add generic x86_32 support #52634
add generic x86_32 support #52634
Conversation
- Add support for i386-i586. - Add `isx86_32` predicate that can replace most uses of `isi686`. - `isi686` is reinterpreted to mean "exactly i686 arch, and not say i585 or i386". - This branch was used to build working i586 kernel running on i586 hardware.
goertzenator
requested review from
Ericson2314,
matthewbauer and
nbp
as code owners
GrahamcOfBorg
added
6.topic: stdenv
|
This pull request has been mentioned on Nix community. There might be relevant details there: https://discourse.nixos.org/t/i586-cross-compiling-ambiguous-i686-use/1482/8 |
lib/systems/doubles.nix
Outdated
| @@ -12,11 +12,14 @@ let | |||
| "mipsel-linux" | |||
|
|
|||
| "i686-cygwin" "i686-freebsd" "i686-linux" "i686-netbsd" "i686-openbsd" | |||
| "i586-cygwin" "i586-freebsd" "i586-linux" "i586-netbsd" "i586-openbsd" | |||
| "i486-cygwin" "i486-freebsd" "i486-linux" "i486-netbsd" "i486-openbsd" | |||
| "i386-cygwin" "i386-freebsd" "i386-linux" "i386-netbsd" "i386-openbsd" | |||
There was a problem hiding this comment.
i386 doesn't work with latest glibc - better leave it off
There was a problem hiding this comment.
Should I purge all i386 support from the PR or just this i386 line?
There was a problem hiding this comment.
Just the i386-linux references, I think. Although I'm not sure whether other OSs actually still support i386. Very few computers still even exist that are i386 so it's largely a irrelevant 😄
There was a problem hiding this comment.
I've followed code to see what doubles.nix actually does, and... it does almost nothing. Only mesaPlatforms appears to be used AFAICT. Anyway I will revert all changes to this file and also for-meta.nix (which looks 100% dead).
lib/systems/inspect.nix
Outdated
| @@ -8,8 +8,12 @@ let abis = lib.mapAttrs (_: abi: builtins.removeAttrs abi [ "assertions" ]) abis | |||
|
|
|||
| rec { | |||
| patterns = rec { | |||
| isi386 = { cpu = cpuTypes.i386; }; | |||
There was a problem hiding this comment.
Do you expect needing these? My thinking would be we would just use x86_32 as much as possible. We are doing a similar thing with isAarch32 as well. Even without these you can always do:
stdenv.hostPlatform.parsed.cpu.name == "i386"
if you really need a condition on a certain micro architecture.
There was a problem hiding this comment.
I don't. I put them in to be symmetric with isi686.
- Remove changes to dead code in `doubles.nix` and `for-meta.nix`. - Remove `isi[345]86` predicates since other cpu families don't have specific model predicates.
| isi686 isx86_64 is64bit isAarch32 isAarch64 isMips isBigEndian; | ||
| isi386 isi486 isi586 isi686 isx86_32 isx86_64 | ||
| is32bit is64bit | ||
| isAarch32 isAarch64 isMips isBigEndian; |
There was a problem hiding this comment.
FWIW, I was trying to not grow this in preference of people using stdenv.hostPlatform.is*. Let others decide whether that's a good goal. It's certainly not our worst source of duplication.
There was a problem hiding this comment.
Also sorry I wrote this (probably on my phone) and then never submitted it.
There was a problem hiding this comment.
Feel free to remove them. At some point we should probably just start removing any stdenv.is* that is unused.
Motivation for this change
Add support for i386-i586. Ref:
In summary:
isx86_32predicate that can replace most uses ofisi686.isi686is reinterpreted to mean "exactly i686 arch, and not say i585 or i386".Things done
sandboxinnix.confon non-NixOS)nix-shell -p nox --run "nox-review wip"./result/bin/)nix path-info -Sbefore and after)