Re: [PATCH 0/3] Fix "attach" infinite loop

[Luis Machado <luis.machado@arm.com>]

On 12/18/23 14:30, Tom Tromey wrote:
> Luis> Attaching to program: /binutils-gdb-arm64-focal/gdb/testsuite/outputs/gdb.threads/attach-many-short-lived-threads/attach-many-short-lived-threads, process 1545611
> Luis> Cannot attach to lwp 1609323: Operation not permitted (1)
> 
> Luis> Are we missing XFAIL-ing the test when we see the ptrace attach failure?
> 
> I see this in the test case:
> 
>     -re "warning: Cannot attach to lwp $decimal: Operation not permitted" {
>         # On Linux, PTRACE_ATTACH sometimes fails with
>         # EPERM, even though /proc/PID/status indicates
>         # the thread is running.
>         set eperm 1
>         exp_continue
>     }
> 
> However, this changed from a warning to an error.

Ah, I failed to spot that.

> 
> Could you try removing the 'warning: ' text and see if that helps?  It
> seems like it should issue an xfail instead.

I played with it for a bit, and it seems the testcase is somewhat broken.

It seems to assume everything is OK if it sees the message "Attaching to program: .*", which explains
the PASS we get from the above test.

If we ever see the EPERM error, we will set the eperm variable, but it will not matter, because the
matching for the attachment message will gobble things all the way to the gdb prompt, and will issue
a PASS, resuming the test (assuming there are live threads) and failing.

I think the testcase needs a bit of a rework to bail out if we fail to attach to the process. I'll
give it a go since it reproduces easily for me.