| 25 Nov 2024 |
 FliegendeWurst | Did anyone see this kind of build failure for openjdk8 before?
zip I/O error: No such file or directory
zip error: Could not create output file (was replacing the original zip file)
make[2]: *** [CreateJars.gmk:659: /build/source/build/linux-x86_64-normal-server-release/images/src.zip] Error 1
(it occurs on the staging branch)
| 11:49:00 |
 emily | you're probably just gonna have to bisect it | 11:50:30 |
| emily | have you tested with .override { stdenv = gcc13Stdenv; } though? | 11:50:37 |
| FliegendeWurst | In reply to @emilazy:matrix.org
have you tested with .override { stdenv = gcc13Stdenv; } though? yes, same error | 14:02:44 |
| emily | excitin | 14:02:59 |
| emily | * exciting | 14:03:00 |
| emily | was the zip package touched? | 14:03:07 |
| FliegendeWurst | in May | 14:03:32 |
| emily | hm | 14:04:21 |
| emily | it might be bisect time :) | 14:04:23 |
| emily | git bisect start --first-parent staging && git bisect run nix build -f . openjdk8, say | 14:05:00 |
| emily | (and be prepared to skip commits…) | 14:05:10 |
| FliegendeWurst | yeah, but it will be a bit annoying for me. I built my prev. staging iterations with strictDepsByDefault | 14:05:55 |
| FliegendeWurst | so I either constantly backport those fixes, or rebuild everything without the setting | 14:06:08 |
| FliegendeWurst | then again, a bisect on staging will rebuild everything very often anyway | 14:06:19 |
| emily | are there strictDeps fixes in the pipeline? :) | 14:06:46 |
| emily | or already merged? | 14:06:48 |
| FliegendeWurst | yes https://github.com/NixOS/nixpkgs/pull/355844
not merged | 14:07:00 |
| FliegendeWurst | some are a bit fishy, like including wayland-scanner in buildInputs only to please the ./configure script | 14:07:56 |
| emily | anyway, it's not that bad, since bisection is O(log n) | 14:07:57 |
| FliegendeWurst | oh good news! | 14:09:17 |
| FliegendeWurst | jdk8.override { stdenv = gcc13Stdenv; zip = zip.override { stdenv = gcc13Stdenv; }; } builds | 14:09:19 |
| FliegendeWurst | (I bet you could remove the first override) | 14:09:30 |
| emily | oh wonderful | 14:16:44 |
| emily | it's probably exposing some UB | 14:16:51 |
| emily | check Arch/Fedora/Gentoo/… for patches | 14:17:09 |
 chayleaf | In reply to @tomodachi94:matrix.org
Wondering if we should drop the programs.java NixOS module, since it's impure and only has two in-tree consumers (packages should make a wrapper that calls jdk/jre, developers should be using devShells) it doesn't seem completely useless as it also sets JAVA_HOME (adding Java to environment.systemPackages does seem like something you'd want to do sometimes), but imo if we were to remove it it's better to start by deprecating or completely removing the binfmt part and seeing if anyone complains, as it's the only non-trivial part of the module | 16:31:36 |
 Tomodachi94 (they/them) | In reply to@tomodachi94:matrix.org
More information about my plans to rip out ant-contrib from the main ant package: https://github.com/NixOS/nixpkgs/issues/357758 ....all of these packages still build if I rip out ant-contrib, even though some of them definitely have targets that use its tasks | 19:08:00 |
| Tomodachi94 (they/them) | Source for the latter: rg -Lz --glob "build.xml" '<(foreach|antcallback|antfetch|assert|foreach|for|if|outoutdate|runtarget|switch|throw|timestampselector|trycatch|post|stopwatch|osfamily|shellscript|math|propertycopy|propertyselector|pathtofileset|propertyregex|sortlist|urlencode|var|forget|limit|compilewithwalls|antclipse|inifile|verifydesign)\b' | less | 19:09:00 |
| Tomodachi94 (they/them) | Source for the latter: rg -Lz --glob "build.xml" '<(foreach|antcallback|antfetch|assert|foreach|for|if|outoutdate|runtarget|switch|throw|timestampselector|trycatch|post|stopwatch|osfamily|shellscript|math|propertycopy|propertyselector|pathtofileset|propertyregex|sortlist|urlencode|var|forget|limit|compilewithwalls|antclipse|inifile|verifydesign)\b' | less over all srcs of Ant dependents | 19:09:24 |
