| 29 Jan 2026 |
 Atemu | Also, while it'd be nice, cross is still rather niche compared to native aarch64. The point of source bootstrap is to reduce the trust necessary for users. I'd first want native bootstrap on the platforms that people actually use. | 09:34:23 |
 K900 | I mean cross as default bootstrap path | 10:54:47 |
| K900 | Instead of relying on bootstrap-tools | 10:54:51 |
 Randy Eckenrode | (Catching up.) Re: bootstrapping from source. Swift currently does (when it builds at all). The rewrite and update to 6.2 will preserve source-based bootstrapping. | 16:21:55 |
| Randy Eckenrode | I have plans for Darwin’s stdenv, but they’re a low priority behind other things. | 16:51:05 |
 dish [Fox/It/She] | are current swift builds using binary bootstrapping, or have I misunderstood you? | 16:51:19 |
| dish [Fox/It/She] | just curious | 16:51:23 |
| Randy Eckenrode | Swift is source-bootstrapped. | 19:08:45 |
| Randy Eckenrode | Swift 5.10 can (more or less) do that out of the box. For Swift 6.2, there is a barely working compiler written in C++ that you can build. | 19:11:40 |
| Randy Eckenrode | For Swift 6.2, I build that then the minimal compiler then the final one. The bootstrapping process has been reworked to be managed in Nix. | 19:12:19 |
 Sergei Zimmerman (xokdvium) | Have there been attempts at having RELR enabled by default on supported systems?
Apparently it has some nice space savings https://maskray.me/blog/2021-10-31-relative-relocations-and-relr | 20:05:30 |
| 30 Jan 2026 |
| Atemu | So, from everything built but gccgo and dependents, it took my 7950x3D ~20min to reach go and I wager most of that was gccgo because a single go compiler takes ~40s | 02:24:39 |
| Atemu | So I think it's safe to say that long compile time for the entire bootstrap chain is unlikely to become an issue | 02:25:31 |
| dish [Fox/It/She] | appreciate the check on that | 04:01:17 |
| dish [Fox/It/She] | glad to confirm its not really an issue | 04:01:23 |
|  zimward changed their display name from zimward to zimward @fosdem. | 16:18:50 |
 aleksi | This could be interesting. Maybe for the non-natively bootstrapped platforms, we could slightly modify the late stages of minimal-bootstrap such that the final gcc is optionally a cross compiler (build=x86_64, host=target=stdenv.buildPlatform) and minimal-bootstrap.*-static would be cross-compiled with it | 19:15:48 |
| 1 Feb 2026 |
|  Fernando Rodrigues changed their display name from SigmaSquadron to Fernando Rodrigues. | 10:41:58 |
| zimward changed their display name from zimward @fosdem to zimward. | 20:36:13 |
| 4 Feb 2026 |
|  tiferrei joined the room. | 02:42:41 |
| tiferrei left the room. | 14:28:34 |
| 5 Feb 2026 |
 Guilherme Lima | Redacted or Malformed Event | 13:39:00 |
| Guilherme Lima | Redacted or Malformed Event | 13:39:41 |
| K900 | The better channel is probably #Nix / NixOS | 13:40:23 |
| K900 | And the answer is there is some delay | 13:40:29 |
| Guilherme Lima | Thanks | 13:40:58 |
| aleksi | There's a PoC bootstrap for aarch64-linux now: https://github.com/NixOS/nixpkgs/pull/487398
To make this work, I had to tell gcc-15 that the host is musl-based. But in fact this breaks building glibc on i686 (but not on x86_64). I think this might be the frankenstein gcc problem? So I guess we need a glibc already built before we build a proper gcc targeting glibc | 21:03:51 |
 Tristan Ross | Awesome, feel free to add me as a reviewer once its ready. | 21:09:34 |
| Tristan Ross | I'd really like to try this out. | 21:09:41 |
