1.48 beta test failures, debuginfo-gdb 9.2 -> 10.1

[infinity0]

This is with 1.48.0-beta.8, and previously these all passed in 1.47.0. On Debian I see these failures on all architectures. The main difference is we are using gdb 10 now, gdb 9.2 previously. The other debuginfo tests pass fine.

test [debuginfo-gdb] debuginfo/extern-c-fn.rs ... FAILED

error: line not found in debugger output: $1 = [...]"abcd\000"
status: exit code: 0
command: "/usr/bin/gdb" "-quiet" "-batch" "-nx" "-command=/<<PKGBUILDDIR>>/build/x86_64-unknown-linux-gnu/test/debuginfo/extern-c-fn.gdb/extern-c-fn.debugger.script"
stdout:
------------------------------------------
GNU gdb (Debian 10.1-1) 10.1
[..]
Breakpoint 1 at 0x122e: file /<<PKGBUILDDIR>>/src/test/debuginfo/extern-c-fn.rs, line 52.
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".

Breakpoint 1, extern_c_fn::fn_with_c_abi (s=0x555555556081, len=20) at /<<PKGBUILDDIR>>/src/test/debuginfo/extern-c-fn.rs:52
52	    zzz(); // #break
$1 = (*mut u8) 0x555555556081
$2 = 20
$3 = 19
$4 = true
$5 = 20.5
[Inferior 1 (process 25039) exited normally]

test [debuginfo-gdb] debuginfo/generator-objects.rs ... FAILED

error: line not found in debugger output: $1 = generator_objects::main::generator-0 {__0: 0x[...], <<variant>>: {__state: 0, 0: generator_objects::main::generator-0::Unresumed, 1: generator_objects::main::generator-0::Returned, 2: generator_objects::main::generator-0::Panicked, 3: generator_objects::main::generator-0::Suspend0 {[...]}, 4: generator_objects::main::generator-0::Suspend1 {[...]}}}
status: exit code: 0
command: "/usr/bin/gdb" "-quiet" "-batch" "-nx" "-command=/<<PKGBUILDDIR>>/build/x86_64-unknown-linux-gnu/test/debuginfo/generator-objects.gdb/generator-objects.debugger.script"
stdout:
------------------------------------------
GNU gdb (Debian 10.1-1) 10.1
[..]
Breakpoint 1 at 0x137e: file /<<PKGBUILDDIR>>/src/test/debuginfo/generator-objects.rs, line 58.
Breakpoint 2 at 0x13ab: file /<<PKGBUILDDIR>>/src/test/debuginfo/generator-objects.rs, line 60.
Breakpoint 3 at 0x13d8: file /<<PKGBUILDDIR>>/src/test/debuginfo/generator-objects.rs, line 62.
Breakpoint 4 at 0x140f: file /<<PKGBUILDDIR>>/src/test/debuginfo/generator-objects.rs, line 64.
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".

Breakpoint 1, generator_objects::main () at /<<PKGBUILDDIR>>/src/test/debuginfo/generator-objects.rs:58
58	    _zzz(); // #break
$1 = (null)

Breakpoint 2, generator_objects::main () at /<<PKGBUILDDIR>>/src/test/debuginfo/generator-objects.rs:60
60	    _zzz(); // #break
$2 = (null)

Breakpoint 3, generator_objects::main () at /<<PKGBUILDDIR>>/src/test/debuginfo/generator-objects.rs:62
62	    _zzz(); // #break
$3 = (null)
6 7 8

Breakpoint 4, generator_objects::main () at /<<PKGBUILDDIR>>/src/test/debuginfo/generator-objects.rs:64
64	    _zzz(); // #break
$4 = (null)
A debugging session is active.

	Inferior 1 [process 25180] will be killed.

