[llvm-bugs] [Bug 49376] LLVM 12 regression: emits bogus inline assembly error without proper error handling

via llvm-bugs llvm-bugs at lists.llvm.org
Tue Mar 2 14:15:39 PST 2021


https://bugs.llvm.org/show_bug.cgi?id=49376

Andrew Kelley <andrew at ziglang.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |INVALID

--- Comment #7 from Andrew Kelley <andrew at ziglang.org> ---
Downstream, Michael Dusan diagnosed this as a problem on the Zig side. With
LLVM12, we revamped our CPU features integration with LLVM, and introduced zig
awareness of 32bit-mode and 16bit-mode. However, the logic for what CPU
features to pass to LLVM explicitly sets every flag. These mode flags operate
as "overrides" of what is determined by architecture and by environment. Zig
CPU feature detection did not include setting these features to defaults.

In summary, Zig regressed by passing to LLVM `-32bit-mode` (incorrectly
disabling 32 bit mode for i386). This is fixed downstream in
https://github.com/ziglang/zig/commit/68c00122b10a7364c52a59f56761942525498ce8.

Thank you for your time on this one. I apologize for the false alarm.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20210302/e1ef8035/attachment.html>


More information about the llvm-bugs mailing list