Re: [MinGW-dvlpr] Progress report on coreutils issue

[Aaron W. L. <aar...@aa...>]

Keith Marshall wrote:

> These are the only requirements I see, for an open call to update any time 
> stamps; i.e. *only* in the specific cases of *creating* a new file, or 
> *truncating* an existing file, should the time stamps definitely be updated.

In addition, see the section for write() where it requires the mtime update.

> In all other cases the behaviour is unspecified, (which to me means that it 
> is permissable, but not mandatory, to leave the time stamps unchanged, so we 
> cannot rely on any particular behaviour).

Neither SuS nor any other copy of IEEE 1003.1 available to me state that 
"the behavior is unspecified" in this case.  I do not believe that the 
lack of a specific prohibition means that the behavior is unspecified, 
as POSIX does not typically state what a function does not do; that is, 
it states the behavior of functions in the positive, not the negative.

However, I think that an implementation could update atime or mtime on 
open() as an extension to POSIX, which means that it would need to be 
explicitly documented--rather than a silent action, as in the case of 
unspecified behavior.

In any case, I am not aware of any implementations that actually modify 
timestamps at open(), so this is probably mostly moot. :)