Quit anyway? (y or n) [answered Y; input not from terminal]

test [debuginfo-gdb] debuginfo/issue-57822.rs ... FAILED

error: line not found in debugger output: $2 = issue_57822::main::generator-3 {__0: issue_57822::main::generator-2 {__0: 2, <<variant>>: {[...]}}, <<variant>>: {[...]}}
status: exit code: 0
command: "/usr/bin/gdb" "-quiet" "-batch" "-nx" "-command=/<<PKGBUILDDIR>>/build/x86_64-unknown-linux-gnu/test/debuginfo/issue-57822.gdb/issue-57822.debugger.script"
stdout:
------------------------------------------
GNU gdb (Debian 10.1-1) 10.1
[..]
Breakpoint 1 at 0x12ef: file /<<PKGBUILDDIR>>/src/test/debuginfo/issue-57822.rs, line 51.
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".

Breakpoint 1, issue_57822::main () at /<<PKGBUILDDIR>>/src/test/debuginfo/issue-57822.rs:51
51	    zzz(); // #break
$1 = issue_57822::main::closure-1 (issue_57822::main::closure-0 (1))
$2 = issue_57822::main::generator-2{__0: 2, __state: 0, 0: issue_57822::main::generator-2::Unresumed}
A debugging session is active.

	Inferior 1 [process 25521] will be killed.

Quit anyway? (y or n) [answered Y; input not from terminal]

test [debuginfo-gdb] debuginfo/pretty-huge-vec.rs ... FAILED

error: line not found in debugger output: $2 = &[u8] {data_ptr: [...]"\000", length: 1000000000}
status: exit code: 0
command: "/usr/bin/gdb" "-quiet" "-batch" "-nx" "-command=/<<PKGBUILDDIR>>/build/x86_64-unknown-linux-gnu/test/debuginfo/pretty-huge-vec.gdb/pretty-huge-vec.debugger.script"
stdout:
------------------------------------------
GNU gdb (Debian 10.1-1) 10.1
[..]
Breakpoint 1 at 0x1fef: file /<<PKGBUILDDIR>>/src/test/debuginfo/pretty-huge-vec.rs, line 27.
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".

Breakpoint 1, pretty_huge_vec::main () at /<<PKGBUILDDIR>>/src/test/debuginfo/pretty-huge-vec.rs:27
27	    zzz(); // #break
$1 = Vec(size=1000000000) = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0...}
$2 = &[u8] {data_ptr: 0x7fffbc266010, length: 1000000000}
A debugging session is active.

	Inferior 1 [process 26156] will be killed.

Quit anyway? (y or n) [answered Y; input not from terminal]

[Mark-Simulacrum]

Can you verify that gdb 9.2 still passes with 1.48 beta? (Not sure how easy that is for you)

Regardless, it seems like this is probably too difficult for us to land a fix for in the next few days in time for 1.48, and it does not seem like a release blocker to me.

[infinity0]

@Mark-Simulacrum It's not easy for me to run the whole build - sadly Debian deletes old versions from the main archive when they get updated. However I still have gdb 9.2 locally and I can confirm running a random sample of these tests by hand against rustc 1.48 & gdb 9.2 does appear to produce the correct expected output.

The differences appear minor except that in generator-objects the output is 9.2/expected:non-null vs 10:null which is slightly concerning, so someone with more in-depth knowledge about gdb & rust should look into this further.

[spastorino]

Assigning P-medium as discussed as part of the Prioritization Working Group procedure and removing I-prioritize.

[jesusprubio]

Same errors here:

• Fedora 33
• rustc 1.51.0-dev
• GNU gdb 10.1-2.fc33

failures:
    [debuginfo-gdb] debuginfo/extern-c-fn.rs
    [debuginfo-gdb] debuginfo/generator-objects.rs
    [debuginfo-gdb] debuginfo/issue-57822.rs
    [debuginfo-gdb] debuginfo/pretty-huge-vec.rs

test result: FAILED. 85 passed; 4 failed; 27 ignored; 0 measured; 0 filtered out; finished in 7.56s

I have also confirmed these errors are not happening using v9. To reproduce:

gdb --version     # GNU gdb (GDB) Fedora  10.1-2.fc33
./x.py test          # those 4 tests failed
sudo dnf downgrade -y gdb
gdb --version    # GNU gdb (GDB) Fedora 9.2-7.fc33
./x.py test          # all tests ok

[Apteryks]

I can reproduce also on GNU Guix, even in 1.49 when using GDB 10. The debuginfo tests pass when using GDB 8.2.