Re: [Bug-tar] bad treatment of symlinks

[Andries E. Brouwer]

On Thu, Oct 20, 2011 at 05:08:09PM +0300, Sergey Poznyakoff wrote:

> The changes Paul referred to are quite recent and therefore are
> still in the repository:

Good. Then we'll see them eventually.

> >  tar: foo: implausibly old time stamp 1970-01-01 01:00:00
> >
> > Clearly that is undesirable. One point of view is that it is none of
> > tar's business to complain about time stamps.
> 
> I disagree.  It is definitely tar's business to complain about failed
> disk operations, isn't it?  Now, what happened is that tar attempted to
> restore timestamp of a file it had recently extracted and failed to do so.

You are mistaken. There was no failure anywhere. The original had this
time stamp. The extracted copy had this time stamp. All was perfect.

Tar thinks that it should complain. None of its business.
Still, a complaint may be useful under certain circumstances.
However, half a million identical complaints is not useful or desirable.
On the other hand, it is *really* slow.

> > In a case like this "the right thing" would be to give this message
> > five times, and upon the 6th say "further timestamp warnings suppressed;
> > to see them all, give the --allfoo option, to see none, use --nofoo.]
> 
> That would be OK if such a flood of warnings were considered
> normal, which it is not.  In my opinion, the right thing is to discover
> the reason for that strange behavior and fix it in the first place.
> Could you give us more detail?

Nothing is wrong. Just a tree that far back came from a commercial system
without Unix-type time stamps. You may recognize dates close to 1970-01-01
essentially as "no date available". The "close to" part here is because of
time zones. This is 0 GMT.


Andries