14:57 <beerman> farkuhar: "sed -i "110s/cp -a/cp/" /usr/ports/contrib/openjdk8/Pkgfile" wouldn't that fail with directories? ;)

15:05 <farkuhar> beerman: thanks for the correction.

15:06 <beerman> looking at the log, the user has to do something gravely wrong elsewhere. ant, rstudio, the jdks, all build fine over here and over there https://crux.ster.zone/packages/3.6/logs/

15:11 <beerman> ¯\_(ツ)_/¯ and if that just that the user was smart enough to setup fakeroot, then fails to remove it for one port... oh well..

15:13 <farkuhar> Line 181 of the ant build log posted by belgacem1958 (https://pastebin.com/9U3L1GAA) shows that the pkgmk user was denied permission writing to /root, and that's why the build requirement "antunit" was reported missing.

15:14 <beerman> thats a configuration problem of his system then, not a port failure. also it makes no sense that jdk8-bin installed fine just after he uninstalled jdk16..

15:19 <farkuhar> I agree it's not a port failure. A misconfigured system is possible, but I'm more inclined to think that the upstream release is not respecting the fakeroot environment.

15:20 <beerman> anyway, it trying to access /root isn't a good thing, but unfortunately i haven't found a good way to tell that to ant ;) preferably i'd like to store all that in PKGMK_SOURCE_DIR for persistance in case of a rebuild etc etc

15:21 <beerman> yep, sure, thats true. afaik, ant expects you to be a unprivileged user and you just do your happy coding in java for yourself.

15:22 <farkuhar> stenur suggested using unshare (from util-linux) to enter a separate namespace where root uid/gid has been remapped to the current user, just for building problematic ports like ant.

15:23 <beerman> let a user that stumbles over this with fakeroot setup an even more obscure build environment, sounds sane to me..

15:31 <farkuhar> "unfortunately i haven't found a good way to tell that to ant" ... do they not have a bug tracker where we could report the attempt to create files in /root , even running under fakeroot?

15:31 <beerman> i am sure they do

15:34 <farkuhar> http://ant.apache.org/bugs.html seems like the place to report it. Better they deal with it than we come up with obscure workarounds like unshare(1) or temporarily changing the prt-get makecommand.

15:35 <beerman> i had to check actually, and pfew, ant is dannys port anyway :þ but i had looked at it at some point in the past, i still have to look at it when rstudio does something new too (and they will do something new sometime in the future, away from qtwebengine towards to electron)

15:35 <beerman> true, wish i had more time, right now i am spending some on the 3.7 toolchain

15:37 <beerman> anyway, recently in a ticket the question was raised why new or longterm users leave CRUX. compiling just takes to much time was one problem. but isn't that what you actually sign up for when you choose CRUX..?

15:38 <beerman> i know thats why i am staying. anyway, i am still eying the CI over gitlab possibilities..

15:40 <beerman> we'd need a CRUX docker buildenv and share the artifact in the end. would be nice for selected big offenders... firefox, libreoffice, rust, llvm, openjdks, heck why not gcc/clang etc?

15:45 <beerman> i imagine this could help users like right now in #crux to not do weird stuff until they understand better. weird in that context can't be a good experience, more like "oh no my X crashed (again)"