Unexpected nulls in debuginfo tests on windows-gnu #129662
Comments
rustbot
added
the
needs-triage
This comment was marked as resolved.
This comment was marked as resolved.
rustbot
added
A-debuginfo
saethlin
removed
the
needs-triage
|
The by-value-non-immediate-argument is #128973 Upon further inspection, I am now confident that the enums test failures are actually a change in layout with targets. Such nonsense is typical of the debuginfo tests. Take this example program, using the enum extracted from the test: enum NamedFields<'a> {
#[allow(dead_code)]
Droid { id: i32, range: i64, internals: &'a isize },
Void
}
fn main() {
let e = NamedFields::Void;
unsafe {
#[cfg(target_pointer_width = "32")]
{
let words = std::mem::transmute::<NamedFields, [std::mem::MaybeUninit<u8>; 16]>(e);
println!("{:?}", words[12].assume_init());
}
#[cfg(target_pointer_width = "64")]
{
let words = std::mem::transmute::<NamedFields, [std::mem::MaybeUninit<u8>; 24]>(e);
println!("{:?}", words[0].assume_init());
}
}
}Which we can run on any target that can print using Miri x86_64-pc-windows-msvc, x86-64-pc-windows-gnu, and x86_64-unknonw-linux-gnu all print 1. So I think the debuggers in the tests are correctly reporting that the discriminant value for |
|
Oh I see!
As far as I can tell, this test has been broken since it was written. You can't rely on field order like this, and indeed as I demonstrated above, fields get reordered differently across targets. The structs and enums also have different field ordering on a single target. |
Rollup merge of rust-lang#129672 - saethlin:enum-debuginfo-tests, r=Mark-Simulacrum Make option-like-enum.rs UB-free and portable Fixes rust-lang#129662 (or, at least the parts of it that aren't rust-lang#128973)
…acrum Make option-like-enum.rs UB-free and portable Fixes rust-lang/rust#129662 (or, at least the parts of it that aren't rust-lang/rust#128973)
…acrum Make option-like-enum.rs UB-free and portable Fixes rust-lang/rust#129662 (or, at least the parts of it that aren't rust-lang/rust#128973)
Since #119229 GDB used on CI is recent enough to run the new tests.
However, there are three failures that have been ignored in #119229:
tests\debuginfo\method-on-enum.rs(happened only on 32-bits)tests\debuginfo\option-like-enum.rs(happened only on 32-bits)tests\debuginfo\by-value-non-immediate-argumentThe text was updated successfully, but these errors were encountered: