1.55.0 new test debuginfo/function-names.rs fails on all architectures in Debian

[infinity0]

Full log, warning very large file: https://buildd.debian.org/status/fetch.php?pkg=rustc&arch=amd64&ver=1.55.0%2Bdfsg1-1%7Eexp1&stamp=1633891056&raw=0

------------------------------------------
NOTE: compiletest thinks it is using GDB version 10001090
executing "/usr/bin/gdb" "-quiet" "-batch" "-nx" "-command=/<<PKGBUILDDIR>>/build/aarch64-unknown-linux-gnu/test/debuginfo/function-names.gdb/function-names.debugger.script"
------stdout------------------------------
GNU gdb (Debian 10.1-2) 10.1.90.20210103-git
[..]

File /<<PKGBUILDDIR>>/src/test/debuginfo/function-names.rs:
92:     static fn function_names::main();
109:    static fn function_names::main::{closure#0}(*mut function_names::main::{closure#0});
113:    static fn function_names::main::{generator#1}(core::pin::Pin<&mut function_names::main::{generator#1}>, ()) -> core::ops::generator::GeneratorState;

File /<<PKGBUILDDIR>>/src/test/debuginfo/function-names.rs:
192:    static fn function_names::generic_func::{closure#0}<i32>(*mut function_names::generic_func::{closure#0});
190:    static fn function_names::generic_func<i32>(i32) -> i32;

File /<<PKGBUILDDIR>>/src/test/debuginfo/function-names.rs:
158:    static fn function_names::GenericStruct<i32, i32>::impl_function<i32, i32>();
146:    static fn function_names::Mod1::TestStruct2::impl_function();
130:    static fn function_names::TestStruct1::impl_function();
160:    static fn function_names::{impl#2}::impl_function::{closure#0}<i32, i32>(*mut function_names::{impl#2}::impl_function::{closure#0});

File /<<PKGBUILDDIR>>/src/test/debuginfo/function-names.rs:
150:    static fn function_names::Mod1::{impl#1}::trait_function();
135:    static fn function_names::{impl#1}::trait_function();
167:    static fn function_names::{impl#3}::trait_function<i32>();
181:    static fn function_names::{impl#5}::trait_function3<function_names::TestStruct1>();
186:    static fn function_names::{impl#6}::trait_function<i32, 1>();

File /<<PKGBUILDDIR>>/src/test/debuginfo/function-names.rs:
198:    static fn function_names::const_generic_fn_bool<false>();
199:    static fn function_names::const_generic_fn_non_int<{CONST#fe3cfa0214ac55c7}>();
200:    static fn function_names::const_generic_fn_signed_int<-7>();
201:    static fn function_names::const_generic_fn_unsigned_int<14>();

------stderr------------------------------

------------------------------------------

error: line not found in debugger output: [...]static fn function_names::generic_func(i32) -> i32;
status: exit status: 0
command: "/usr/bin/gdb" "-quiet" "-batch" "-nx" "-command=/<<PKGBUILDDIR>>/build/aarch64-unknown-linux-gnu/test/debuginfo/function-names.gdb/function-names.debugger.script"
stdout:
------------------------------------------

Note the difference: generic_func(i32) -> i32 expected vs generic_func<i32>(i32) -> i32 actual output.

All the other debuginfo tests are fine so I guess it's not as simple as "Debian is using the wrong GDB version"; CC @dpaoliello and @michaelwoerister who worked on the test.

[Mark-Simulacrum]

IIRC we found recently this test is just not being run in our CI, so it's probably just broken.

[michaelwoerister]

I'm looking into it.

[michaelwoerister]

I opened #89772 to fix this. But we'll need a better long-term solution for dealing with debuginfo tests.

[infinity0]

Is fe3cfa0214ac55c7 endian-independent? Guess we'll see in a few days when I run this on s390x.

[infinity0]

The patch works on big-endian too, test passed on s390x.

[Enselic]

Triage: My understanding is that the failing test was fixed eventually. Let's close this issue as fixed.