On 16.09.2020 18:08, Tom Tromey wrote:
>>>>>> "Kamil" == Kamil Rytarowski <n54@gmx.com> writes:
> 
> While switching the target wait flags to be a flag enum type, I found a
> small problem in the new NetBSD gdbserver work.
> 
> Kamil> +/* Implement a safe wrapper around waitpid().  */
> Kamil> +
> Kamil> +static pid_t
> Kamil> +netbsd_waitpid (ptid_t ptid, struct target_waitstatus *ourstatus, int options)
> 
> Here options is either 0 or:
> 
> #define TARGET_WNOHANG 1
> 
> Kamil> +{
> Kamil> +  int status;
> Kamil> +
> Kamil> +  pid_t pid = gdb::handle_eintr (::waitpid, ptid.pid (), &status, options);
> 
> ... but it is passed directly to ::waitpid.  This is fine for now if the
> system WNOHANG happens to be 1.  However, I think it would be better not
> to rely on this.  For one thing, there's no guarantee that the value of
> TARGET_WNOHANG will never change.
> 
> Kamil> +  pid_t pid = netbsd_waitpid (ptid, ourstatus, target_options);
> 
> Here's a spot making the call; this shows that the target option is
> being passed untranslated.
> 

I see. It happens that WNOHANG is equal to TARGET_WNOHANG on NetBSD and
it went unnoticed.

> thanks,
> Tom
>