LLVM 14 code gen failure

[nhrrsn]

Context

• Operating System & Odin Version:
  Odin: dev-2025-05:843648c81
  OS: Pop!_OS 22.04 LTS, Linux 6.6.6-76060606-generic
  CPU: 11th Gen Intel(R) Core(TM) i7-11800H @ 2.30GHz
  RAM: 64033 MiB
  Backend: LLVM 14.0.0

Expected Behavior

No errors

Current Behavior

When compiling from the current master branch I get an LLVM code gen failure when calling certain Odin functions.
This error doesn't occur when using release branch dev-2025-04

LLVM CODE GEN FAILED FOR PROCEDURE: raylib::Vector3Angle
define float @"raylib::Vector3Angle"({ <2 x float>, float } %0, { <2 x float>, float } %1) {
decls:

Full error below.

I have seen this error in other places as well. For example when calling net::pack_dns_header

This is specifically an issue with llvm version 14 as I updated to llvm 20.1.7 and the error stopped happening.

Failure Information (for bugs)

Steps to Reproduce

I built Odin from master using make release-native
I made a demo branch containing the Odin binary and a test that fails with the below error when I run ./odin test ./tests/vendor/raylib/

Failure Logs

LLVM CODE GEN FAILED FOR PROCEDURE: raylib::Vector3Angle
define float @"raylib::Vector3Angle"({ <2 x float>, float } %0, { <2 x float>, float } %1) {
decls:
  %2 = alloca { <2 x float>, float }, align 8
  %3 = alloca [3 x float], align 16
  %4 = alloca { <2 x float>, float }, align 8
  %5 = alloca [3 x float], align 16
  %6 = alloca { <2 x float>, float }, align 16
  %7 = alloca { <2 x float>, float }, align 16
  br label %entry

entry:                                            ; preds = %decls
  store { <2 x float>, float } %0, { <2 x float>, float }* %2, align 8
  call void @llvm.memcpy.inline.p0i8.p0i8.i64([3 x float]* %3, { <2 x float>, float }* %2, i64 12, i1 false)
  %8 = load [3 x float], [3 x float]* %3, align 4
  store { <2 x float>, float } %1, { <2 x float>, float }* %4, align 8
  call void @llvm.memcpy.inline.p0i8.p0i8.i64([3 x float]* %5, { <2 x float>, float }* %4, i64 12, i1 false)
  %9 = load [3 x float], [3 x float]* %5, align 4
  %10 = bitcast { <2 x float>, float }* %6 to [3 x float]*
  store [3 x float] %8, [3 x float]* %10, align 4
  %11 = load { <2 x float>, float }, { <2 x float>, float }* %6, align 8
  %12 = bitcast { <2 x float>, float }* %7 to [3 x float]*
  store [3 x float] %9, [3 x float]* %12, align 4
  %13 = load { <2 x float>, float }, { <2 x float>, float }* %7, align 8
  %14 = call float @"linalg::vector_angle_between:proc\22contextless\22(a:[3]f32,b:[3]f32)->(:f32)"({ <2 x float>, float } %11, { <2 x float>, float } %13)
  ret float %14
}




Call parameter type does not match function signature!
  %3 = alloca [3 x float], align 16
 i8*  call void @llvm.memcpy.inline.p0i8.p0i8.i64([3 x float]* %3, { <2 x float>, float }* %2, i64 12, i1 false)
Call parameter type does not match function signature!
  %5 = alloca [3 x float], align 16
 i8*  call void @llvm.memcpy.inline.p0i8.p0i8.i64([3 x float]* %5, { <2 x float>, float }* %4, i64 12, i1 false)