonder what has changed since b2"19:48
frinnst.. 2 years :)19:48
jaegerI haven't yet... worth a look19:49
*** onodera has quit IRC21:45
teK_jaeger, Romster: I would have come up with something simpler in terms of Pkgfile syntax..21:49
teK_I would simply hardcode (yes! he said it! hardcode!) a check if the current source array elemnt contains github.com/ and then check if it starts with v or not and then just contains numbers and dots for the file name, if so... strip the v or not and prepend $name, done21:50
teK_which has the downside of a switch all or nothing or get hellish md5 mismatches21:51
teK_we have to discuss & decide :)21:51
teK_on the sunny side, there's no change required in the pkgfiles themselves21:52
jaegerI'd rather have something generic and not specifically aimed at github.com, myself21:52
teK_which other sites do have that problem?21:53
teK_but yeah, you could skip the check if the source elemtn contains github.com.. the second check is rather generic, I think21:53
jaegerNot sure of any others off the top of my head but why not do it properly from the start instead of possibly having to fix more cases later?21:55
pedjaI think I fixed the problem go has with ca-certificates by symlinking cert.pem to ca-bundle.pem21:56
teK_because there are more important fields we could add to Pkgfile and we dont but for this ugly clutch we do? Meh.21:56
teK_I am talking about the depends on "field" here21:57
jaegerpedja: interesting, that didn't work for me22:01
jaegerteK_: I'm not advocating we add any fields22:01
teK_no but changing the current format22:02
jaegeronly of items in the source line22:03
teK_which you also do if you'd add a depends on field ;)22:03
teK_yes.22:03
teK_still :)22:03
jaegerWe already have depends on, I'm confused22:03
teK_as a comment22:03
teK_and just an example22:03
jaegeroh, you want to convert it to a field?22:03
teK_I understand that it's parsed by prt-get and not pkgmk which is shell script and doesnt need it.. I am using it for illustration22:04
teK_no not 'want'22:04
teK_but if I could pick I'd leave source the way it is and make depends a field22:04
jaegerI must have missed some other conversation22:05
jaegerthis sounds like 2 unrelated things22:05
pedjajaeger: it was either the symlink, or this http://816622275f370fd6.paste.se/22:06
pedjaI can't remember, tbh22:06
jaegerpedja: I tried both, heh22:06
jaegerI must have missed something else, I guess22:06
jaegerwill give it another look22:06
teK_I am using the depends field as illustration in the discussion about source() right now and only now ;)22:07
pedjago is ...weird...22:07
jaegerteK_: I guess I don't see the correlation22:07
teK_thats my guess too :D22:08
jaegerthe github source issue has nothing to do with depends, it's simply bad source archive naming22:08
teK_sorry, my bad22:08
jaegerouch22:08
teK_yeah it doesnt22:08
teK_BUT22:08
teK_the source field should not be modified so the Pkgfile format stays stable (and less ugly)22:08
teK_so IF I would give up the stability of the format for only field I'd always chose depends on over every other field (and thus source)22:09
jaegerI don't feel like a behind-the-scenes fix for github urls is clean22:09
teK_define clean ;-)22:09
jaegerit's a hack for a single site and not obvious from the perspective of the Pkgfile22:09
teK_the changes are easy to implement either way so we could always remove the GH fix and opt for the second option IF more sites would pop up having this problem (which wont I guess)22:10
teK_true22:10
jaegerI don't mind if Pkgfile formats change a little over time, myself. It's not like we do it for no reason22:10
jaegerI'd even be fine with depends turning into an array like source, if that's needed22:11
teK_I mind it more than you do, obviously ;)22:11
jaegerIndeed :)22:11
teK_it's too minor in importance imho22:11
teK_again pkgmk is bash and sources the Pkgfile, so prt-get still needs to parse the depends manually in one or another way22:11
jaegerI'd classify it as a "quality of life" change more than anything else. It's not required.22:14
teK_which change?22:14
teK_the syntax in source()?22:14
jaegerThe source change22:14
teK_then I disagree22:14
teK_:-)22:14
teK_things _will_ break once two ports have the same file name (here: version)22:15
jaegerYes, but that's easily fixed by the maintainer22:15
teK_read break: md5sum mismatch as long as you don't rm the source file before building port #2 with the same name22:15
teK_by using the syntax? Yeah.. but its still not a quality of life fix but a (very unlikely) fix b0rkage for people fix :]22:16
jaegerno, by renaming the source archive22:16
teK_but it doesnt matter, we'll have to opt for one route and push thi22:16
teK_s22:16
teK_yeah renaming by using the new syntax22:17
jaegerNo, by renaming the source archive itself.22:17
teK_and self host?22:17
jaegerYes22:17
teK_:)22:17
teK_of course22:17
jaegerI did not say this is ideal, simply that it's easy to work around22:18
teK_fucking GH22:18
teK_-_-22:18
jaegerTo be clear, I very much would like to see this issue fixed22:18
teK_so much work / discussion for shit22:18
jaegerI just prefer a way that works for everything22:18
teK_I get that22:18
frinnstI dont think we should spend time working around github crap22:22
frinnsti'd rather se a crux-wide github ban :)22:22
teK_:)))22:22
frinnstis it possible to download a tarball from github with wget/curl and get a normal filename?22:24
frinnstif you use a browser it seems to work fine - javascript voodoo?22:24
teK_you sure?22:25
frinnstno, but iirc22:26
frinnsthttps://github.com/fontforge/libspiro/archive/0.4.20150701.tar.gz - libspiro-0.4.20150701.tar.gz22:28
teK_those are sick sick people22:28
jaegerIt does some odd shit with a header22:30
frinnsthttp://www.zoobab.com/github-sucks-wget-unfriendly22:30
jaeger$ curl -sL --head https://github.com/docker/docker/archive/v1.11.1.tar.gz | grep -i disposition22:30
jaegerContent-Disposition: attachment; filename=docker-1.11.1.tar.gz22:30
teK_parse all the headers \o/22:30
jaegerThe Content-Disposition response-header field has been proposed as a means for the origin server to suggest a default filename if the user requests that the content is saved to a file. This usage is derived from the definition of Content-Disposition in RFC 1806 [35].22:31
jaegercurl has -J to try to use that22:32
jaegerwget has --content-disposition which is marked experimental22:32
teK_nice22:32
jaegerI wonder if those are safe to use by default22:33
jaegerThere's still the issue of the name being different in source=() vs. the downloaded file22:33
teK_true22:33
teK_meow22:33
teK_if there's a way for pkgmk to check if there was renaming going on because of the header it could adjust the file name in the array22:35
teK_which would be a general fix ...22:36
teK_but tricky maybe :]22:36
teK_and things break if the server forgets to send the header22:39

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