Re: [ptrace] please review follow fork/exec changes

Well, that didn't work... Not sure why since it broke existing gdb. My guess is we're not getting the exec stops we used to get. Might have to wait till tomorrow to get more details.


On 02/07/2012 12:45 PM, Dmitry Mikulin wrote:
>
>
>> So, do you in fact need to distinguish exec stops from syscall exit
>> against exec stops from PT_FOLLOW_EXEC,
>
> This is pretty much what I need. It's the same stop in syscall return right? I don't want to change when the stop happens, I want to have an lwpinfo flag that tells me when a stop occurred in a process under PT_FOLLOW_EXEC.
>
>> _at__at_ -889,7 +890,9 _at__at_ exec_fail_dealloc:
>>
>>       if (error == 0) {
>>           PROC_LOCK(p);
>> -        td->td_dbgflags |= TDB_EXEC;
>> +        if ((p->p_flag&  P_TRACED) != 0&&
>> +            ((P_FOLLOWEXEC) != 0 || (p->p_stops&  S_PT_SCX) != 0))
>> +            td->td_dbgflags |= TDB_EXEC;
>>           PROC_UNLOCK(p);
>
> There's a small bug in the patch that makes it not work. The check for P_FOLLOWEXEC should be:
>
> +        ((p->p_flag&  P_FOLLOWEXEC) != 0 || (p->p_stops&  S_PT_SCX) != 0))
>
>
> Looks like the patch should work for me but I need to verify.