Re: [PATCH v1 2/2] gdb: rely on the first non-exited thread TPID when reading Linux procfs files

On 20/08/2026 16:48, Simon Marchi wrote:
> On 7/28/26 10:33 AM, Matthieu Longo wrote:
>> diff --git a/gdb/inferior.h b/gdb/inferior.h
>> index 5c7a52319e7..91dc4e636fd 100644
>> --- a/gdb/inferior.h
>> +++ b/gdb/inferior.h
>> @@ -513,6 +513,17 @@ class inferior : public refcounted_object,
>>    /* Find (non-exited) thread PTID of this inferior.  */
>>    thread_info *find_thread (ptid_t ptid);
>>  
>> +  /* Return the first non-exited thread of this inferior.
>> +
>> +     This is only a best-effort choice of a thread that is expected to still
>> +     exist in the target.  A thread may have exited after GDB last updated its
>> +     thread list, or GDB/gdbserver may have observed the exit but not yet
>> +     propagated it through all layers.  Therefore the returned thread is not
>> +     guaranteed to still be alive when it is later accessed.  This avoids the
>> +     common case where the current thread has exited, but callers must still be
>> +     prepared for the selected thread to no longer exist.  */
>> +  ptid_t first_non_exited_thread () const;
> 
> I think it would be more consistent with the other "find thread" methods
> to return the `thread_info *`
> 

Makes sense. Fixed.

>> diff --git a/gdb/linux-tdep.c b/gdb/linux-tdep.c
>> index 8c53ffd5e89..9bdcc55a0e1 100644
>> --- a/gdb/linux-tdep.c
>> +++ b/gdb/linux-tdep.c
>> @@ -457,6 +457,44 @@ linux_has_shared_address_space (struct gdbarch *gdbarch)
>>    return linux_is_uclinux ();
>>  }
>>  
>> +/* Return a PTID that identifies the current process and can be used to
>> +   access procfs safely.
>> +
>> +   The returned PTID is that of the thread-group leader whenever it is
>> +   still alive.  If the leader has already exited, the PTID of the first
>> +   non-exited thread in the current inferior is returned instead.
>> +   This ensures that the returned PTID always refers to a live thread
>> +   whose procfs entries are present and populated.  */
>> +static ptid_t
>> +get_process_reference_ptid (bool verbose = false)
>> +{
>> +  /* Get the current thread.  */
>> +  thread_info *thr = inferior_thread ();
>> +
>> +  /* Construct the PTID of the thread-group leader.  On Linux,
>> +     the leader's LWP ID is equal to the process ID.  */
>> +  ptid_t leader_ptid (thr->ptid.pid (), thr->ptid.pid ());
>> +
>> +  /* Use the thread-group leader if it is still alive.  Otherwise, use
>> +     the first thread that has not exited.  */
>> +  thread_info *leader_thr
>> +    = current_inferior ()->find_thread (leader_ptid);
>> +  ptid_t ptid = (leader_thr == nullptr
>> +		 ? current_inferior ()->first_non_exited_thread ()
>> +		 : leader_ptid);
>> +
>> +  if (!verbose)
>> +    return ptid;
>> +
>> +  if (leader_thr != nullptr)
>> +    gdb_printf (_("process %d\n"), leader_ptid.pid ());
>> +  else
>> +    gdb_printf (_("process %d [Note: information where gathered from LWP %ld " \
>> +		  "as the thread-group leader (LWP=%ld) already exited.]\n"),
>> +		ptid.pid (), ptid.lwp (), leader_ptid.lwp ());
> 
> I don't think the printfs should be part of this getter function.  Move
> it to a separate function called by whoever needs it.  Or just inline
> it, it looks like there is just a single caller that needs it
> (linux_info_proc).
> 

The idea is to inform the user if a fallback occurs.
Now, if I move the printing to another function, I don't have the fetched information anymore, and
cannot figure out easily if a fallback happened and using which thread.

It is ugly but I could not find better.
Do you have a better approach ?

> And really (IIUC), this function only really needs to return a tid (the
> id of one LWP), not a ptid.

Yes, at one exception in linux_fill_prpsinfo() where it needs the PID too.

> 
> Simon

Matthieu