| 14 Jan 2024 |
 raitobezarius | I would recommend not focusing too much on the specifics and trying to find a path | 21:37:22 |
| raitobezarius | then, you can easily improve your metadata quality | 21:37:28 |
 trofi | As long as you pick a tuple it should be fine :) | 21:37:28 |
| raitobezarius | otherwise you are trying to solve two problems at the same time | 21:37:35 |
 samueldr | that's part of what I hinted at earlier: I don't even know what I'm supposed to be doing :) | 21:37:58 |
| raitobezarius | pick any quadruplet that you like or a triplet, modify the parser to make it accepted, then move on to lib/systems/examples.nix | 21:38:38 |
| raitobezarius | you can do btron = { config = "i386-unknown-unknown-btron"; libc = "btron"; ... } | 21:38:49 |
| raitobezarius | there | 21:38:50 |
| raitobezarius | and then you have pkgsCross.btron.hello in theory | 21:38:57 |
| raitobezarius | (modulo the cross stdenv chooser for cc of course) | 21:39:05 |
| raitobezarius | (you need to wire your GCC 2.X there) | 21:39:09 |
| raitobezarius | but yeah if you have a branch, I can easily write patches over it and show you how I see things | 21:39:39 |
| samueldr | the gcc is unmodified, and hints I found yesterday points at versions more recent than our oldest still working to produce binaries | 21:41:02 |
| samueldr | the only important part is the libc | 21:41:06 |
| samueldr | and I verified that gcc is unmodified, the tarball, once extracted, matches entirely (except docs) to the 2.95.2 checkout | 21:41:45 |
| samueldr | I think knowing that "it's 'okay'" to force a triplet by hacking together Nixpkgs helps me here | 21:42:24 |
| trofi | gentoo had a somewhat working gcc-2.95.3 to be buildable against a modern toolchain as of 7 years ago, a fix or two might be useful: https://github.com/gentoo/gcc-patches/tree/master/2.95.3/gentoo | 21:43:04 |
| raitobezarius | the libc rerouting can be done by touching the lib/systems/examples.nix and the big switch case we have seen yesterday I believe, is there any other concern you had over libc? | 21:43:11 |
| samueldr | I'll know at this point in time | 21:43:31 |
| samueldr | trofi you missed it yesterday, I think, but I have a hacked-up "improper" setup building their samples already | 21:44:21 |
| trofi | aha, nice! | 21:44:33 |
| samueldr | but it's not relying on Nixpkgs to do cross | 21:44:35 |
| samueldr | it's "just" the packages and some environmental hacking to make it work like their "SDK" does | 21:45:07 |
| samueldr | it's relatively straightforward, I think | 21:45:16 |
| trofi | nod | 21:45:21 |
| samueldr | I meant, their SDK is straightforward | 21:45:35 |
| trofi | gcc before 3.3 had a slightly different bin directory layout from cross-compiling. i wonder if you had to deal with it. | 21:45:38 |
| samueldr | http://www.chokanji.com/developer/download.html | 21:46:41 |
| samueldr | knowing that, I understand that I was observing | 21:47:17 |
| samueldr | and I had to deal with it, but it wasn't that much of a deal, since I had their SDK to ape | 21:47:31 |
