Re: [PATCH] Fix execution_direction's type

[Pedro Alves <palves@redhat.com>]

On 10/09/2015 06:14 PM, Simon Marchi wrote:
> On 15-10-09 11:34 AM, Pedro Alves wrote:
>> @@ -3817,6 +3828,7 @@ fetch_inferior_event (void *client_data)
>>    struct cleanup *old_chain = make_cleanup (null_cleanup, NULL);
>>    struct cleanup *ts_old_chain;
>>    int was_sync = sync_execution;
>> +  enum exec_direction_kind save_exec_dir = execution_direction;
>>    int cmd_done = 0;
>>    ptid_t waiton_ptid = minus_one_ptid;
>>  
>> @@ -3849,7 +3861,7 @@ fetch_inferior_event (void *client_data)
>>       event.  */
>>    target_dcache_invalidate ();
>>  
>> -  make_cleanup_restore_integer (&execution_direction);
>> +  make_cleanup (restore_execution_direction, &save_exec_dir);
>>    execution_direction = target_execution_direction ();
>>  
>>    ecs->ptid = do_target_wait (waiton_ptid, &ecs->ws,
> 
> Isn't it dangerous to record a cleanup with a pointer to a local variable?
> 
> What if something throws the exception is caught in a parent frame?  The
> content of the memory location pointed by the cleanup will be undefined
> at that time, won't it?

No, the cleanups are run _before_ the throwing scope is left (that is,
before longjmp), from within throw_exception.

Thanks,
Pedro Alves