Re: [PATCH 2/3] gdb: add some asserts to thread_db_notice_clone

[Keith Seitz <keiths@redhat.com>]

Hi,

On 4/7/26 3:22 AM, Andrew Burgess wrote:
> This commit adds some asserts to thread_db_notice_clone, and reorders
> some code.  This is really just a refactor ahead of the next commit.
> The changes made here are:
> 
>    1. Add an assert that STOPPED is not NULL.  We pass STOPPED as the
>       first argument to thread_from_lwp, which already assumes that the
>       pointer is not NULL.  If STOPPED was ever NULL then we would have
>       hit a segfault in thread_from_lwp long ago.
> 
>    2. Move the thread_from_lwp call relating to the PARENT thread
>       before the call relating to the CHILD thread, and assert that the
>       value we get back is STOPPED.  The thread_from_lwp might also
>       gather thread-db related information about PARENT, so the call is
>       important, but even if PARENT cannot be managed via thread-db,
>       the result we get back should always still be STOPPED (see
>       record_thread for more details).  Moving this call before the
>       call relating to the CHILD thread makes the next commit cleaner.
> 
> There should be no user visible changes after this commit as I believe
> both of the things being asserted have always been true.

I'm no expert at this, but your assumption seems correct as far as I
understand it. So this LGTM.

Reviewed-By: Keith Seitz <keiths@redhat.com>

Keith

> ---
>   gdb/linux-thread-db.c | 13 +++++++++----
>   1 file changed, 9 insertions(+), 4 deletions(-)
> 
> diff --git a/gdb/linux-thread-db.c b/gdb/linux-thread-db.c
> index cd25c55a898..1807352ed76 100644
> --- a/gdb/linux-thread-db.c
> +++ b/gdb/linux-thread-db.c
> @@ -440,12 +440,17 @@ thread_db_notice_clone (ptid_t parent, ptid_t child)
>       return false;
>   
>     thread_info *stopped = linux_target->find_thread (parent);
> -
> -  thread_from_lwp (stopped, child);
> +  gdb_assert (stopped != nullptr);
>   
>     /* If we do not know about the main thread's pthread info yet, this
> -     would be a good time to find it.  */
> -  thread_from_lwp (stopped, parent);
> +     would be a good time to find it.  This should return the same
> +     thread_info as STOPPED, but a side effect of this call is that the
> +     pthread/thread-db information might have been filled in if it was not
> +     already known.  */
> +  thread_info *parent_info = thread_from_lwp (stopped, parent);
> +  gdb_assert (parent_info == stopped);
> +
> +  thread_from_lwp (stopped, child);
>     return true;
>   }
>